2
* Copyright (c) 2008, 2009, 2010, 2011, 2012 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->nw_tos = ntohl(tc_flow) >> 20;
152
flow->ipv6_label = tc_flow & htonl(IPV6_LABEL_MASK);
153
flow->nw_ttl = nh->ip6_hlim;
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
flow->nw_frag = FLOW_NW_FRAG_ANY;
207
if ((frag_hdr->ip6f_offlg & IP6F_OFF_MASK) != htons(0)) {
208
flow->nw_frag |= FLOW_NW_FRAG_LATER;
209
nexthdr = IPPROTO_FRAGMENT;
215
flow->nw_proto = nexthdr;
220
parse_tcp(struct ofpbuf *packet, struct ofpbuf *b, struct flow *flow)
222
const struct tcp_header *tcp = pull_tcp(b);
224
flow->tp_src = tcp->tcp_src;
225
flow->tp_dst = tcp->tcp_dst;
226
packet->l7 = b->data;
231
parse_udp(struct ofpbuf *packet, struct ofpbuf *b, struct flow *flow)
233
const struct udp_header *udp = pull_udp(b);
235
flow->tp_src = udp->udp_src;
236
flow->tp_dst = udp->udp_dst;
237
packet->l7 = b->data;
242
parse_icmpv6(struct ofpbuf *b, struct flow *flow)
244
const struct icmp6_hdr *icmp = pull_icmpv6(b);
250
/* The ICMPv6 type and code fields use the 16-bit transport port
251
* fields, so we need to store them in 16-bit network byte order. */
252
flow->tp_src = htons(icmp->icmp6_type);
253
flow->tp_dst = htons(icmp->icmp6_code);
255
if (icmp->icmp6_code == 0 &&
256
(icmp->icmp6_type == ND_NEIGHBOR_SOLICIT ||
257
icmp->icmp6_type == ND_NEIGHBOR_ADVERT)) {
258
const struct in6_addr *nd_target;
260
nd_target = ofpbuf_try_pull(b, sizeof *nd_target);
264
flow->nd_target = *nd_target;
266
while (b->size >= 8) {
267
/* The minimum size of an option is 8 bytes, which also is
268
* the size of Ethernet link-layer options. */
269
const struct nd_opt_hdr *nd_opt = b->data;
270
int opt_len = nd_opt->nd_opt_len * 8;
272
if (!opt_len || opt_len > b->size) {
276
/* Store the link layer address if the appropriate option is
277
* provided. It is considered an error if the same link
278
* layer option is specified twice. */
279
if (nd_opt->nd_opt_type == ND_OPT_SOURCE_LINKADDR
281
if (eth_addr_is_zero(flow->arp_sha)) {
282
memcpy(flow->arp_sha, nd_opt + 1, ETH_ADDR_LEN);
286
} else if (nd_opt->nd_opt_type == ND_OPT_TARGET_LINKADDR
288
if (eth_addr_is_zero(flow->arp_tha)) {
289
memcpy(flow->arp_tha, nd_opt + 1, ETH_ADDR_LEN);
295
if (!ofpbuf_try_pull(b, opt_len)) {
304
memset(&flow->nd_target, 0, sizeof(flow->nd_target));
305
memset(flow->arp_sha, 0, sizeof(flow->arp_sha));
306
memset(flow->arp_tha, 0, sizeof(flow->arp_tha));
312
/* Initializes 'flow' members from 'packet', 'tun_id', and 'ofp_in_port'.
313
* Initializes 'packet' header pointers as follows:
315
* - packet->l2 to the start of the Ethernet header.
317
* - packet->l3 to just past the Ethernet header, or just past the
318
* vlan_header if one is present, to the first byte of the payload of the
321
* - packet->l4 to just past the IPv4 header, if one is present and has a
322
* correct length, and otherwise NULL.
324
* - packet->l7 to just past the TCP or UDP or ICMP header, if one is
325
* present and has a correct length, and otherwise NULL.
