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>
23 #include <netinet/in.h>
24 #include <netinet/icmp6.h>
25 #include <netinet/ip6.h>
29 #include "byte-order.h"
32 #include "dynamic-string.h"
35 #include "openflow/openflow.h"
37 #include "unaligned.h"
40 VLOG_DEFINE_THIS_MODULE(flow
);
42 COVERAGE_DEFINE(flow_extract
);
43 COVERAGE_DEFINE(miniflow_malloc
);
45 static struct arp_eth_header
*
46 pull_arp(struct ofpbuf
*packet
)
48 return ofpbuf_try_pull(packet
, ARP_ETH_HEADER_LEN
);
51 static struct ip_header
*
52 pull_ip(struct ofpbuf
*packet
)
54 if (packet
->size
>= IP_HEADER_LEN
) {
55 struct ip_header
*ip
= packet
->data
;
56 int ip_len
= IP_IHL(ip
->ip_ihl_ver
) * 4;
57 if (ip_len
>= IP_HEADER_LEN
&& packet
->size
>= ip_len
) {
58 return ofpbuf_pull(packet
, ip_len
);
64 static struct tcp_header
*
65 pull_tcp(struct ofpbuf
*packet
)
67 if (packet
->size
>= TCP_HEADER_LEN
) {
68 struct tcp_header
*tcp
= packet
->data
;
69 int tcp_len
= TCP_OFFSET(tcp
->tcp_ctl
) * 4;
70 if (tcp_len
>= TCP_HEADER_LEN
&& packet
->size
>= tcp_len
) {
71 return ofpbuf_pull(packet
, tcp_len
);
77 static struct udp_header
*
78 pull_udp(struct ofpbuf
*packet
)
80 return ofpbuf_try_pull(packet
, UDP_HEADER_LEN
);
83 static struct icmp_header
*
84 pull_icmp(struct ofpbuf
*packet
)
86 return ofpbuf_try_pull(packet
, ICMP_HEADER_LEN
);
89 static struct icmp6_hdr
*
90 pull_icmpv6(struct ofpbuf
*packet
)
92 return ofpbuf_try_pull(packet
, sizeof(struct icmp6_hdr
));
96 parse_vlan(struct ofpbuf
*b
, struct flow
*flow
)
99 ovs_be16 eth_type
; /* ETH_TYPE_VLAN */
103 if (b
->size
>= sizeof(struct qtag_prefix
) + sizeof(ovs_be16
)) {
104 struct qtag_prefix
*qp
= ofpbuf_pull(b
, sizeof *qp
);
105 flow
->vlan_tci
= qp
->tci
| htons(VLAN_CFI
);
110 parse_ethertype(struct ofpbuf
*b
)
112 struct llc_snap_header
*llc
;
115 proto
= *(ovs_be16
*) ofpbuf_pull(b
, sizeof proto
);
116 if (ntohs(proto
) >= ETH_TYPE_MIN
) {
120 if (b
->size
< sizeof *llc
) {
121 return htons(FLOW_DL_TYPE_NONE
);
125 if (llc
->llc
.llc_dsap
!= LLC_DSAP_SNAP
126 || llc
->llc
.llc_ssap
!= LLC_SSAP_SNAP
127 || llc
->llc
.llc_cntl
!= LLC_CNTL_SNAP
128 || memcmp(llc
->snap
.snap_org
, SNAP_ORG_ETHERNET
,
129 sizeof llc
->snap
.snap_org
)) {
130 return htons(FLOW_DL_TYPE_NONE
);
133 ofpbuf_pull(b
, sizeof *llc
);
134 return llc
->snap
.snap_type
;
138 parse_ipv6(struct ofpbuf
*packet
, struct flow
*flow
)
140 const struct ip6_hdr
*nh
;
144 nh
= ofpbuf_try_pull(packet
, sizeof *nh
);
149 nexthdr
= nh
->ip6_nxt
;
151 flow
->ipv6_src
= nh
->ip6_src
;
152 flow
->ipv6_dst
= nh
->ip6_dst
;
154 tc_flow
= get_unaligned_be32(&nh
->ip6_flow
);
155 flow
->nw_tos
= ntohl(tc_flow
) >> 20;
156 flow
->ipv6_label
= tc_flow
& htonl(IPV6_LABEL_MASK
);
157 flow
->nw_ttl
= nh
->ip6_hlim
;
158 flow
->nw_proto
= IPPROTO_NONE
;
161 if ((nexthdr
!= IPPROTO_HOPOPTS
)
162 && (nexthdr
!= IPPROTO_ROUTING
)
163 && (nexthdr
!= IPPROTO_DSTOPTS
)
164 && (nexthdr
!= IPPROTO_AH
)
165 && (nexthdr
!= IPPROTO_FRAGMENT
)) {
166 /* It's either a terminal header (e.g., TCP, UDP) or one we
167 * don't understand. In either case, we're done with the
168 * packet, so use it to fill in 'nw_proto'. */
172 /* We only verify that at least 8 bytes of the next header are
173 * available, but many of these headers are longer. Ensure that
174 * accesses within the extension header are within those first 8
175 * bytes. All extension headers are required to be at least 8
177 if (packet
->size
< 8) {
181 if ((nexthdr
== IPPROTO_HOPOPTS
)
182 || (nexthdr
== IPPROTO_ROUTING
)
183 || (nexthdr
== IPPROTO_DSTOPTS
)) {
184 /* These headers, while different, have the fields we care about
185 * in the same location and with the same interpretation. */
