2 * Copyright (c) 2008, 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.
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"
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)
137 const struct ip6_hdr *nh;
141 nh = ofpbuf_try_pull(packet, sizeof *nh);
146 nexthdr = nh->ip6_nxt;
148 flow->ipv6_src = nh->ip6_src;
149 flow->ipv6_dst = nh->ip6_dst;
151 tc_flow = get_unaligned_be32(&nh->ip6_flow);
152 flow->nw_tos = ntohl(tc_flow) >> 20;
153 flow->ipv6_label = tc_flow & htonl(IPV6_LABEL_MASK);
154 flow->nw_ttl = nh->ip6_hlim;
155 flow->nw_proto = IPPROTO_NONE;
158 if ((nexthdr != IPPROTO_HOPOPTS)
159 && (nexthdr != IPPROTO_ROUTING)
160 && (nexthdr != IPPROTO_DSTOPTS)
161 && (nexthdr != IPPROTO_AH)
162 && (nexthdr != IPPROTO_FRAGMENT)) {
163 /* It's either a terminal header (e.g., TCP, UDP) or one we
164 * don't understand. In either case, we're done with the
165 * packet, so use it to fill in 'nw_proto'. */
169 /* We only verify that at least 8 bytes of the next header are
170 * available, but many of these headers are longer. Ensure that
171 * accesses within the extension header are within those first 8
172 * bytes. All extension headers are required to be at least 8
174 if (packet->size < 8) {
178 if ((nexthdr == IPPROTO_HOPOPTS)
179 || (nexthdr == IPPROTO_ROUTING)
180 || (nexthdr == IPPROTO_DSTOPTS)) {
181 /* These headers, while different, have the fields we care about
182 * in the same location and with the same interpretation. */
183 const struct ip6_ext *ext_hdr = packet->data;
184 nexthdr = ext_hdr->ip6e_nxt;
185 if (!ofpbuf_try_pull(packet, (ext_hdr->ip6e_len + 1) * 8)) {
188 } else if (nexthdr == IPPROTO_AH) {
189 /* A standard AH definition isn't available, but the fields
190 * we care about are in the same location as the generic
191 * option header--only the header length is calculated
193 const struct ip6_ext *ext_hdr = packet->data;
194 nexthdr = ext_hdr->ip6e_nxt;
195 if (!ofpbuf_try_pull(packet, (ext_hdr->ip6e_len + 2) * 4)) {
198 } else if (nexthdr == IPPROTO_FRAGMENT) {
199 const struct ip6_frag *frag_hdr = packet->data;
201 nexthdr = frag_hdr->ip6f_nxt;
202 if (!ofpbuf_try_pull(packet, sizeof *frag_hdr)) {
206 /* We only process the first fragment. */
207 if (frag_hdr->ip6f_offlg != htons(0)) {
208 if ((frag_hdr->ip6f_offlg & IP6F_OFF_MASK) == htons(0)) {
209 flow->nw_frag = FLOW_NW_FRAG_ANY;
211 flow->nw_frag |= FLOW_NW_FRAG_LATER;
212 nexthdr = IPPROTO_FRAGMENT;
219 flow->nw_proto = nexthdr;
224 parse_tcp(struct ofpbuf *packet, struct ofpbuf *b, struct flow *flow)
226 const struct tcp_header *tcp = pull_tcp(b);
228 flow->tp_src = tcp->tcp_src;
229 flow->tp_dst = tcp->tcp_dst;
230 packet->l7 = b->data;
235 parse_udp(struct ofpbuf *packet, struct ofpbuf *b, struct flow *flow)
237 const struct udp_header *udp = pull_udp(b);
239 flow->tp_src = udp->udp_src;
240 flow->tp_dst = udp->udp_dst;
241 packet->l7 = b->data;
246 parse_icmpv6(struct ofpbuf *b, struct flow *flow)
248 const struct icmp6_hdr *icmp = pull_icmpv6(b);
254 /* The ICMPv6 type and code fields use the 16-bit transport port
255 * fields, so we need to store them in 16-bit network byte order. */
256 flow->tp_src = htons(icmp->icmp6_type);
257 flow->tp_dst = htons(icmp->icmp6_code);
259 if (icmp->icmp6_code == 0 &&
260 (icmp->icmp6_type == ND_NEIGHBOR_SOLICIT ||
261 icmp->icmp6_type == ND_NEIGHBOR_ADVERT)) {
262 const struct in6_addr *nd_target;
264 nd_target = ofpbuf_try_pull(b, sizeof *nd_target);
268 flow->nd_target = *nd_target;
270 while (b->size >= 8) {
271 /* The minimum size of an option is 8 bytes, which also is
272 * the size of Ethernet link-layer options. */
273 const struct nd_opt_hdr *nd_opt = b->data;
274 int opt_len = nd_opt->nd_opt_len * 8;
276 if (!opt_len || opt_len > b->size) {
280 /* Store the link layer address if the appropriate option is
281 * provided. It is considered an error if the same link
282 * layer option is specified twice. */
283 if (nd_opt->nd_opt_type == ND_OPT_SOURCE_LINKADDR
285 if (eth_addr_is_zero(flow->arp_sha)) {
286 memcpy(flow->arp_sha, nd_opt + 1, ETH_ADDR_LEN);
290 } else if (nd_opt->nd_opt_type == ND_OPT_TARGET_LINKADDR
292 if (eth_addr_is_zero(flow->arp_tha)) {
293 memcpy(flow->arp_tha, nd_opt + 1, ETH_ADDR_LEN);
299 if (!ofpbuf_try_pull(b, opt_len)) {
308 memset(&flow->nd_target, 0, sizeof(flow->nd_target));
309 memset(flow->arp_sha, 0, sizeof(flow->arp_sha));
310 memset(flow->arp_tha, 0, sizeof(flow->arp_tha));
316 /* Initializes 'flow' members from 'packet', 'skb_priority', 'tun_id', and
319 * Initializes 'packet' header pointers as follows:
321 * - packet->l2 to the start of the Ethernet header.