328
flow_extract(struct ofpbuf *packet, uint32_t priority, ovs_be64 tun_id,
329
uint16_t ofp_in_port, struct flow *flow)
331
struct ofpbuf b = *packet;
332
struct eth_header *eth;
334
COVERAGE_INC(flow_extract);
336
memset(flow, 0, sizeof *flow);
337
flow->tun_id = tun_id;
338
flow->in_port = ofp_in_port;
339
flow->priority = priority;
346
if (b.size < sizeof *eth) {
352
memcpy(flow->dl_src, eth->eth_src, ETH_ADDR_LEN);
353
memcpy(flow->dl_dst, eth->eth_dst, ETH_ADDR_LEN);
355
/* dl_type, vlan_tci. */
356
ofpbuf_pull(&b, ETH_ADDR_LEN * 2);
357
if (eth->eth_type == htons(ETH_TYPE_VLAN)) {
358
parse_vlan(&b, flow);
360
flow->dl_type = parse_ethertype(&b);
364
if (flow->dl_type == htons(ETH_TYPE_IP)) {
365
const struct ip_header *nh = pull_ip(&b);
369
flow->nw_src = get_unaligned_be32(&nh->ip_src);
370
flow->nw_dst = get_unaligned_be32(&nh->ip_dst);
371
flow->nw_proto = nh->ip_proto;
373
flow->nw_tos = nh->ip_tos;
374
if (IP_IS_FRAGMENT(nh->ip_frag_off)) {
375
flow->nw_frag = FLOW_NW_FRAG_ANY;
376
if (nh->ip_frag_off & htons(IP_FRAG_OFF_MASK)) {
377
flow->nw_frag |= FLOW_NW_FRAG_LATER;
380
flow->nw_ttl = nh->ip_ttl;
382
if (!(nh->ip_frag_off & htons(IP_FRAG_OFF_MASK))) {
383
if (flow->nw_proto == IPPROTO_TCP) {
384
parse_tcp(packet, &b, flow);
385
} else if (flow->nw_proto == IPPROTO_UDP) {
386
parse_udp(packet, &b, flow);
387
} else if (flow->nw_proto == IPPROTO_ICMP) {
388
const struct icmp_header *icmp = pull_icmp(&b);
390
flow->tp_src = htons(icmp->icmp_type);
391
flow->tp_dst = htons(icmp->icmp_code);
397
} else if (flow->dl_type == htons(ETH_TYPE_IPV6)) {
398
if (parse_ipv6(&b, flow)) {
403
if (flow->nw_proto == IPPROTO_TCP) {
404
parse_tcp(packet, &b, flow);
405
} else if (flow->nw_proto == IPPROTO_UDP) {
406
parse_udp(packet, &b, flow);
407
} else if (flow->nw_proto == IPPROTO_ICMPV6) {
408
if (parse_icmpv6(&b, flow)) {
412
} else if (flow->dl_type == htons(ETH_TYPE_ARP)) {
413
const struct arp_eth_header *arp = pull_arp(&b);
414
if (arp && arp->ar_hrd == htons(1)
415
&& arp->ar_pro == htons(ETH_TYPE_IP)
416
&& arp->ar_hln == ETH_ADDR_LEN
417
&& arp->ar_pln == 4) {
418
/* We only match on the lower 8 bits of the opcode. */
419
if (ntohs(arp->ar_op) <= 0xff) {
420
flow->nw_proto = ntohs(arp->ar_op);
423
if ((flow->nw_proto == ARP_OP_REQUEST)
424
|| (flow->nw_proto == ARP_OP_REPLY)) {
425
flow->nw_src = arp->ar_spa;
426
flow->nw_dst = arp->ar_tpa;
427
memcpy(flow->arp_sha, arp->ar_sha, ETH_ADDR_LEN);
428
memcpy(flow->arp_tha, arp->ar_tha, ETH_ADDR_LEN);
434
/* For every bit of a field that is wildcarded in 'wildcards', sets the
435
* corresponding bit in 'flow' to zero. */
437
flow_zero_wildcards(struct flow *flow, const struct flow_wildcards *wildcards)
439
const flow_wildcards_t wc = wildcards->wildcards;