186 const struct ip6_ext
*ext_hdr
= packet
->data
;
187 nexthdr
= ext_hdr
->ip6e_nxt
;
188 if (!ofpbuf_try_pull(packet
, (ext_hdr
->ip6e_len
+ 1) * 8)) {
191 } else if (nexthdr
== IPPROTO_AH
) {
192 /* A standard AH definition isn't available, but the fields
193 * we care about are in the same location as the generic
194 * option header--only the header length is calculated
196 const struct ip6_ext
*ext_hdr
= packet
->data
;
197 nexthdr
= ext_hdr
->ip6e_nxt
;
198 if (!ofpbuf_try_pull(packet
, (ext_hdr
->ip6e_len
+ 2) * 4)) {
201 } else if (nexthdr
== IPPROTO_FRAGMENT
) {
202 const struct ip6_frag
*frag_hdr
= packet
->data
;
204 nexthdr
= frag_hdr
->ip6f_nxt
;
205 if (!ofpbuf_try_pull(packet
, sizeof *frag_hdr
)) {
209 /* We only process the first fragment. */
210 if (frag_hdr
->ip6f_offlg
!= htons(0)) {
211 if ((frag_hdr
->ip6f_offlg
& IP6F_OFF_MASK
) == htons(0)) {
212 flow
->nw_frag
= FLOW_NW_FRAG_ANY
;
214 flow
->nw_frag
|= FLOW_NW_FRAG_LATER
;
215 nexthdr
= IPPROTO_FRAGMENT
;
222 flow
->nw_proto
= nexthdr
;
227 parse_tcp(struct ofpbuf
*packet
, struct ofpbuf
*b
, struct flow
*flow
)
229 const struct tcp_header
*tcp
= pull_tcp(b
);
231 flow
->tp_src
= tcp
->tcp_src
;
232 flow
->tp_dst
= tcp
->tcp_dst
;
233 packet
->l7
= b
->data
;
238 parse_udp(struct ofpbuf
*packet
, struct ofpbuf
*b
, struct flow
*flow
)
240 const struct udp_header
*udp
= pull_udp(b
);
242 flow
->tp_src
= udp
->udp_src
;
243 flow
->tp_dst
= udp
->udp_dst
;
244 packet
->l7
= b
->data
;
249 parse_icmpv6(struct ofpbuf
*b
, struct flow
*flow
)
251 const struct icmp6_hdr
*icmp
= pull_icmpv6(b
);
257 /* The ICMPv6 type and code fields use the 16-bit transport port
258 * fields, so we need to store them in 16-bit network byte order. */
259 flow
->tp_src
= htons(icmp
->icmp6_type
);
260 flow
->tp_dst
= htons(icmp
->icmp6_code
);
262 if (icmp
->icmp6_code
== 0 &&
263 (icmp
->icmp6_type
== ND_NEIGHBOR_SOLICIT
||
264 icmp
->icmp6_type
== ND_NEIGHBOR_ADVERT
)) {
265 const struct in6_addr
*nd_target
;
267 nd_target
= ofpbuf_try_pull(b
, sizeof *nd_target
);
271 flow
->nd_target
= *nd_target
;
273 while (b
->size
>= 8) {
274 /* The minimum size of an option is 8 bytes, which also is
275 * the size of Ethernet link-layer options. */
276 const struct nd_opt_hdr
*nd_opt
= b
->data
;
277 int opt_len
= nd_opt
->nd_opt_len
* 8;
279 if (!opt_len
|| opt_len
> b
->size
) {
283 /* Store the link layer address if the appropriate option is
284 * provided. It is considered an error if the same link
285 * layer option is specified twice. */
286 if (nd_opt
->nd_opt_type
== ND_OPT_SOURCE_LINKADDR
288 if (eth_addr_is_zero(flow
->arp_sha
)) {
289 memcpy(flow
->arp_sha
, nd_opt
+ 1, ETH_ADDR_LEN
);
293 } else if (nd_opt
->nd_opt_type
== ND_OPT_TARGET_LINKADDR
295 if (eth_addr_is_zero(flow
->arp_tha
)) {
296 memcpy(flow
->arp_tha
, nd_opt
+ 1, ETH_ADDR_LEN
);
302 if (!ofpbuf_try_pull(b
, opt_len
)) {
311 memset(&flow
->nd_target
, 0, sizeof(flow
->nd_target
));
312 memset(flow
->arp_sha
, 0, sizeof(flow
->arp_sha
));
313 memset(flow
->arp_tha
, 0, sizeof(flow
->arp_tha
));
319 /* Initializes 'flow' members from 'packet', 'skb_priority', 'tun_id', and
322 * Initializes 'packet' header pointers as follows:
324 * - packet->l2 to the start of the Ethernet header.
326 * - packet->l3 to just past the Ethernet header, or just past the
327 * vlan_header if one is present, to the first byte of the payload of the
330 * - packet->l4 to just past the IPv4 header, if one is present and has a
331 * correct length, and otherwise NULL.
333 * - packet->l7 to just past the TCP or UDP or ICMP header, if one is
334 * present and has a correct length, and otherwise NULL.
