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"
36 #include "openflow/openflow.h"
38 #include "unaligned.h"
41 VLOG_DEFINE_THIS_MODULE(flow
);
43 COVERAGE_DEFINE(flow_extract
);
44 COVERAGE_DEFINE(miniflow_malloc
);
46 static struct arp_eth_header
*
47 pull_arp(struct ofpbuf
*packet
)
49 return ofpbuf_try_pull(packet
, ARP_ETH_HEADER_LEN
);
52 static struct ip_header
*
53 pull_ip(struct ofpbuf
*packet
)
55 if (packet
->size
>= IP_HEADER_LEN
) {
56 struct ip_header
*ip
= packet
->data
;
57 int ip_len
= IP_IHL(ip
->ip_ihl_ver
) * 4;
58 if (ip_len
>= IP_HEADER_LEN
&& packet
->size
>= ip_len
) {
59 return ofpbuf_pull(packet
, ip_len
);
65 static struct tcp_header
*
66 pull_tcp(struct ofpbuf
*packet
)
68 if (packet
->size
>= TCP_HEADER_LEN
) {
69 struct tcp_header
*tcp
= packet
->data
;
70 int tcp_len
= TCP_OFFSET(tcp
->tcp_ctl
) * 4;
71 if (tcp_len
>= TCP_HEADER_LEN
&& packet
->size
>= tcp_len
) {
72 return ofpbuf_pull(packet
, tcp_len
);
78 static struct udp_header
*
79 pull_udp(struct ofpbuf
*packet
)
81 return ofpbuf_try_pull(packet
, UDP_HEADER_LEN
);
84 static struct icmp_header
*
85 pull_icmp(struct ofpbuf
*packet
)
87 return ofpbuf_try_pull(packet
, ICMP_HEADER_LEN
);
90 static struct icmp6_hdr
*
91 pull_icmpv6(struct ofpbuf
*packet
)
93 return ofpbuf_try_pull(packet
, sizeof(struct icmp6_hdr
));
97 parse_vlan(struct ofpbuf
*b
, struct flow
*flow
)
100 ovs_be16 eth_type
; /* ETH_TYPE_VLAN */
104 if (b
->size
>= sizeof(struct qtag_prefix
) + sizeof(ovs_be16
)) {
105 struct qtag_prefix
*qp
= ofpbuf_pull(b
, sizeof *qp
);
106 flow
->vlan_tci
= qp
->tci
| htons(VLAN_CFI
);
111 parse_ethertype(struct ofpbuf
*b
)
113 struct llc_snap_header
*llc
;
116 proto
= *(ovs_be16
*) ofpbuf_pull(b
, sizeof proto
);
117 if (ntohs(proto
) >= ETH_TYPE_MIN
) {
121 if (b
->size
< sizeof *llc
) {
122 return htons(FLOW_DL_TYPE_NONE
);
126 if (llc
->llc
.llc_dsap
!= LLC_DSAP_SNAP
127 || llc
->llc
.llc_ssap
!= LLC_SSAP_SNAP
128 || llc
->llc
.llc_cntl
!= LLC_CNTL_SNAP
129 || memcmp(llc
->snap
.snap_org
, SNAP_ORG_ETHERNET
,
130 sizeof llc
->snap
.snap_org
)) {
131 return htons(FLOW_DL_TYPE_NONE
);
134 ofpbuf_pull(b
, sizeof *llc
);
135 return llc
->snap
.snap_type
;
139 parse_ipv6(struct ofpbuf
*packet
, struct flow
*flow
)
141 const struct ip6_hdr
*nh
;
145 nh
= ofpbuf_try_pull(packet
, sizeof *nh
);
150 nexthdr
= nh
->ip6_nxt
;
152 flow
->ipv6_src
= nh
->ip6_src
;
153 flow
->ipv6_dst
= nh
->ip6_dst
;
155 tc_flow
= get_unaligned_be32(&nh
->ip6_flow
);
156 flow
->nw_tos
= ntohl(tc_flow
) >> 20;
157 flow
->ipv6_label
= tc_flow
& htonl(IPV6_LABEL_MASK
);
158 flow
->nw_ttl
= nh
->ip6_hlim
;
159 flow
->nw_proto
= IPPROTO_NONE
;
162 if ((nexthdr
!= IPPROTO_HOPOPTS
)
163 && (nexthdr
!= IPPROTO_ROUTING
)
164 && (nexthdr
!= IPPROTO_DSTOPTS
)
165 && (nexthdr
!= IPPROTO_AH
)
166 && (nexthdr
!= IPPROTO_FRAGMENT
)) {
167 /* It's either a terminal header (e.g., TCP, UDP) or one we
168 * don't understand. In either case, we're done with the
169 * packet, so use it to fill in 'nw_proto'. */
173 /* We only verify that at least 8 bytes of the next header are
174 * available, but many of these headers are longer. Ensure that
175 * accesses within the extension header are within those first 8
176 * bytes. All extension headers are required to be at least 8
178 if (packet
->size
< 8) {
182 if ((nexthdr