442
BUILD_ASSERT_DECL(FLOW_WC_SEQ == 7);
444
for (i = 0; i < FLOW_N_REGS; i++) {
445
flow->regs[i] &= wildcards->reg_masks[i];
447
flow->tun_id &= wildcards->tun_id_mask;
448
flow->nw_src &= wildcards->nw_src_mask;
449
flow->nw_dst &= wildcards->nw_dst_mask;
450
if (wc & FWW_IN_PORT) {
453
flow->vlan_tci &= wildcards->vlan_tci_mask;
454
if (wc & FWW_DL_TYPE) {
455
flow->dl_type = htons(0);
457
if (wc & FWW_TP_SRC) {
458
flow->tp_src = htons(0);
460
if (wc & FWW_TP_DST) {
461
flow->tp_dst = htons(0);
463
if (wc & FWW_DL_SRC) {
464
memset(flow->dl_src, 0, sizeof flow->dl_src);
466
if (wc & FWW_DL_DST) {
467
flow->dl_dst[0] &= 0x01;
468
memset(&flow->dl_dst[1], 0, 5);
470
if (wc & FWW_ETH_MCAST) {
471
flow->dl_dst[0] &= 0xfe;
473
if (wc & FWW_NW_PROTO) {
476
if (wc & FWW_IPV6_LABEL) {
477
flow->ipv6_label = htonl(0);
479
if (wc & FWW_NW_DSCP) {
480
flow->nw_tos &= ~IP_DSCP_MASK;
482
if (wc & FWW_NW_ECN) {
483
flow->nw_tos &= ~IP_ECN_MASK;
485
if (wc & FWW_NW_TTL) {
488
flow->nw_frag &= wildcards->nw_frag_mask;
489
if (wc & FWW_ARP_SHA) {
490
memset(flow->arp_sha, 0, sizeof flow->arp_sha);
492
if (wc & FWW_ARP_THA) {
493
memset(flow->arp_tha, 0, sizeof flow->arp_tha);
495
flow->ipv6_src = ipv6_addr_bitand(&flow->ipv6_src,
496
&wildcards->ipv6_src_mask);
497
flow->ipv6_dst = ipv6_addr_bitand(&flow->ipv6_dst,
498
&wildcards->ipv6_dst_mask);
499
if (wc & FWW_ND_TARGET) {
500
memset(&flow->nd_target, 0, sizeof flow->nd_target);
506
flow_to_string(const struct flow *flow)
508
struct ds ds = DS_EMPTY_INITIALIZER;
509
flow_format(&ds, flow);
514
flow_format(struct ds *ds, const struct flow *flow)
516
ds_put_format(ds, "priority%"PRIu32
520
ntohll(flow->tun_id),
523
ds_put_format(ds, ":tci(");
524
if (flow->vlan_tci) {
525
ds_put_format(ds, "vlan%"PRIu16",pcp%d",
526
vlan_tci_to_vid(flow->vlan_tci),
527
vlan_tci_to_pcp(flow->vlan_tci));
529
ds_put_char(ds, '0');
531
ds_put_format(ds, ") mac"ETH_ADDR_FMT"->"ETH_ADDR_FMT
533
ETH_ADDR_ARGS(flow->dl_src),
534
ETH_ADDR_ARGS(flow->dl_dst),
535
ntohs(flow->dl_type));
537
if (flow->dl_type == htons(ETH_TYPE_IPV6)) {
538
ds_put_format(ds, " label%#"PRIx32" proto%"PRIu8" tos%#"PRIx8
540
ntohl(flow->ipv6_label), flow->nw_proto,
541
flow->nw_tos, flow->nw_ttl);
542
print_ipv6_addr(ds, &flow->ipv6_src);
543
ds_put_cstr(ds, "->");
544
print_ipv6_addr(ds, &flow->ipv6_dst);
547
ds_put_format(ds, " proto%"PRIu8" tos%#"PRIx8" ttl%"PRIu8
548
" ip"IP_FMT"->"IP_FMT,
549
flow->nw_proto, flow->nw_tos, flow->nw_ttl,
550
IP_ARGS(&flow->nw_src), IP_ARGS(&flow->nw_dst));
553
ds_put_format(ds, " frag(%s)",
554
flow->nw_frag == FLOW_NW_FRAG_ANY ? "first"
555
: flow->nw_frag == (FLOW_NW_FRAG_ANY | FLOW_NW_FRAG_LATER)
556
? "later" : "<error>");
558
if (flow->tp_src || flow->tp_dst) {
559