323 * - packet->l3 to just past the Ethernet header, or just past the
324 * vlan_header if one is present, to the first byte of the payload of the
327 * - packet->l4 to just past the IPv4 header, if one is present and has a
328 * correct length, and otherwise NULL.
330 * - packet->l7 to just past the TCP or UDP or ICMP header, if one is
331 * present and has a correct length, and otherwise NULL.
334 flow_extract(struct ofpbuf *packet, uint32_t skb_priority, ovs_be64 tun_id,
335 uint16_t ofp_in_port, struct flow *flow)
337 struct ofpbuf b = *packet;
338 struct eth_header *eth;
340 COVERAGE_INC(flow_extract);
342 memset(flow, 0, sizeof *flow);
343 flow->tun_id = tun_id;
344 flow->in_port = ofp_in_port;
345 flow->skb_priority = skb_priority;
352 if (b.size < sizeof *eth) {
358 memcpy(flow->dl_src, eth->eth_src, ETH_ADDR_LEN);
359 memcpy(flow->dl_dst, eth->eth_dst, ETH_ADDR_LEN);
361 /* dl_type, vlan_tci. */
362 ofpbuf_pull(&b, ETH_ADDR_LEN * 2);
363 if (eth->eth_type == htons(ETH_TYPE_VLAN)) {
364 parse_vlan(&b, flow);
366 flow->dl_type = parse_ethertype(&b);
370 if (flow->dl_type == htons(ETH_TYPE_IP)) {
371 const struct ip_header *nh = pull_ip(&b);
375 flow->nw_src = get_unaligned_be32(&nh->ip_src);
376 flow->nw_dst = get_unaligned_be32(&nh->ip_dst);
377 flow->nw_proto = nh->ip_proto;
379 flow->nw_tos = nh->ip_tos;
380 if (IP_IS_FRAGMENT(nh->ip_frag_off)) {
381 flow->nw_frag = FLOW_NW_FRAG_ANY;
382 if (nh->ip_frag_off & htons(IP_FRAG_OFF_MASK)) {
383 flow->nw_frag |= FLOW_NW_FRAG_LATER;
386 flow->nw_ttl = nh->ip_ttl;
388 if (!(nh->ip_frag_off & htons(IP_FRAG_OFF_MASK))) {
389 if (flow->nw_proto == IPPROTO_TCP) {
390 parse_tcp(packet, &b, flow);
391 } else if (flow->nw_proto == IPPROTO_UDP) {
392 parse_udp(packet, &b, flow);
393 } else if (flow->nw_proto == IPPROTO_ICMP) {
394 const struct icmp_header *icmp = pull_icmp(&b);
396 flow->tp_src = htons(icmp->icmp_type);
397 flow->tp_dst = htons(icmp->icmp_code);
403 } else if (flow->dl_type == htons(ETH_TYPE_IPV6)) {
404 if (parse_ipv6(&b, flow)) {
409 if (flow->nw_proto == IPPROTO_TCP) {
410 parse_tcp(packet, &b, flow);
411 } else if (flow->nw_proto == IPPROTO_UDP) {
412 parse_udp(packet, &b, flow);
413 } else if (flow->nw_proto == IPPROTO_ICMPV6) {
414 if (parse_icmpv6(&b, flow)) {
418 } else if (flow->dl_type == htons(ETH_TYPE_ARP)) {
419 const struct arp_eth_header *arp = pull_arp(&b);
420 if (arp && arp->ar_hrd == htons(1)
421 && arp->ar_pro == htons(ETH_TYPE_IP)
422 && arp->ar_hln == ETH_ADDR_LEN
423 && arp->ar_pln == 4) {
424 /* We only match on the lower 8 bits of the opcode. */
425 if (ntohs(arp->ar_op) <= 0xff) {
426 flow->nw_proto = ntohs(arp->ar_op);
429 if ((flow->nw_proto == ARP_OP_REQUEST)
430 || (flow->nw_proto == ARP_OP_REPLY)) {
431 flow->nw_src = arp->ar_spa;
432 flow->nw_dst = arp->ar_tpa;
433 memcpy(flow->arp_sha, arp->ar_sha, ETH_ADDR_LEN);
434 memcpy(flow->arp_tha, arp->ar_tha, ETH_ADDR_LEN);
440 /* For every bit of a field that is wildcarded in 'wildcards', sets the
441 * corresponding bit in 'flow' to zero. */
443 flow_zero_wildcards(struct flow *flow, const struct flow_wildcards *wildcards)
445 const flow_wildcards_t wc = wildcards->wildcards;
448 BUILD_ASSERT_DECL(FLOW_WC_SEQ == 14);