337 flow_extract(struct ofpbuf
*packet
, uint32_t skb_priority
, ovs_be64 tun_id
,
338 uint16_t ofp_in_port
, struct flow
*flow
)
340 struct ofpbuf b
= *packet
;
341 struct eth_header
*eth
;
343 COVERAGE_INC(flow_extract
);
345 memset(flow
, 0, sizeof *flow
);
346 flow
->tun_id
= tun_id
;
347 flow
->in_port
= ofp_in_port
;
348 flow
->skb_priority
= skb_priority
;
355 if (b
.size
< sizeof *eth
) {
361 memcpy(flow
->dl_src
, eth
->eth_src
, ETH_ADDR_LEN
);
362 memcpy(flow
->dl_dst
, eth
->eth_dst
, ETH_ADDR_LEN
);
364 /* dl_type, vlan_tci. */
365 ofpbuf_pull(&b
, ETH_ADDR_LEN
* 2);
366 if (eth
->eth_type
== htons(ETH_TYPE_VLAN
)) {
367 parse_vlan(&b
, flow
);
369 flow
->dl_type
= parse_ethertype(&b
);
373 if (flow
->dl_type
== htons(ETH_TYPE_IP
)) {
374 const struct ip_header
*nh
= pull_ip(&b
);
378 flow
->nw_src
= get_unaligned_be32(&nh
->ip_src
);
379 flow
->nw_dst
= get_unaligned_be32(&nh
->ip_dst
);
380 flow
->nw_proto
= nh
->ip_proto
;
382 flow
->nw_tos
= nh
->ip_tos
;
383 if (IP_IS_FRAGMENT(nh
->ip_frag_off
)) {
384 flow
->nw_frag
= FLOW_NW_FRAG_ANY
;
385 if (nh
->ip_frag_off
& htons(IP_FRAG_OFF_MASK
)) {
386 flow
->nw_frag
|= FLOW_NW_FRAG_LATER
;
389 flow
->nw_ttl
= nh
->ip_ttl
;
391 if (!(nh
->ip_frag_off
& htons(IP_FRAG_OFF_MASK
))) {
392 if (flow
->nw_proto
== IPPROTO_TCP
) {
393 parse_tcp(packet
, &b
, flow
);
394 } else if (flow
->nw_proto
== IPPROTO_UDP
) {
395 parse_udp(packet
, &b
, flow
);
396 } else if (flow
->nw_proto
== IPPROTO_ICMP
) {
397 const struct icmp_header
*icmp
= pull_icmp(&b
);
399 flow
->tp_src
= htons(icmp
->icmp_type
);
400 flow
->tp_dst
= htons(icmp
->icmp_code
);
406 } else if (flow
->dl_type
== htons(ETH_TYPE_IPV6
)) {
407 if (parse_ipv6(&b
, flow
)) {
412 if (flow
->nw_proto
== IPPROTO_TCP
) {
413 parse_tcp(packet
, &b
, flow
);
414 } else if (flow
->nw_proto
== IPPROTO_UDP
) {
415 parse_udp(packet
, &b
, flow
);
416 } else if (flow
->nw_proto
== IPPROTO_ICMPV6
) {
417 if (parse_icmpv6(&b
, flow
)) {
421 } else if (flow
->dl_type
== htons(ETH_TYPE_ARP
)) {
422 const struct arp_eth_header
*arp
= pull_arp(&b
);
423 if (arp
&& arp
->ar_hrd
== htons(1)
424 && arp
->ar_pro
== htons(ETH_TYPE_IP
)
425 && arp
->ar_hln
== ETH_ADDR_LEN
426 && arp
->ar_pln
== 4) {
427 /* We only match on the lower 8 bits of the opcode. */
428 if (ntohs(arp
->ar_op
) <= 0xff) {
429 flow
->nw_proto
= ntohs(arp
->ar_op
);
432 if ((flow
->nw_proto
== ARP_OP_REQUEST
)
433 || (flow
->nw_proto
== ARP_OP_REPLY
)) {
434 flow
->nw_src
= arp
->ar_spa
;
435 flow
->nw_dst
= arp
->ar_tpa
;
436 memcpy(flow
->arp_sha
, arp
->ar_sha
, ETH_ADDR_LEN
);
437 memcpy(flow
->arp_tha
, arp
->ar_tha
, ETH_ADDR_LEN
);
443 /* For every bit of a field that is wildcarded in 'wildcards', sets the
444 * corresponding bit in 'flow' to zero. */
446 flow_zero_wildcards(struct flow
*flow
, const struct flow_wildcards
*wildcards
)
448 uint32_t *flow_u32
= (uint32_t *) flow
;
449 const uint32_t *wc_u32
= (const uint32_t *) &wildcards
->masks
;
452 for (i
= 0; i
< FLOW_U32S
; i
++) {
453 flow_u32
[i
] &= wc_u32
[i
];
457 /* Initializes 'fmd' with the metadata found in 'flow'. */
459 flow_get_metadata(const struct flow
*flow
, struct flow_metadata
*fmd
)
461 BUILD_ASSERT_DECL(FLOW_WC_SEQ
== 17);
463 fmd
->tun_id
= flow
->tun_id
;
464 fmd
->metadata
= flow
->metadata
;
465 memcpy(fmd
->regs
, flow
->regs
, sizeof fmd
->regs
);
466 fmd
->in_port
= flow
->in_port
;
470 flow_to_string(const struct flow
*flow
)
472 struct ds ds
= DS_EMPTY_INITIALIZER
;
473 flow_format(&ds
, flow
);
478 flow_format(struct ds
*ds
, const struct flow
*flow
)
480 ds_put_format(ds
, "priority:%"PRIu32
483 ",in_port:%04"PRIx16
,
485 ntohll(flow
->tun_id
),
486 ntohll(flow
->metadata
),
489 ds_put_format(ds
, ",tci(");
490 if (flow
->vlan_tci
) {
491 ds_put_format(ds
, "vlan:%"PRIu16
",pcp:%d",
492 vlan_tci_to_vid(flow
->vlan_tci
),
493 vlan_tci_to_pcp(flow
->vlan_tci
));
495 ds_put_char(ds
, '0');
497 ds_put_format(ds
, ") mac("ETH_ADDR_FMT
"->"ETH_ADDR_FMT
499 ETH_ADDR_ARGS(flow
->dl_src
),