ds_put_format(ds, " port%"PRIu16"->%"PRIu16,
560
ntohs(flow->tp_src), ntohs(flow->tp_dst));
562
if (!eth_addr_is_zero(flow->arp_sha) || !eth_addr_is_zero(flow->arp_tha)) {
563
ds_put_format(ds, " arp_ha"ETH_ADDR_FMT"->"ETH_ADDR_FMT,
564
ETH_ADDR_ARGS(flow->arp_sha),
565
ETH_ADDR_ARGS(flow->arp_tha));
570
flow_print(FILE *stream, const struct flow *flow)
572
char *s = flow_to_string(flow);
577
/* flow_wildcards functions. */
579
/* Initializes 'wc' as a set of wildcards that matches every packet. */
581
flow_wildcards_init_catchall(struct flow_wildcards *wc)
583
BUILD_ASSERT_DECL(FLOW_WC_SEQ == 7);
585
wc->wildcards = FWW_ALL;
586
wc->tun_id_mask = htonll(0);
587
wc->nw_src_mask = htonl(0);
588
wc->nw_dst_mask = htonl(0);
589
wc->ipv6_src_mask = in6addr_any;
590
wc->ipv6_dst_mask = in6addr_any;
591
memset(wc->reg_masks, 0, sizeof wc->reg_masks);
592
wc->vlan_tci_mask = htons(0);
593
wc->nw_frag_mask = 0;
594
memset(wc->zeros, 0, sizeof wc->zeros);
597
/* Initializes 'wc' as an exact-match set of wildcards; that is, 'wc' does not
598
* wildcard any bits or fields. */
600
flow_wildcards_init_exact(struct flow_wildcards *wc)
602
BUILD_ASSERT_DECL(FLOW_WC_SEQ == 7);
605
wc->tun_id_mask = htonll(UINT64_MAX);
606
wc->nw_src_mask = htonl(UINT32_MAX);
607
wc->nw_dst_mask = htonl(UINT32_MAX);
608
wc->ipv6_src_mask = in6addr_exact;
609
wc->ipv6_dst_mask = in6addr_exact;
610
memset(wc->reg_masks, 0xff, sizeof wc->reg_masks);
611
wc->vlan_tci_mask = htons(UINT16_MAX);
612
wc->nw_frag_mask = UINT8_MAX;
613
memset(wc->zeros, 0, sizeof wc->zeros);
616
/* Returns true if 'wc' is exact-match, false if 'wc' wildcards any bits or
619
flow_wildcards_is_exact(const struct flow_wildcards *wc)
623
BUILD_ASSERT_DECL(FLOW_WC_SEQ == 7);
626
|| wc->tun_id_mask != htonll(UINT64_MAX)
627
|| wc->nw_src_mask != htonl(UINT32_MAX)
628
|| wc->nw_dst_mask != htonl(UINT32_MAX)
629
|| wc->vlan_tci_mask != htons(UINT16_MAX)
630
|| !ipv6_mask_is_exact(&wc->ipv6_src_mask)
631
|| !ipv6_mask_is_exact(&wc->ipv6_dst_mask)
632
|| wc->nw_frag_mask != UINT8_MAX) {
636
for (i = 0; i < FLOW_N_REGS; i++) {
637
if (wc->reg_masks[i] != UINT32_MAX) {
645
/* Returns true if 'wc' matches every packet, false if 'wc' fixes any bits or
648
flow_wildcards_is_catchall(const struct flow_wildcards *wc)
652
BUILD_ASSERT_DECL(FLOW_WC_SEQ == 7);
654
if (wc->wildcards != FWW_ALL
655
|| wc->tun_id_mask != htonll(0)
656
|| wc->nw_src_mask != htonl(0)
657
|| wc->nw_dst_mask != htonl(0)
658
|| wc->vlan_tci_mask != htons(0)
659
|| !ipv6_mask_is_any(&wc->ipv6_src_mask)
660
|| !ipv6_mask_is_any(&wc->ipv6_dst_mask)
661
|| wc->nw_frag_mask != 0) {
665
for (i = 0; i < FLOW_N_REGS; i++) {
666
if (wc->reg_masks[i] != 0) {
674
/* Initializes 'dst' as the combination of wildcards in 'src1' and 'src2'.