450 for (i = 0; i < FLOW_N_REGS; i++) {
451 flow->regs[i] &= wildcards->reg_masks[i];
453 flow->tun_id &= wildcards->tun_id_mask;
454 flow->metadata &= wildcards->metadata_mask;
455 flow->nw_src &= wildcards->nw_src_mask;
456 flow->nw_dst &= wildcards->nw_dst_mask;
457 if (wc & FWW_IN_PORT) {
460 flow->vlan_tci &= wildcards->vlan_tci_mask;
461 if (wc & FWW_DL_TYPE) {
462 flow->dl_type = htons(0);
464 flow->tp_src &= wildcards->tp_src_mask;
465 flow->tp_dst &= wildcards->tp_dst_mask;
466 eth_addr_bitand(flow->dl_src, wildcards->dl_src_mask, flow->dl_src);
467 eth_addr_bitand(flow->dl_dst, wildcards->dl_dst_mask, flow->dl_dst);
468 if (wc & FWW_NW_PROTO) {
471 flow->ipv6_label &= wildcards->ipv6_label_mask;
472 if (wc & FWW_NW_DSCP) {
473 flow->nw_tos &= ~IP_DSCP_MASK;
475 if (wc & FWW_NW_ECN) {
476 flow->nw_tos &= ~IP_ECN_MASK;
478 if (wc & FWW_NW_TTL) {
481 flow->nw_frag &= wildcards->nw_frag_mask;
482 eth_addr_bitand(flow->arp_sha, wildcards->arp_sha_mask, flow->arp_sha);
483 eth_addr_bitand(flow->arp_tha, wildcards->arp_tha_mask, flow->arp_tha);
484 flow->ipv6_src = ipv6_addr_bitand(&flow->ipv6_src,
485 &wildcards->ipv6_src_mask);
486 flow->ipv6_dst = ipv6_addr_bitand(&flow->ipv6_dst,
487 &wildcards->ipv6_dst_mask);
488 flow->nd_target = ipv6_addr_bitand(&flow->nd_target,
489 &wildcards->nd_target_mask);
490 flow->skb_priority = 0;
493 /* Initializes 'fmd' with the metadata found in 'flow'. */
495 flow_get_metadata(const struct flow *flow, struct flow_metadata *fmd)
497 BUILD_ASSERT_DECL(FLOW_WC_SEQ == 14);
499 fmd->tun_id = flow->tun_id;
500 fmd->tun_id_mask = htonll(UINT64_MAX);
502 fmd->metadata = flow->metadata;
503 fmd->metadata_mask = htonll(UINT64_MAX);
505 memcpy(fmd->regs, flow->regs, sizeof fmd->regs);
506 memset(fmd->reg_masks, 0xff, sizeof fmd->reg_masks);
508 fmd->in_port = flow->in_port;
512 flow_to_string(const struct flow *flow)
514 struct ds ds = DS_EMPTY_INITIALIZER;
515 flow_format(&ds, flow);
520 flow_format(struct ds *ds, const struct flow *flow)
522 ds_put_format(ds, "priority:%"PRIu32
525 ",in_port:%04"PRIx16,
527 ntohll(flow->tun_id),
528 ntohll(flow->metadata),
531 ds_put_format(ds, ",tci(");
532 if (flow->vlan_tci) {
533 ds_put_format(ds, "vlan:%"PRIu16",pcp:%d",
534 vlan_tci_to_vid(flow->vlan_tci),
535 vlan_tci_to_pcp(flow->vlan_tci));
537 ds_put_char(ds, '0');
539 ds_put_format(ds, ") mac("ETH_ADDR_FMT"->"ETH_ADDR_FMT
541 ETH_ADDR_ARGS(flow->dl_src),
542 ETH_ADDR_ARGS(flow->dl_dst),
543 ntohs(flow->dl_type));
545 if (flow->dl_type == htons(ETH_TYPE_IPV6)) {
546 ds_put_format(ds, " label:%#"PRIx32" proto:%"PRIu8" tos:%#"PRIx8
547 " ttl:%"PRIu8" ipv6(",
548 ntohl(flow->ipv6_label), flow->nw_proto,
549 flow->nw_tos, flow->nw_ttl);
550 print_ipv6_addr(ds, &flow->ipv6_src);
551 ds_put_cstr(ds, "->");
552 print_ipv6_addr(ds, &flow->ipv6_dst);
553 ds_put_char(ds, ')');
555 ds_put_format(ds, " proto:%"PRIu8" tos:%#"PRIx8" ttl:%"PRIu8
556 " ip("IP_FMT"->"IP_FMT")",
557 flow->nw_proto, flow->nw_tos, flow->nw_ttl,
558 IP_ARGS(&flow->nw_src), IP_ARGS(&flow->nw_dst));
561 ds_put_format(ds, " frag(%s)",
562 flow->nw_frag == FLOW_NW_FRAG_ANY ? "first"
563 : flow->nw_frag == (FLOW_NW_FRAG_ANY | FLOW_NW_FRAG_LATER)