500 ETH_ADDR_ARGS(flow
->dl_dst
),
501 ntohs(flow
->dl_type
));
503 if (flow
->dl_type
== htons(ETH_TYPE_IPV6
)) {
504 ds_put_format(ds
, " label:%#"PRIx32
" proto:%"PRIu8
" tos:%#"PRIx8
505 " ttl:%"PRIu8
" ipv6(",
506 ntohl(flow
->ipv6_label
), flow
->nw_proto
,
507 flow
->nw_tos
, flow
->nw_ttl
);
508 print_ipv6_addr(ds
, &flow
->ipv6_src
);
509 ds_put_cstr(ds
, "->");
510 print_ipv6_addr(ds
, &flow
->ipv6_dst
);
511 ds_put_char(ds
, ')');
512 } else if (flow
->dl_type
== htons(ETH_TYPE_IP
) ||
513 flow
->dl_type
== htons(ETH_TYPE_ARP
)) {
514 ds_put_format(ds
, " proto:%"PRIu8
" tos:%#"PRIx8
" ttl:%"PRIu8
515 " ip("IP_FMT
"->"IP_FMT
")",
516 flow
->nw_proto
, flow
->nw_tos
, flow
->nw_ttl
,
517 IP_ARGS(&flow
->nw_src
), IP_ARGS(&flow
->nw_dst
));
520 ds_put_format(ds
, " frag(%s)",
521 flow
->nw_frag
== FLOW_NW_FRAG_ANY
? "first"
522 : flow
->nw_frag
== (FLOW_NW_FRAG_ANY
| FLOW_NW_FRAG_LATER
)
523 ? "later" : "<error>");
525 if (flow
->tp_src
|| flow
->tp_dst
) {
526 ds_put_format(ds
, " port(%"PRIu16
"->%"PRIu16
")",
527 ntohs(flow
->tp_src
), ntohs(flow
->tp_dst
));
529 if (!eth_addr_is_zero(flow
->arp_sha
) || !eth_addr_is_zero(flow
->arp_tha
)) {
530 ds_put_format(ds
, " arp_ha("ETH_ADDR_FMT
"->"ETH_ADDR_FMT
")",
531 ETH_ADDR_ARGS(flow
->arp_sha
),
532 ETH_ADDR_ARGS(flow
->arp_tha
));
537 flow_print(FILE *stream
, const struct flow
*flow
)
539 char *s
= flow_to_string(flow
);
544 /* flow_wildcards functions. */
546 /* Initializes 'wc' as a set of wildcards that matches every packet. */
548 flow_wildcards_init_catchall(struct flow_wildcards
*wc
)
550 memset(&wc
->masks
, 0, sizeof wc
->masks
);
553 /* Initializes 'wc' as an exact-match set of wildcards; that is, 'wc' does not
554 * wildcard any bits or fields. */
556 flow_wildcards_init_exact(struct flow_wildcards
*wc
)
558 memset(&wc
->masks
, 0xff, sizeof wc
->masks
);
559 memset(wc
->masks
.zeros
, 0, sizeof wc
->masks
.zeros
);
562 /* Returns true if 'wc' matches every packet, false if 'wc' fixes any bits or
565 flow_wildcards_is_catchall(const struct flow_wildcards
*wc
)
567 const uint32_t *wc_u32
= (const uint32_t *) &wc
->masks
;
570 for (i
= 0; i
< FLOW_U32S
; i
++) {
578 /* Initializes 'dst' as the combination of wildcards in 'src1' and 'src2'.
579 * That is, a bit or a field is wildcarded in 'dst' if it is wildcarded in
580 * 'src1' or 'src2' or both. */
582 flow_wildcards_combine(struct flow_wildcards
*dst
,
583 const struct flow_wildcards
*src1
,
584 const struct flow_wildcards
*src2
)
586 uint32_t *dst_u32
= (uint32_t *) &dst
->masks
;
587 const uint32_t *src1_u32
= (const uint32_t *) &src1
->masks
;
588 const uint32_t *src2_u32
= (const uint32_t *) &src2
->masks
;
591 for (i
= 0; i
< FLOW_U32S
; i
++) {
592 dst_u32
[i
] = src1_u32
[i
] & src2_u32
[i
];
596 /* Returns a hash of the wildcards in 'wc'. */
598 flow_wildcards_hash(const struct flow_wildcards
*wc
, uint32_t basis
)
600 return flow_hash(&wc
->masks
, 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
)
609 return flow_equal(&a
->masks
, &b
->masks
);
612 /* Returns true if at least one bit or field is wildcarded in 'a' but not in
613 * 'b', false otherwise. */
615 flow_wildcards_has_extra(const struct flow_wildcards
*a
,
616 const struct flow_wildcards
*b
)
618 const uint32_t *a_u32
= (const uint32_t *) &a
->masks
;
619 const uint32_t *b_u32
= (const uint32_t *) &b
->masks
;
622 for (i
= 0; i
< FLOW_U32S
; i
++) {
623 if ((a_u32
[i
] & b_u32
[i
]) != b_u32
[i
]) {
630 /* Returns true if 'a' and 'b' are equal, except that 0-bits (wildcarded bits)
631 * in 'wc' do not need to be equal in 'a' and 'b'. */
633 flow_equal_except(const struct flow
*a
, const struct flow
*b
,
634 const struct flow_wildcards
*wc
)
636 const uint32_t *a_u32
= (const uint32_t *) a
;
637 const uint32_t *b_u32
= (const uint32_t *) b
;
638 const uint32_t *wc_u32
= (const uint32_t *) &wc
->masks
;
641 for (i
= 0; i
< FLOW_U32S
; i
++) {
642 if ((a_u32
[i
] ^ b_u32
[i
]) & wc_u32
[i
]) {
649 /* Sets the wildcard mask for register 'idx' in 'wc' to 'mask'.