675
* That is, a bit or a field is wildcarded in 'dst' if it is wildcarded in
676
* 'src1' or 'src2' or both. */
678
flow_wildcards_combine(struct flow_wildcards *dst,
679
const struct flow_wildcards *src1,
680
const struct flow_wildcards *src2)
684
dst->wildcards = src1->wildcards | src2->wildcards;
685
dst->tun_id_mask = src1->tun_id_mask & src2->tun_id_mask;
686
dst->nw_src_mask = src1->nw_src_mask & src2->nw_src_mask;
687
dst->nw_dst_mask = src1->nw_dst_mask & src2->nw_dst_mask;
688
dst->ipv6_src_mask = ipv6_addr_bitand(&src1->ipv6_src_mask,
689
&src2->ipv6_src_mask);
690
dst->ipv6_dst_mask = ipv6_addr_bitand(&src1->ipv6_dst_mask,
691
&src2->ipv6_dst_mask);
692
for (i = 0; i < FLOW_N_REGS; i++) {
693
dst->reg_masks[i] = src1->reg_masks[i] & src2->reg_masks[i];
695
dst->vlan_tci_mask = src1->vlan_tci_mask & src2->vlan_tci_mask;
696
dst->nw_frag_mask = src1->nw_frag_mask & src2->nw_frag_mask;
699
/* Returns a hash of the wildcards in 'wc'. */
701
flow_wildcards_hash(const struct flow_wildcards *wc, uint32_t basis)
703
/* If you change struct flow_wildcards and thereby trigger this
704
* assertion, please check that the new struct flow_wildcards has no holes
705
* in it before you update the assertion. */
706
BUILD_ASSERT_DECL(sizeof *wc == 60 + FLOW_N_REGS * 4);
707
return hash_bytes(wc, sizeof *wc, basis);
710
/* Returns true if 'a' and 'b' represent the same wildcards, false if they are
713
flow_wildcards_equal(const struct flow_wildcards *a,
714
const struct flow_wildcards *b)
718
if (a->wildcards != b->wildcards
719
|| a->tun_id_mask != b->tun_id_mask
720
|| a->nw_src_mask != b->nw_src_mask
721
|| a->nw_dst_mask != b->nw_dst_mask
722
|| a->vlan_tci_mask != b->vlan_tci_mask
723
|| !ipv6_addr_equals(&a->ipv6_src_mask, &b->ipv6_src_mask)
724
|| !ipv6_addr_equals(&a->ipv6_dst_mask, &b->ipv6_dst_mask)
725
|| a->nw_frag_mask != b->nw_frag_mask) {
729
for (i = 0; i < FLOW_N_REGS; i++) {
730
if (a->reg_masks[i] != b->reg_masks[i]) {
738
/* Returns true if at least one bit or field is wildcarded in 'a' but not in
739
* 'b', false otherwise. */
741
flow_wildcards_has_extra(const struct flow_wildcards *a,
742
const struct flow_wildcards *b)
745
struct in6_addr ipv6_masked;
747
for (i = 0; i < FLOW_N_REGS; i++) {
748
if ((a->reg_masks[i] & b->reg_masks[i]) != b->reg_masks[i]) {
753
ipv6_masked = ipv6_addr_bitand(&a->ipv6_src_mask, &b->ipv6_src_mask);
754
if (!ipv6_addr_equals(&ipv6_masked, &b->ipv6_src_mask)) {
758
ipv6_masked = ipv6_addr_bitand(&a->ipv6_dst_mask, &b->ipv6_dst_mask);
759
if (!ipv6_addr_equals(&ipv6_masked, &b->ipv6_dst_mask)) {
763
return (a->wildcards & ~b->wildcards
764
|| (a->tun_id_mask & b->tun_id_mask) != b->tun_id_mask
765
|| (a->nw_src_mask & b->nw_src_mask) != b->nw_src_mask
766
|| (a->nw_dst_mask & b->nw_dst_mask) != b->nw_dst_mask
767
|| (a->vlan_tci_mask & b->vlan_tci_mask) != b->vlan_tci_mask
768
|| (a->nw_frag_mask & b->nw_frag_mask) != b->nw_frag_mask);
771
/* Sets the wildcard mask for register 'idx' in 'wc' to 'mask'.