564 ? "later" : "<error>");
566 if (flow->tp_src || flow->tp_dst) {
567 ds_put_format(ds, " port(%"PRIu16"->%"PRIu16")",
568 ntohs(flow->tp_src), ntohs(flow->tp_dst));
570 if (!eth_addr_is_zero(flow->arp_sha) || !eth_addr_is_zero(flow->arp_tha)) {
571 ds_put_format(ds, " arp_ha("ETH_ADDR_FMT"->"ETH_ADDR_FMT")",
572 ETH_ADDR_ARGS(flow->arp_sha),
573 ETH_ADDR_ARGS(flow->arp_tha));
578 flow_print(FILE *stream, const struct flow *flow)
580 char *s = flow_to_string(flow);
585 /* flow_wildcards functions. */
587 /* Initializes 'wc' as a set of wildcards that matches every packet. */
589 flow_wildcards_init_catchall(struct flow_wildcards *wc)
591 BUILD_ASSERT_DECL(FLOW_WC_SEQ == 14);
593 wc->wildcards = FWW_ALL;
594 wc->tun_id_mask = htonll(0);
595 wc->nw_src_mask = htonl(0);
596 wc->nw_dst_mask = htonl(0);
597 wc->ipv6_src_mask = in6addr_any;
598 wc->ipv6_dst_mask = in6addr_any;
599 wc->ipv6_label_mask = htonl(0);
600 wc->nd_target_mask = in6addr_any;
601 memset(wc->reg_masks, 0, sizeof wc->reg_masks);
602 wc->metadata_mask = htonll(0);
603 wc->vlan_tci_mask = htons(0);
604 wc->nw_frag_mask = 0;
605 wc->tp_src_mask = htons(0);
606 wc->tp_dst_mask = htons(0);
607 memset(wc->dl_src_mask, 0, ETH_ADDR_LEN);
608 memset(wc->dl_dst_mask, 0, ETH_ADDR_LEN);
609 memset(wc->arp_sha_mask, 0, ETH_ADDR_LEN);
610 memset(wc->arp_tha_mask, 0, ETH_ADDR_LEN);
611 memset(wc->zeros, 0, sizeof wc->zeros);
614 /* Initializes 'wc' as an exact-match set of wildcards; that is, 'wc' does not
615 * wildcard any bits or fields. */
617 flow_wildcards_init_exact(struct flow_wildcards *wc)
619 BUILD_ASSERT_DECL(FLOW_WC_SEQ == 14);
622 wc->tun_id_mask = htonll(UINT64_MAX);
623 wc->nw_src_mask = htonl(UINT32_MAX);
624 wc->nw_dst_mask = htonl(UINT32_MAX);
625 wc->ipv6_src_mask = in6addr_exact;
626 wc->ipv6_dst_mask = in6addr_exact;
627 wc->ipv6_label_mask = htonl(UINT32_MAX);
628 wc->nd_target_mask = in6addr_exact;
629 memset(wc->reg_masks, 0xff, sizeof wc->reg_masks);
630 wc->metadata_mask = htonll(UINT64_MAX);
631 wc->vlan_tci_mask = htons(UINT16_MAX);
632 wc->nw_frag_mask = UINT8_MAX;
633 wc->tp_src_mask = htons(UINT16_MAX);
634 wc->tp_dst_mask = htons(UINT16_MAX);
635 memset(wc->dl_src_mask, 0xff, ETH_ADDR_LEN);
636 memset(wc->dl_dst_mask, 0xff, ETH_ADDR_LEN);
637 memset(wc->arp_sha_mask, 0xff, ETH_ADDR_LEN);
638 memset(wc->arp_tha_mask, 0xff, ETH_ADDR_LEN);
639 memset(wc->zeros, 0, sizeof wc->zeros);
642 /* Returns true if 'wc' is exact-match, false if 'wc' wildcards any bits or
645 flow_wildcards_is_exact(const struct flow_wildcards *wc)
649 BUILD_ASSERT_DECL(FLOW_WC_SEQ == 14);
652 || wc->tun_id_mask != htonll(UINT64_MAX)
653 || wc->nw_src_mask != htonl(UINT32_MAX)
654 || wc->nw_dst_mask != htonl(UINT32_MAX)
655 || wc->tp_src_mask != htons(UINT16_MAX)
656 || wc->tp_dst_mask != htons(UINT16_MAX)
657 || wc->vlan_tci_mask != htons(UINT16_MAX)
658 || wc->metadata_mask != htonll(UINT64_MAX)
659 || !eth_mask_is_exact(wc->dl_src_mask)
660 || !eth_mask_is_exact(wc->dl_dst_mask)
661 || !eth_mask_is_exact(wc->arp_sha_mask)
662 || !eth_mask_is_exact(wc->arp_tha_mask)
663 || !ipv6_mask_is_exact(&wc->ipv6_src_mask)
664 || !ipv6_mask_is_exact(&wc->ipv6_dst_mask)