650 * (A 0-bit indicates a wildcard bit.) */
652 flow_wildcards_set_reg_mask(struct flow_wildcards
*wc
, int idx
, uint32_t mask
)
654 wc
->masks
.regs
[idx
] = mask
;
657 /* Hashes 'flow' based on its L2 through L4 protocol information. */
659 flow_hash_symmetric_l4(const struct flow
*flow
, uint32_t basis
)
664 struct in6_addr ipv6_addr
;
669 uint8_t eth_addr
[ETH_ADDR_LEN
];
675 memset(&fields
, 0, sizeof fields
);
676 for (i
= 0; i
< ETH_ADDR_LEN
; i
++) {
677 fields
.eth_addr
[i
] = flow
->dl_src
[i
] ^ flow
->dl_dst
[i
];
679 fields
.vlan_tci
= flow
->vlan_tci
& htons(VLAN_VID_MASK
);
680 fields
.eth_type
= flow
->dl_type
;
682 /* UDP source and destination port are not taken into account because they
683 * will not necessarily be symmetric in a bidirectional flow. */
684 if (fields
.eth_type
== htons(ETH_TYPE_IP
)) {
685 fields
.ipv4_addr
= flow
->nw_src
^ flow
->nw_dst
;
686 fields
.ip_proto
= flow
->nw_proto
;
687 if (fields
.ip_proto
== IPPROTO_TCP
) {
688 fields
.tp_port
= flow
->tp_src
^ flow
->tp_dst
;
690 } else if (fields
.eth_type
== htons(ETH_TYPE_IPV6
)) {
691 const uint8_t *a
= &flow
->ipv6_src
.s6_addr
[0];
692 const uint8_t *b
= &flow
->ipv6_dst
.s6_addr
[0];
693 uint8_t *ipv6_addr
= &fields
.ipv6_addr
.s6_addr
[0];
695 for (i
=0; i
<16; i
++) {
696 ipv6_addr
[i
] = a
[i
] ^ b
[i
];
698 fields
.ip_proto
= flow
->nw_proto
;
699 if (fields
.ip_proto
== IPPROTO_TCP
) {
700 fields
.tp_port
= flow
->tp_src
^ flow
->tp_dst
;
703 return hash_bytes(&fields
, sizeof fields
, basis
);
706 /* Hashes the portions of 'flow' designated by 'fields'. */
708 flow_hash_fields(const struct flow
*flow
, enum nx_hash_fields fields
,
713 case NX_HASH_FIELDS_ETH_SRC
:
714 return hash_bytes(flow
->dl_src
, sizeof flow
->dl_src
, basis
);
716 case NX_HASH_FIELDS_SYMMETRIC_L4
:
717 return flow_hash_symmetric_l4(flow
, basis
);
723 /* Returns a string representation of 'fields'. */
725 flow_hash_fields_to_str(enum nx_hash_fields fields
)
728 case NX_HASH_FIELDS_ETH_SRC
: return "eth_src";
729 case NX_HASH_FIELDS_SYMMETRIC_L4
: return "symmetric_l4";
730 default: return "<unknown>";
734 /* Returns true if the value of 'fields' is supported. Otherwise false. */
736 flow_hash_fields_valid(enum nx_hash_fields fields
)
738 return fields
== NX_HASH_FIELDS_ETH_SRC
739 || fields
== NX_HASH_FIELDS_SYMMETRIC_L4
;
742 /* Sets the VLAN VID that 'flow' matches to 'vid', which is interpreted as an
743 * OpenFlow 1.0 "dl_vlan" value:
745 * - If it is in the range 0...4095, 'flow->vlan_tci' is set to match
746 * that VLAN. Any existing PCP match is unchanged (it becomes 0 if
747 * 'flow' previously matched packets without a VLAN header).
749 * - If it is OFP_VLAN_NONE, 'flow->vlan_tci' is set to match a packet
750 * without a VLAN tag.
752 * - Other values of 'vid' should not be used. */
754 flow_set_dl_vlan(struct flow
*flow
, ovs_be16 vid
)
756 if (vid
== htons(OFP10_VLAN_NONE
)) {
757 flow
->vlan_tci
= htons(0);
759 vid
&= htons(VLAN_VID_MASK
);
760 flow
->vlan_tci
&= ~htons(VLAN_VID_MASK
);
761 flow
->vlan_tci
|= htons(VLAN_CFI
) | vid
;
765 /* Sets the VLAN VID that 'flow' matches to 'vid', which is interpreted as an
766 * OpenFlow 1.2 "vlan_vid" value, that is, the low 13 bits of 'vlan_tci' (VID
769 flow_set_vlan_vid(struct flow
*flow
, ovs_be16 vid
)
771 ovs_be16 mask
= htons(VLAN_VID_MASK
| VLAN_CFI
);
772 flow
->vlan_tci
&= ~mask
;
773 flow
->vlan_tci
|= vid
& mask
;
776 /* Sets the VLAN PCP that 'flow' matches to 'pcp', which should be in the
779 * This function has no effect on the VLAN ID that 'flow' matches.