772
* (A 0-bit indicates a wildcard bit.) */
774
flow_wildcards_set_reg_mask(struct flow_wildcards *wc, int idx, uint32_t mask)
776
wc->reg_masks[idx] = mask;
779
/* Returns the wildcard bitmask for the Ethernet destination address
780
* that 'wc' specifies. The bitmask has a 0 in each bit that is wildcarded
781
* and a 1 in each bit that must match. */
783
flow_wildcards_to_dl_dst_mask(flow_wildcards_t wc)
785
static const uint8_t no_wild[] = {0xff, 0xff, 0xff, 0xff, 0xff, 0xff};
786
static const uint8_t addr_wild[] = {0x01, 0x00, 0x00, 0x00, 0x00, 0x00};
787
static const uint8_t mcast_wild[] = {0xfe, 0xff, 0xff, 0xff, 0xff, 0xff};
788
static const uint8_t all_wild[] = {0x00, 0x00, 0x00, 0x00, 0x00, 0x00};
790
switch (wc & (FWW_DL_DST | FWW_ETH_MCAST)) {
791
case 0: return no_wild;
792
case FWW_DL_DST: return addr_wild;
793
case FWW_ETH_MCAST: return mcast_wild;
794
case FWW_DL_DST | FWW_ETH_MCAST: return all_wild;
799
/* Returns true if 'mask' is a valid wildcard bitmask for the Ethernet
800
* destination address. Valid bitmasks are either all-bits-0 or all-bits-1,
801
* except that the multicast bit may differ from the rest of the bits. So,
802
* there are four possible valid bitmasks:
804
* - 00:00:00:00:00:00
805
* - 01:00:00:00:00:00
806
* - fe:ff:ff:ff:ff:ff
807
* - ff:ff:ff:ff:ff:ff
809
* All other bitmasks are invalid. */
811
flow_wildcards_is_dl_dst_mask_valid(const uint8_t mask[ETH_ADDR_LEN])
816
return (mask[1] | mask[2] | mask[3] | mask[4] | mask[5]) == 0x00;
820
return (mask[1] & mask[2] & mask[3] & mask[4] & mask[5]) == 0xff;
827
/* Returns 'wc' with the FWW_DL_DST and FWW_ETH_MCAST bits modified
828
* appropriately to match 'mask'.