665 || wc->ipv6_label_mask != htonl(UINT32_MAX)
666 || !ipv6_mask_is_exact(&wc->nd_target_mask)
667 || wc->nw_frag_mask != UINT8_MAX) {
671 for (i = 0; i < FLOW_N_REGS; i++) {
672 if (wc->reg_masks[i] != UINT32_MAX) {
680 /* Returns true if 'wc' matches every packet, false if 'wc' fixes any bits or
683 flow_wildcards_is_catchall(const struct flow_wildcards *wc)
687 BUILD_ASSERT_DECL(FLOW_WC_SEQ == 14);
689 if (wc->wildcards != FWW_ALL
690 || wc->tun_id_mask != htonll(0)
691 || wc->nw_src_mask != htonl(0)
692 || wc->nw_dst_mask != htonl(0)
693 || wc->tp_src_mask != htons(0)
694 || wc->tp_dst_mask != htons(0)
695 || wc->vlan_tci_mask != htons(0)
696 || wc->metadata_mask != htonll(0)
697 || !eth_addr_is_zero(wc->dl_src_mask)
698 || !eth_addr_is_zero(wc->dl_dst_mask)
699 || !eth_addr_is_zero(wc->arp_sha_mask)
700 || !eth_addr_is_zero(wc->arp_tha_mask)
701 || !ipv6_mask_is_any(&wc->ipv6_src_mask)
702 || !ipv6_mask_is_any(&wc->ipv6_dst_mask)
703 || wc->ipv6_label_mask != htonl(0)
704 || !ipv6_mask_is_any(&wc->nd_target_mask)
705 || wc->nw_frag_mask != 0) {
709 for (i = 0; i < FLOW_N_REGS; i++) {
710 if (wc->reg_masks[i] != 0) {
718 /* Initializes 'dst' as the combination of wildcards in 'src1' and 'src2'.
719 * That is, a bit or a field is wildcarded in 'dst' if it is wildcarded in
720 * 'src1' or 'src2' or both. */
722 flow_wildcards_combine(struct flow_wildcards *dst,
723 const struct flow_wildcards *src1,
724 const struct flow_wildcards *src2)
728 BUILD_ASSERT_DECL(FLOW_WC_SEQ == 14);
730 dst->wildcards = src1->wildcards | src2->wildcards;
731 dst->tun_id_mask = src1->tun_id_mask & src2->tun_id_mask;
732 dst->nw_src_mask = src1->nw_src_mask & src2->nw_src_mask;
733 dst->nw_dst_mask = src1->nw_dst_mask & src2->nw_dst_mask;
734 dst->ipv6_src_mask = ipv6_addr_bitand(&src1->ipv6_src_mask,
735 &src2->ipv6_src_mask);
736 dst->ipv6_dst_mask = ipv6_addr_bitand(&src1->ipv6_dst_mask,
737 &src2->ipv6_dst_mask);
738 dst->ipv6_label_mask = src1->ipv6_label_mask & src2->ipv6_label_mask;
739 dst->nd_target_mask = ipv6_addr_bitand(&src1->nd_target_mask,
740 &src2->nd_target_mask);
741 for (i = 0; i < FLOW_N_REGS; i++) {
742 dst->reg_masks[i] = src1->reg_masks[i] & src2->reg_masks[i];
744 dst->metadata_mask = src1->metadata_mask & src2->metadata_mask;
745 dst->vlan_tci_mask = src1->vlan_tci_mask & src2->vlan_tci_mask;
746 dst->tp_src_mask = src1->tp_src_mask & src2->tp_src_mask;
747 dst->tp_dst_mask = src1->tp_dst_mask & src2->tp_dst_mask;
748 dst->nw_frag_mask = src1->nw_frag_mask & src2->nw_frag_mask;
749 eth_addr_bitand(src1->dl_src_mask, src2->dl_src_mask, dst->dl_src_mask);
750 eth_addr_bitand(src1->dl_dst_mask, src2->dl_dst_mask, dst->dl_dst_mask);
751 eth_addr_bitand(src1->arp_sha_mask, src2->arp_sha_mask, dst->arp_sha_mask);
752 eth_addr_bitand(src1->arp_tha_mask, src2->arp_tha_mask, dst->arp_tha_mask);
755 /* Returns a hash of the wildcards in 'wc'. */
757 flow_wildcards_hash(const struct flow_wildcards *wc, uint32_t basis)
759 /* If you change struct flow_wildcards and thereby trigger this
760 * assertion, please check that the new struct flow_wildcards has no holes
761 * in it before you update the assertion. */
762 BUILD_ASSERT_DECL(sizeof *wc == 112 + FLOW_N_REGS * 4);