781 * After calling this function, 'flow' will not match packets without a VLAN
784 flow_set_vlan_pcp(struct flow
*flow
, uint8_t pcp
)
787 flow
->vlan_tci
&= ~htons(VLAN_PCP_MASK
);
788 flow
->vlan_tci
|= htons((pcp
<< VLAN_PCP_SHIFT
) | VLAN_CFI
);
791 /* Puts into 'b' a packet that flow_extract() would parse as having the given
794 * (This is useful only for testing, obviously, and the packet isn't really
795 * valid. It hasn't got some checksums filled in, for one, and lots of fields
796 * are just zeroed.) */
798 flow_compose(struct ofpbuf
*b
, const struct flow
*flow
)
800 eth_compose(b
, flow
->dl_dst
, flow
->dl_src
, ntohs(flow
->dl_type
), 0);
801 if (flow
->dl_type
== htons(FLOW_DL_TYPE_NONE
)) {
802 struct eth_header
*eth
= b
->l2
;
803 eth
->eth_type
= htons(b
->size
);
807 if (flow
->vlan_tci
& htons(VLAN_CFI
)) {
808 eth_push_vlan(b
, flow
->vlan_tci
);
811 if (flow
->dl_type
== htons(ETH_TYPE_IP
)) {
812 struct ip_header
*ip
;
814 b
->l3
= ip
= ofpbuf_put_zeros(b
, sizeof *ip
);
815 ip
->ip_ihl_ver
= IP_IHL_VER(5, 4);
816 ip
->ip_tos
= flow
->nw_tos
;
817 ip
->ip_proto
= flow
->nw_proto
;
818 ip
->ip_src
= flow
->nw_src
;
819 ip
->ip_dst
= flow
->nw_dst
;
821 if (flow
->nw_frag
& FLOW_NW_FRAG_ANY
) {
822 ip
->ip_frag_off
|= htons(IP_MORE_FRAGMENTS
);
823 if (flow
->nw_frag
& FLOW_NW_FRAG_LATER
) {
824 ip
->ip_frag_off
|= htons(100);
827 if (!(flow
->nw_frag
& FLOW_NW_FRAG_ANY
)
828 || !(flow
->nw_frag
& FLOW_NW_FRAG_LATER
)) {
829 if (flow
->nw_proto
== IPPROTO_TCP
) {
830 struct tcp_header
*tcp
;
832 b
->l4
= tcp
= ofpbuf_put_zeros(b
, sizeof *tcp
);
833 tcp
->tcp_src
= flow
->tp_src
;
834 tcp
->tcp_dst
= flow
->tp_dst
;
835 tcp
->tcp_ctl
= TCP_CTL(0, 5);
836 } else if (flow
->nw_proto
== IPPROTO_UDP
) {
837 struct udp_header
*udp
;
839 b
->l4
= udp
= ofpbuf_put_zeros(b
, sizeof *udp
);
840 udp
->udp_src
= flow
->tp_src
;
841 udp
->udp_dst
= flow
->tp_dst
;
842 } else if (flow
->nw_proto
== IPPROTO_ICMP
) {
843 struct icmp_header
*icmp
;
845 b
->l4
= icmp
= ofpbuf_put_zeros(b
, sizeof *icmp
);
846 icmp
->icmp_type
= ntohs(flow
->tp_src
);
847 icmp
->icmp_code
= ntohs(flow
->tp_dst
);
848 icmp
->icmp_csum
= csum(icmp
, ICMP_HEADER_LEN
);
853 ip
->ip_tot_len
= htons((uint8_t *) b
->data
+ b
->size
854 - (uint8_t *) b
->l3
);
855 ip
->ip_csum
= csum(ip
, sizeof *ip
);
856 } else if (flow
->dl_type
== htons(ETH_TYPE_IPV6
)) {
858 } else if (flow
->dl_type
== htons(ETH_TYPE_ARP
)) {
859 struct arp_eth_header
*arp
;
861 b
->l3
= arp
= ofpbuf_put_zeros(b
, sizeof *arp
);
862 arp
->ar_hrd
= htons(1);
863 arp
->ar_pro
= htons(ETH_TYPE_IP
);
864 arp
->ar_hln
= ETH_ADDR_LEN
;
866 arp
->ar_op
= htons(flow
->nw_proto
);
868 if (flow
->nw_proto
== ARP_OP_REQUEST
||
869 flow
->nw_proto
== ARP_OP_REPLY
) {
870 arp
->ar_spa
= flow
->nw_src
;
871 arp
->ar_tpa
= flow
->nw_dst
;
872 memcpy(arp
->ar_sha
, flow
->arp_sha
, ETH_ADDR_LEN
);
873 memcpy(arp
->ar_tha
, flow
->arp_tha
, ETH_ADDR_LEN
);
878 /* Compressed flow. */
881 miniflow_n_values(const struct miniflow
*flow
)
886 for (i
= 0; i
< MINI_N_MAPS
; i
++) {
887 n
+= popcount(flow
->map
[i
]);
893 miniflow_alloc_values(struct miniflow
*flow
, int n
)
895 if (n
<= MINI_N_INLINE
) {
896 return flow
->inline_values
;
898 COVERAGE_INC(miniflow_malloc
);
899 return xmalloc(n
* sizeof *flow
->values
);
903 /* Initializes 'dst' as a copy of 'src'. The caller must eventually free 'dst'
904 * with miniflow_destroy(). */
906 miniflow_init(struct miniflow
*dst
, const struct flow
*src
)
908 const uint32_t *src_u32
= (const uint32_t *) src
;
913 /* Initialize dst->map, counting the number of nonzero elements. */
915 memset(dst
->map
, 0, sizeof dst
->map
);
916 for (i
= 0; i
< FLOW_U32S
; i
++) {
918 dst
->map
[i
/ 32] |= 1u << (i
% 32);
923 /* Initialize dst->values. */
924 dst
->values
= miniflow_alloc_values(dst
, n
);
926 for (i
= 0; i
< MINI_N_MAPS
; i
++) {
929 for (map
= dst
->map
[i
]; map
; map
= zero_rightmost_1bit(map
)) {
930 dst
->values
[ofs
++] = src_u32
[raw_ctz(map
) + i
* 32];
935 /* Initializes 'dst' as a copy of 'src'. The caller must eventually free 'dst'
936 * with miniflow_destroy(). */
938 miniflow_clone(struct miniflow
*dst
, const struct miniflow
*src
)
940 int n
= miniflow_n_values(src
);
941 memcpy(dst
->map
, src
->map
, sizeof dst
->map
);
942 dst
->values
= miniflow_alloc_values(dst
, n
);
943 memcpy(dst
->values
, src
->values
, n
* sizeof *dst