830
* This function will assert-fail if 'mask' is invalid. Only 'mask' values
831
* accepted by flow_wildcards_is_dl_dst_mask_valid() are allowed. */
833
flow_wildcards_set_dl_dst_mask(flow_wildcards_t wc,
834
const uint8_t mask[ETH_ADDR_LEN])
836
assert(flow_wildcards_is_dl_dst_mask_valid(mask));
840
return wc | FWW_DL_DST | FWW_ETH_MCAST;
843
return (wc | FWW_DL_DST) & ~FWW_ETH_MCAST;
846
return (wc & ~FWW_DL_DST) | FWW_ETH_MCAST;
849
return wc & ~(FWW_DL_DST | FWW_ETH_MCAST);
856
/* Hashes 'flow' based on its L2 through L4 protocol information. */
858
flow_hash_symmetric_l4(const struct flow *flow, uint32_t basis)
863
struct in6_addr ipv6_addr;
868
uint8_t eth_addr[ETH_ADDR_LEN];
874
memset(&fields, 0, sizeof fields);
875
for (i = 0; i < ETH_ADDR_LEN; i++) {
876
fields.eth_addr[i] = flow->dl_src[i] ^ flow->dl_dst[i];
878
fields.vlan_tci = flow->vlan_tci & htons(VLAN_VID_MASK);
879
fields.eth_type = flow->dl_type;
881
/* UDP source and destination port are not taken into account because they
882
* will not necessarily be symmetric in a bidirectional flow. */
883
if (fields.eth_type == htons(ETH_TYPE_IP)) {
884
fields.ipv4_addr = flow->nw_src ^ flow->nw_dst;
885
fields.ip_proto = flow->nw_proto;
886
if (fields.ip_proto == IPPROTO_TCP) {
887
fields.tp_addr = flow->tp_src ^ flow->tp_dst;
889
} else if (fields.eth_type == htons(ETH_TYPE_IPV6)) {
890
const uint8_t *a = &flow->ipv6_src.s6_addr[0];
891
const uint8_t *b = &flow->ipv6_dst.s6_addr[0];
892
uint8_t *ipv6_addr = &fields.ipv6_addr.s6_addr[0];
894
for (i=0; i<16; i++) {
895
ipv6_addr[i] = a[i] ^ b[i];
897
fields.ip_proto = flow->nw_proto;
898
if (fields.ip_proto == IPPROTO_TCP) {
899
fields.tp_addr = flow->tp_src ^ flow->tp_dst;
902
return hash_bytes(&fields, sizeof fields, basis);
905
/* Hashes the portions of 'flow' designated by 'fields'. */
907
flow_hash_fields(const struct flow *flow, enum nx_hash_fields fields,
912
case NX_HASH_FIELDS_ETH_SRC:
913
return hash_bytes(flow->dl_src, sizeof flow->dl_src, basis);
915
case NX_HASH_FIELDS_SYMMETRIC_L4:
916
return flow_hash_symmetric_l4(flow, basis);
922
/* Returns a string representation of 'fields'. */
924
flow_hash_fields_to_str(enum nx_hash_fields fields)
927
case NX_HASH_FIELDS_ETH_SRC: return "eth_src";
928
case NX_HASH_FIELDS_SYMMETRIC_L4: return "symmetric_l4";
929
default: return "<unknown>";
933
/* Returns true if the value of 'fields' is supported. Otherwise false. */
935
flow_hash_fields_valid(enum nx_hash_fields fields)
937
return fields == NX_HASH_FIELDS_ETH_SRC
938
|| fields == NX_HASH_FIELDS_SYMMETRIC_L4;
941
/* Sets the VLAN VID that 'flow' matches to 'vid', which is interpreted as an
942
* OpenFlow 1.0 "dl_vlan" value:
944
* - If it is in the range 0...4095, 'flow->vlan_tci' is set to match
945
* that VLAN. Any existing PCP match is unchanged (it becomes 0 if
946
* 'flow' previously matched packets without a VLAN header).
948
* - If it is OFP_VLAN_NONE, 'flow->vlan_tci' is set to match a packet
949
* without a VLAN tag.
951
* - Other values of 'vid' should not be used. */
953
flow_set_vlan_vid(struct flow *flow, ovs_be16 vid)
955
if (vid == htons(OFP_VLAN_NONE)) {
956
flow->vlan_tci = htons(0);
958
vid &= htons(VLAN_VID_MASK);
959
flow->vlan_tci &= ~htons(VLAN_VID_MASK);
960
flow->vlan_tci |= htons(VLAN_CFI) | vid;
964
/* Sets the VLAN PCP that 'flow' matches to 'pcp', which should be in the
967
* This function has no effect on the VLAN ID that 'flow' matches.