763 return hash_bytes(wc, sizeof *wc, basis);
766 /* Returns true if 'a' and 'b' represent the same wildcards, false if they are
769 flow_wildcards_equal(const struct flow_wildcards *a,
770 const struct flow_wildcards *b)
774 BUILD_ASSERT_DECL(FLOW_WC_SEQ == 14);
776 if (a->wildcards != b->wildcards
777 || a->tun_id_mask != b->tun_id_mask
778 || a->nw_src_mask != b->nw_src_mask
779 || a->nw_dst_mask != b->nw_dst_mask
780 || a->vlan_tci_mask != b->vlan_tci_mask
781 || a->metadata_mask != b->metadata_mask
782 || !ipv6_addr_equals(&a->ipv6_src_mask, &b->ipv6_src_mask)
783 || !ipv6_addr_equals(&a->ipv6_dst_mask, &b->ipv6_dst_mask)
784 || a->ipv6_label_mask != b->ipv6_label_mask
785 || !ipv6_addr_equals(&a->nd_target_mask, &b->nd_target_mask)
786 || a->tp_src_mask != b->tp_src_mask
787 || a->tp_dst_mask != b->tp_dst_mask
788 || a->nw_frag_mask != b->nw_frag_mask
789 || !eth_addr_equals(a->dl_src_mask, b->dl_src_mask)
790 || !eth_addr_equals(a->dl_dst_mask, b->dl_dst_mask)
791 || !eth_addr_equals(a->arp_sha_mask, b->arp_sha_mask)
792 || !eth_addr_equals(a->arp_tha_mask, b->arp_tha_mask)) {
796 for (i = 0; i < FLOW_N_REGS; i++) {
797 if (a->reg_masks[i] != b->reg_masks[i]) {
805 /* Returns true if at least one bit or field is wildcarded in 'a' but not in
806 * 'b', false otherwise. */
808 flow_wildcards_has_extra(const struct flow_wildcards *a,
809 const struct flow_wildcards *b)
812 uint8_t eth_masked[ETH_ADDR_LEN];
813 struct in6_addr ipv6_masked;
815 BUILD_ASSERT_DECL(FLOW_WC_SEQ == 14);
817 for (i = 0; i < FLOW_N_REGS; i++) {
818 if ((a->reg_masks[i] & b->reg_masks[i]) != b->reg_masks[i]) {
823 eth_addr_bitand(a->dl_src_mask, b->dl_src_mask, eth_masked);
824 if (!eth_addr_equals(eth_masked, b->dl_src_mask)) {
828 eth_addr_bitand(a->dl_dst_mask, b->dl_dst_mask, eth_masked);
829 if (!eth_addr_equals(eth_masked, b->dl_dst_mask)) {
833 eth_addr_bitand(a->arp_sha_mask, b->arp_sha_mask, eth_masked);
834 if (!eth_addr_equals(eth_masked, b->arp_sha_mask)) {
838 eth_addr_bitand(a->arp_tha_mask, b->arp_tha_mask, eth_masked);
839 if (!eth_addr_equals(eth_masked, b->arp_tha_mask)) {
843 ipv6_masked = ipv6_addr_bitand(&a->ipv6_src_mask, &b->ipv6_src_mask);
844 if (!ipv6_addr_equals(&ipv6_masked, &b->ipv6_src_mask)) {
848 ipv6_masked = ipv6_addr_bitand(&a->ipv6_dst_mask, &b->ipv6_dst_mask);
849 if (!ipv6_addr_equals(&ipv6_masked, &b->ipv6_dst_mask)) {
853 ipv6_masked = ipv6_addr_bitand(&a->nd_target_mask, &b->nd_target_mask);
854 if (!ipv6_addr_equals(&ipv6_masked, &b->nd_target_mask)) {
858 return (a->wildcards & ~b->wildcards
859 || (a->tun_id_mask & b->tun_id_mask) != b->tun_id_mask
860 || (a->nw_src_mask & b->nw_src_mask) != b->nw_src_mask
861 || (a->nw_dst_mask & b->nw_dst_mask) != b->nw_dst_mask
862 || (a->ipv6_label_mask & b->ipv6_label_mask) != b->ipv6_label_mask
863 || (a->vlan_tci_mask & b->vlan_tci_mask) != b->vlan_tci_mask
864 || (a->metadata_mask & b->metadata_mask) != b->metadata_mask
865 || (a->tp_src_mask & b->tp_src_mask) != b->tp_src_mask
866 || (a->tp_dst_mask & b->tp_dst_mask) != b->tp_dst_mask
867 || (a->nw_frag_mask & b->nw_frag_mask) != b->nw_frag_mask);
870 /* Sets the wildcard mask for register 'idx' in 'wc' to 'mask'.