->values
);
946 /* Frees any memory owned by 'flow'. Does not free the storage in which 'flow'
947 * itself resides; the caller is responsible for that. */
949 miniflow_destroy(struct miniflow
*flow
)
951 if (flow
->values
!= flow
->inline_values
) {
956 /* Initializes 'dst' as a copy of 'src'. */
958 miniflow_expand(const struct miniflow
*src
, struct flow
*dst
)
960 uint32_t *dst_u32
= (uint32_t *) dst
;
964 memset(dst_u32
, 0, sizeof *dst
);
967 for (i
= 0; i
< MINI_N_MAPS
; i
++) {
970 for (map
= src
->map
[i
]; map
; map
= zero_rightmost_1bit(map
)) {
971 dst_u32
[raw_ctz(map
) + i
* 32] = src
->values
[ofs
++];
976 static const uint32_t *
977 miniflow_get__(const struct miniflow
*flow
, unsigned int u32_ofs
)
979 if (!(flow
->map
[u32_ofs
/ 32] & (1u << (u32_ofs
% 32)))) {
980 static const uint32_t zero
= 0;
983 const uint32_t *p
= flow
->values
;
985 BUILD_ASSERT(MINI_N_MAPS
== 2);
987 p
+= popcount(flow
->map
[0] & ((1u << u32_ofs
) - 1));
989 p
+= popcount(flow
->map
[0]);
990 p
+= popcount(flow
->map
[1] & ((1u << (u32_ofs
- 32)) - 1));
996 /* Returns the uint32_t that would be at byte offset '4 * u32_ofs' if 'flow'
997 * were expanded into a "struct flow". */
999 miniflow_get(const struct miniflow
*flow
, unsigned int u32_ofs
)
1001 return *miniflow_get__(flow
, u32_ofs
);
1004 /* Returns the ovs_be16 that would be at byte offset 'u8_ofs' if 'flow' were
1005 * expanded into a "struct flow". */
1007 miniflow_get_be16(const struct miniflow
*flow
, unsigned int u8_ofs
)
1009 const uint32_t *u32p
= miniflow_get__(flow
, u8_ofs
/ 4);
1010 const ovs_be16
*be16p
= (const ovs_be16
*) u32p
;
1011 return be16p
[u8_ofs
% 4 != 0];
1014 /* Returns the VID within the vlan_tci member of the "struct flow" represented
1017 miniflow_get_vid(const struct miniflow
*flow
)
1019 ovs_be16 tci
= miniflow_get_be16(flow
, offsetof(struct flow
, vlan_tci
));
1020 return vlan_tci_to_vid(tci
);
1023 /* Returns true if 'a' and 'b' are the same flow, false otherwise. */
1025 miniflow_equal(const struct miniflow
*a
, const struct miniflow
*b
)
1029 for (i
= 0; i
< MINI_N_MAPS
; i
++) {
1030 if (a
->map
[i
] != b
->map
[i
]) {
1035 return !memcmp(a
->values
, b
->values
,
1036 miniflow_n_values(a
) * sizeof *a
->values
);
1039 /* Returns true if 'a' and 'b' are equal at the places where there are 1-bits
1040 * in 'mask', false if they differ. */
1042 miniflow_equal_in_minimask(const struct miniflow
*a
, const struct miniflow
*b
,
1043 const struct minimask
*mask
)
1048 p
= mask
->masks
.values
;
1049 for (i
= 0; i
< MINI_N_MAPS
; i
++) {
1052 for (map
= mask
->masks
.map
[i
]; map
; map
= zero_rightmost_1bit(map
)) {
1053 int ofs
= raw_ctz(map
) + i
* 32;
1055 if ((miniflow_get(a
, ofs
) ^ miniflow_get(b
, ofs
)) & *p
) {
1065 /* Returns true if 'a' and 'b' are equal at the places where there are 1-bits
1066 * in 'mask', false if they differ. */
1068 miniflow_equal_flow_in_minimask(const struct miniflow
*a
, const struct flow
*b
,
1069 const struct minimask
*mask
)
1071 const uint32_t *b_u32
= (const uint32_t *) b
;
1075 p
= mask
->masks
.values
;
1076 for (i
= 0; i
< MINI_N_MAPS
; i
++) {
1079 for (map
= mask
->masks
.map
[i
]; map
; map
= zero_rightmost_1bit(map
)) {
1080 int ofs
= raw_ctz(map
) + i
* 32;
1082 if ((miniflow_get(a
, ofs
) ^ b_u32
[ofs
]) & *p
) {
1092 /* Returns a hash value for 'flow', given 'basis'. */
1094 miniflow_hash(const struct miniflow
*flow
, uint32_t basis
)
1096 BUILD_ASSERT_DECL(MINI_N_MAPS
== 2);
1097 return hash_3words(flow
->map
[0], flow
->map
[1],
1098 hash_words(flow
->values
, miniflow_n_values(flow
),
1102 /* Returns a hash value for the bits of 'flow' where there are 1-bits in
1103 * 'mask', given 'basis'.
1105 * The hash values returned by this function are the same as those returned by
1106 * flow_hash_in_minimask(), only the form of the arguments differ. */
1108 miniflow_hash_in_minimask(const struct miniflow
*flow
,
1109 const struct minimask
*mask
, uint32_t basis
)
1111 const uint32_t *p
= mask
->masks
.values
;
1116 for (i
= 0; i
< MINI_N_MAPS
; i
++) {
1119 for (map
= mask
->masks
.map
[i
]; map
; map
= zero_rightmost_1bit(map
)) {
1120 int ofs
= raw_ctz(map
) + i
* 32;
1122 hash
= mhash_add(hash
, miniflow_get(flow
, ofs
) & *p
);
1127 return mhash_finish(hash
, p
- mask
->masks
.values
);
1130 /* Returns a hash value for the bits of 'flow' where there are 1-bits in
1131 * 'mask', given 'basis'.