969
* After calling this function, 'flow' will not match packets without a VLAN
972
flow_set_vlan_pcp(struct flow *flow, uint8_t pcp)
975
flow->vlan_tci &= ~htons(VLAN_PCP_MASK);
976
flow->vlan_tci |= htons((pcp << VLAN_PCP_SHIFT) | VLAN_CFI);
979
/* Puts into 'b' a packet that flow_extract() would parse as having the given
982
* (This is useful only for testing, obviously, and the packet isn't really
983
* valid. It hasn't got any checksums filled in, for one, and lots of fields
984
* are just zeroed.) */
986
flow_compose(struct ofpbuf *b, const struct flow *flow)
988
eth_compose(b, flow->dl_dst, flow->dl_src, ntohs(flow->dl_type), 0);
989
if (flow->dl_type == htons(FLOW_DL_TYPE_NONE)) {
990
struct eth_header *eth = b->l2;
991
eth->eth_type = htons(b->size);
995
if (flow->vlan_tci & htons(VLAN_CFI)) {
996
eth_push_vlan(b, flow->vlan_tci & ~htons(VLAN_CFI));
999
if (flow->dl_type == htons(ETH_TYPE_IP)) {
1000
struct ip_header *ip;
1002
b->l3 = ip = ofpbuf_put_zeros(b, sizeof *ip);
1003
ip->ip_ihl_ver = IP_IHL_VER(5, 4);
1004
ip->ip_tos = flow->nw_tos;
1005
ip->ip_proto = flow->nw_proto;
1006
ip->ip_src = flow->nw_src;
1007
ip->ip_dst = flow->nw_dst;
1009
if (flow->nw_frag & FLOW_NW_FRAG_ANY) {
1010
ip->ip_frag_off |= htons(IP_MORE_FRAGMENTS);
1011
if (flow->nw_frag & FLOW_NW_FRAG_LATER) {
1012
ip->ip_frag_off |= htons(100);
1015
if (!(flow->nw_frag & FLOW_NW_FRAG_ANY)
1016
|| !(flow->nw_frag & FLOW_NW_FRAG_LATER)) {
1017
if (flow->nw_proto == IPPROTO_TCP) {
1018
struct tcp_header *tcp;
1020
b->l4 = tcp = ofpbuf_put_zeros(b, sizeof *tcp);
1021
tcp->tcp_src = flow->tp_src;
1022
tcp->tcp_dst = flow->tp_dst;
1023
} else if (flow->nw_proto == IPPROTO_UDP) {
1024
struct udp_header *udp;
1026
b->l4 = udp = ofpbuf_put_zeros(b, sizeof *udp);
1027
udp->udp_src = flow->tp_src;
1028
udp->udp_dst = flow->tp_dst;
1029
} else if (flow->nw_proto == IPPROTO_ICMP) {
1030
struct icmp_header *icmp;
1032
b->l4 = icmp = ofpbuf_put_zeros(b, sizeof *icmp);
1033
icmp->icmp_type = ntohs(flow->tp_src);
1034
icmp->icmp_code = ntohs(flow->tp_dst);
1037
} else if (flow->dl_type == htons(ETH_TYPE_IPV6)) {
1039
} else if (flow->dl_type == htons(ETH_TYPE_ARP)) {
1040
struct arp_eth_header *arp;
1042
b->l3 = arp = ofpbuf_put_zeros(b, sizeof *arp);
1043
arp->ar_hrd = htons(1);
1044
arp->ar_pro = htons(ETH_TYPE_IP);
1045
arp->ar_hln = ETH_ADDR_LEN;
1047
arp->ar_op = htons(flow->nw_proto);
1049
if (flow->nw_proto == ARP_OP_REQUEST ||
1050
flow->nw_proto == ARP_OP_REPLY) {
1051
arp->ar_spa = flow->nw_src;
1052
arp->ar_tpa = flow->nw_dst;
1053
memcpy(arp->ar_sha, flow->arp_sha, ETH_ADDR_LEN);
1054
memcpy(arp->ar_tha, flow->arp_tha, ETH_ADDR_LEN);