871 * (A 0-bit indicates a wildcard bit.) */
873 flow_wildcards_set_reg_mask(struct flow_wildcards *wc, int idx, uint32_t mask)
875 wc->reg_masks[idx] = mask;
878 /* Hashes 'flow' based on its L2 through L4 protocol information. */
880 flow_hash_symmetric_l4(const struct flow *flow, uint32_t basis)
885 struct in6_addr ipv6_addr;
890 uint8_t eth_addr[ETH_ADDR_LEN];
896 memset(&fields, 0, sizeof fields);
897 for (i = 0; i < ETH_ADDR_LEN; i++) {
898 fields.eth_addr[i] = flow->dl_src[i] ^ flow->dl_dst[i];
900 fields.vlan_tci = flow->vlan_tci & htons(VLAN_VID_MASK);
901 fields.eth_type = flow->dl_type;
903 /* UDP source and destination port are not taken into account because they
904 * will not necessarily be symmetric in a bidirectional flow. */
905 if (fields.eth_type == htons(ETH_TYPE_IP)) {
906 fields.ipv4_addr = flow->nw_src ^ flow->nw_dst;
907 fields.ip_proto = flow->nw_proto;
908 if (fields.ip_proto == IPPROTO_TCP) {
909 fields.tp_port = flow->tp_src ^ flow->tp_dst;
911 } else if (fields.eth_type == htons(ETH_TYPE_IPV6)) {
912 const uint8_t *a = &flow->ipv6_src.s6_addr[0];
913 const uint8_t *b = &flow->ipv6_dst.s6_addr[0];
914 uint8_t *ipv6_addr = &fields.ipv6_addr.s6_addr[0];
916 for (i=0; i<16; i++) {
917 ipv6_addr[i] = a[i] ^ b[i];
919 fields.ip_proto = flow->nw_proto;
920 if (fields.ip_proto == IPPROTO_TCP) {
921 fields.tp_port = flow->tp_src ^ flow->tp_dst;
924 return hash_bytes(&fields, sizeof fields, basis);
927 /* Hashes the portions of 'flow' designated by 'fields'. */
929 flow_hash_fields(const struct flow *flow, enum nx_hash_fields fields,
934 case NX_HASH_FIELDS_ETH_SRC:
935 return hash_bytes(flow->dl_src, sizeof flow->dl_src, basis);
937 case NX_HASH_FIELDS_SYMMETRIC_L4:
938 return flow_hash_symmetric_l4(flow, basis);
944 /* Returns a string representation of 'fields'. */
946 flow_hash_fields_to_str(enum nx_hash_fields fields)
949 case NX_HASH_FIELDS_ETH_SRC: return "eth_src";
950 case NX_HASH_FIELDS_SYMMETRIC_L4: return "symmetric_l4";
951 default: return "<unknown>";
955 /* Returns true if the value of 'fields' is supported. Otherwise false. */
957 flow_hash_fields_valid(enum nx_hash_fields fields)
959 return fields == NX_HASH_FIELDS_ETH_SRC
960 || fields == NX_HASH_FIELDS_SYMMETRIC_L4;
963 /* Sets the VLAN VID that 'flow' matches to 'vid', which is interpreted as an
964 * OpenFlow 1.0 "dl_vlan" value:
966 * - If it is in the range 0...4095, 'flow->vlan_tci' is set to match
967 * that VLAN. Any existing PCP match is unchanged (it becomes 0 if
968 * 'flow' previously matched packets without a VLAN header).
970 * - If it is OFP_VLAN_NONE, 'flow->vlan_tci' is set to match a packet
971 * without a VLAN tag.
973 * - Other values of 'vid' should not be used. */
975 flow_set_dl_vlan(struct flow *flow, ovs_be16 vid)
977 if (vid == htons(OFP10_VLAN_NONE)) {
978 flow->vlan_tci = htons(0);
980 vid &= htons(VLAN_VID_MASK);
981 flow->vlan_tci &= ~htons(VLAN_VID_MASK);
982 flow->vlan_tci |= htons(VLAN_CFI) | vid;
986 /* Sets the VLAN VID that 'flow' matches to 'vid', which is interpreted as an
987 * OpenFlow 1.2 "vlan_vid" value, that is, the low 13 bits of 'vlan_tci' (VID
990 flow_set_vlan_vid(struct flow *flow, ovs_be16 vid)
992 ovs_be16 mask = htons(VLAN_VID_MASK | VLAN_CFI);
993 flow->vlan_tci &= ~mask;
994 flow->vlan_tci |= vid & mask;
997 /* Sets the VLAN PCP that 'flow' matches to 'pcp', which should be in the
1000 * This function has no effect on the VLAN ID that 'flow' matches.