1133 * The hash values returned by this function are the same as those returned by
1134 * miniflow_hash_in_minimask(), only the form of the arguments differ. */
1136 flow_hash_in_minimask(const struct flow
*flow
, const struct minimask
*mask
,
1139 const uint32_t *flow_u32
= (const uint32_t *) flow
;
1140 const uint32_t *p
= mask
->masks
.values
;
1145 for (i
= 0; i
< MINI_N_MAPS
; i
++) {
1148 for (map
= mask
->masks
.map
[i
]; map
; map
= zero_rightmost_1bit(map
)) {
1149 int ofs
= raw_ctz(map
) + i
* 32;
1151 hash
= mhash_add(hash
, flow_u32
[ofs
] & *p
);
1156 return mhash_finish(hash
, p
- mask
->masks
.values
);
1159 /* Initializes 'dst' as a copy of 'src'. The caller must eventually free 'dst'
1160 * with minimask_destroy(). */
1162 minimask_init(struct minimask
*mask
, const struct flow_wildcards
*wc
)
1164 miniflow_init(&mask
->masks
, &wc
->masks
);
1167 /* Initializes 'dst' as a copy of 'src'. The caller must eventually free 'dst'
1168 * with minimask_destroy(). */
1170 minimask_clone(struct minimask
*dst
, const struct minimask
*src
)
1172 miniflow_clone(&dst
->masks
, &src
->masks
);
1175 /* Initializes 'dst_' as the bit-wise "and" of 'a_' and 'b_'.
1177 * The caller must provide room for FLOW_U32S "uint32_t"s in 'storage', for use
1178 * by 'dst_'. The caller must *not* free 'dst_' with minimask_destroy(). */
1180 minimask_combine(struct minimask
*dst_
,
1181 const struct minimask
*a_
, const struct minimask
*b_
,
1182 uint32_t storage
[FLOW_U32S
])
1184 struct miniflow
*dst
= &dst_
->masks
;
1185 const struct miniflow
*a
= &a_
->masks
;
1186 const struct miniflow
*b
= &b_
->masks
;
1190 dst
->values
= storage
;
1191 for (i
= 0; i
< MINI_N_MAPS
; i
++) {
1195 for (map
= a
->map
[i
] & b
->map
[i
]; map
;
1196 map
= zero_rightmost_1bit(map
)) {
1197 int ofs
= raw_ctz(map
) + i
* 32;
1198 uint32_t mask
= miniflow_get(a
, ofs
) & miniflow_get(b
, ofs
);
1201 dst
->map
[i
] |= rightmost_1bit(map
);
1202 dst
->values
[n
++] = mask
;
1208 /* Frees any memory owned by 'mask'. Does not free the storage in which 'mask'
1209 * itself resides; the caller is responsible for that. */
1211 minimask_destroy(struct minimask
*mask
)
1213 miniflow_destroy(&mask
->masks
);
1216 /* Initializes 'dst' as a copy of 'src'. */
1218 minimask_expand(const struct minimask
*mask
, struct flow_wildcards
*wc
)
1220 miniflow_expand(&mask
->masks
, &wc
->masks
);
1223 /* Returns the uint32_t that would be at byte offset '4 * u32_ofs' if 'mask'
1224 * were expanded into a "struct flow_wildcards". */
1226 minimask_get(const struct minimask
*mask
, unsigned int u32_ofs
)
1228 return miniflow_get(&mask
->masks
, u32_ofs
);
1231 /* Returns the VID mask within the vlan_tci member of the "struct
1232 * flow_wildcards" represented by 'mask'. */
1234 minimask_get_vid_mask(const struct minimask
*mask
)
1236 return miniflow_get_vid(&mask
->masks
);
1239 /* Returns true if 'a' and 'b' are the same flow mask, false otherwise. */
1241 minimask_equal(const struct minimask
*a
, const struct minimask
*b
)
1243 return miniflow_equal(&a
->masks
, &b
->masks
);
1246 /* Returns a hash value for 'mask', given 'basis'. */
1248 minimask_hash(const struct minimask
*mask
, uint32_t basis
)
1250 return miniflow_hash(&mask
->masks
, basis
);
1253 /* Returns true if at least one bit is wildcarded in 'a_' but not in 'b_',
1254 * false otherwise. */
1256 minimask_has_extra(const struct minimask
*a_
, const struct minimask
*b_
)
1258 const struct miniflow
*a
= &a_
->masks
;
1259 const struct miniflow
*b
= &b_
->masks
;
1262 for (i
= 0; i
< MINI_N_MAPS
; i
++) {
1265 for (map
= a
->map
[i
] | b
->map
[i
]; map
;
1266 map
= zero_rightmost_1bit(map
)) {
1267 int ofs
= raw_ctz(map
) + i
* 32;
1268 uint32_t a_u32
= miniflow_get(a
, ofs
);
1269 uint32_t b_u32
= miniflow_get(b
, ofs
);
1271 if ((a_u32
& b_u32
) != b_u32
) {
1280 /* Returns true if 'mask' matches every packet, false if 'mask' fixes any bits
1283 minimask_is_catchall(const struct minimask
*mask_
)
1285 const struct miniflow
*mask
= &mask_
->masks
;
1287 BUILD_ASSERT(MINI_N_MAPS
== 2);
1288 return !(mask
->map
[0] | mask
->map
[1]);