1002 * After calling this function, 'flow' will not match packets without a VLAN
1005 flow_set_vlan_pcp(struct flow *flow, uint8_t pcp)
1008 flow->vlan_tci &= ~htons(VLAN_PCP_MASK);
1009 flow->vlan_tci |= htons((pcp << VLAN_PCP_SHIFT) | VLAN_CFI);
1012 /* Puts into 'b' a packet that flow_extract() would parse as having the given
1015 * (This is useful only for testing, obviously, and the packet isn't really
1016 * valid. It hasn't got some checksums filled in, for one, and lots of fields
1017 * are just zeroed.) */
1019 flow_compose(struct ofpbuf *b, const struct flow *flow)
1021 eth_compose(b, flow->dl_dst, flow->dl_src, ntohs(flow->dl_type), 0);
1022 if (flow->dl_type == htons(FLOW_DL_TYPE_NONE)) {
1023 struct eth_header *eth = b->l2;
1024 eth->eth_type = htons(b->size);
1028 if (flow->vlan_tci & htons(VLAN_CFI)) {
1029 eth_push_vlan(b, flow->vlan_tci);
1032 if (flow->dl_type == htons(ETH_TYPE_IP)) {
1033 struct ip_header *ip;
1035 b->l3 = ip = ofpbuf_put_zeros(b, sizeof *ip);
1036 ip->ip_ihl_ver = IP_IHL_VER(5, 4);
1037 ip->ip_tos = flow->nw_tos;
1038 ip->ip_proto = flow->nw_proto;
1039 ip->ip_src = flow->nw_src;
1040 ip->ip_dst = flow->nw_dst;
1042 if (flow->nw_frag & FLOW_NW_FRAG_ANY) {
1043 ip->ip_frag_off |= htons(IP_MORE_FRAGMENTS);
1044 if (flow->nw_frag & FLOW_NW_FRAG_LATER) {
1045 ip->ip_frag_off |= htons(100);
1048 if (!(flow->nw_frag & FLOW_NW_FRAG_ANY)
1049 || !(flow->nw_frag & FLOW_NW_FRAG_LATER)) {
1050 if (flow->nw_proto == IPPROTO_TCP) {
1051 struct tcp_header *tcp;
1053 b->l4 = tcp = ofpbuf_put_zeros(b, sizeof *tcp);
1054 tcp->tcp_src = flow->tp_src;
1055 tcp->tcp_dst = flow->tp_dst;
1056 tcp->tcp_ctl = TCP_CTL(0, 5);
1057 } else if (flow->nw_proto == IPPROTO_UDP) {
1058 struct udp_header *udp;
1060 b->l4 = udp = ofpbuf_put_zeros(b, sizeof *udp);
1061 udp->udp_src = flow->tp_src;
1062 udp->udp_dst = flow->tp_dst;
1063 } else if (flow->nw_proto == IPPROTO_ICMP) {
1064 struct icmp_header *icmp;
1066 b->l4 = icmp = ofpbuf_put_zeros(b, sizeof *icmp);
1067 icmp->icmp_type = ntohs(flow->tp_src);
1068 icmp->icmp_code = ntohs(flow->tp_dst);
1069 icmp->icmp_csum = csum(icmp, ICMP_HEADER_LEN);
1074 ip->ip_tot_len = htons((uint8_t *) b->data + b->size
1075 - (uint8_t *) b->l3);
1076 ip->ip_csum = csum(ip, sizeof *ip);
1077 } else if (flow->dl_type == htons(ETH_TYPE_IPV6)) {
1079 } else if (flow->dl_type == htons(ETH_TYPE_ARP)) {
1080 struct arp_eth_header *arp;
1082 b->l3 = arp = ofpbuf_put_zeros(b, sizeof *arp);
1083 arp->ar_hrd = htons(1);
1084 arp->ar_pro = htons(ETH_TYPE_IP);
1085 arp->ar_hln = ETH_ADDR_LEN;
1087 arp->ar_op = htons(flow->nw_proto);
1089 if (flow->nw_proto == ARP_OP_REQUEST ||
1090 flow->nw_proto == ARP_OP_REPLY) {
1091 arp->ar_spa = flow->nw_src;
1092 arp->ar_tpa = flow->nw_dst;
1093 memcpy(arp->ar_sha, flow->arp_sha, ETH_ADDR_LEN);
1094 memcpy(arp->ar_tha, flow->arp_tha, ETH_ADDR_LEN);