== IPPROTO_HOPOPTS
)
183 || (nexthdr
== IPPROTO_ROUTING
)
184 || (nexthdr
== IPPROTO_DSTOPTS
)) {
185 /* These headers, while different, have the fields we care about
186 * in the same location and with the same interpretation. */
187 const struct ip6_ext
*ext_hdr
= packet
->data
;
188 nexthdr
= ext_hdr
->ip6e_nxt
;
189 if (!ofpbuf_try_pull(packet
, (ext_hdr
->ip6e_len
+ 1) * 8)) {
192 } else if (nexthdr
== IPPROTO_AH
) {
193 /* A standard AH definition isn't available, but the fields
194 * we care about are in the same location as the generic
195 * option header--only the header length is calculated
197 const struct ip6_ext
*ext_hdr
= packet
->data
;
198 nexthdr
= ext_hdr
->ip6e_nxt
;
199 if (!ofpbuf_try_pull(packet
, (ext_hdr
->ip6e_len
+ 2) * 4)) {
202 } else if (nexthdr
== IPPROTO_FRAGMENT
) {
203 const struct ip6_frag
*frag_hdr
= packet
->data
;
205 nexthdr
= frag_hdr
->ip6f_nxt
;
206 if (!ofpbuf_try_pull(packet
, sizeof *frag_hdr
)) {
210 /* We only process the first fragment. */
211 if (frag_hdr
->ip6f_offlg
!= htons(0)) {
212 if ((frag_hdr
->ip6f_offlg
& IP6F_OFF_MASK
) == htons(0)) {
213 flow
->nw_frag
= FLOW_NW_FRAG_ANY
;
215 flow
->nw_frag
|= FLOW_NW_FRAG_LATER
;
216 nexthdr
= IPPROTO_FRAGMENT
;
223 flow
->nw_proto
= nexthdr
;
228 parse_tcp(struct ofpbuf
*packet
, struct ofpbuf
*b
, struct flow
*flow
)
230 const struct tcp_header
*tcp
= pull_tcp(b
);
232 flow
->tp_src
= tcp
->tcp_src
;
233 flow
->tp_dst
= tcp
->tcp_dst
;
234 packet
->l7
= b
->data
;
239 parse_udp(struct ofpbuf
*packet
, struct ofpbuf
*b
, struct flow
*flow
)
241 const struct udp_header
*udp
= pull_udp(b
);
243 flow
->tp_src
= udp
->udp_src
;
244 flow
->tp_dst
= udp
->udp_dst
;
245 packet
->l7
= b
->data
;
250 parse_icmpv6(struct ofpbuf
*b
, struct flow
*flow
)
252 const struct icmp6_hdr
*icmp
= pull_icmpv6(b
);
258 /* The ICMPv6 type and code fields use the 16-bit transport port
259 * fields, so we need to store them in 16-bit network byte order. */
260 flow
->tp_src
= htons(icmp
->icmp6_type
);
261 flow
->tp_dst
= htons(icmp
->icmp6_code
);
263 if (icmp
->icmp6_code
== 0 &&
264 (icmp
->icmp6_type
== ND_NEIGHBOR_SOLICIT
||
265 icmp
->icmp6_type
== ND_NEIGHBOR_ADVERT
)) {
266 const struct in6_addr
*nd_target
;
268 nd_target
= ofpbuf_try_pull(b
, sizeof *nd_target
);
272 flow
->nd_target
= *nd_target
;
274 while (b
->size
>= 8) {
275 /* The minimum size of an option is 8 bytes, which also is
276 * the size of Ethernet link-layer options. */
277 const struct nd_opt_hdr
*nd_opt
= b
->data
;
278 int opt_len
= nd_opt
->nd_opt_len
* 8;
280 if (!opt_len
|| opt_len
> b
->size
) {
284 /* Store the link layer address if the appropriate option is
285 * provided. It is considered an error if the same link
286 * layer option is specified twice. */
287 if (nd_opt
->nd_opt_type
== ND_OPT_SOURCE_LINKADDR
289 if (eth_addr_is_zero(flow
->arp_sha
)) {
290 memcpy(flow
->arp_sha
, nd_opt
+ 1, ETH_ADDR_LEN
);
294 } else if (nd_opt
->nd_opt_type
== ND_OPT_TARGET_LINKADDR
296 if (eth_addr_is_zero(flow
->arp_tha
)) {
297 memcpy(flow
->arp_tha
, nd_opt
+ 1, ETH_ADDR_LEN
);
303 if (!ofpbuf_try_pull(b
, opt_len
)) {
312 memset(&flow
->nd_target
, 0, sizeof(flow
->nd_target
));
313 memset(flow
->arp_sha
, 0, sizeof(flow
->arp_sha
));
314 memset(flow
->arp_tha
, 0, sizeof(flow
->arp_tha
));
320 /* Initializes 'flow' members from 'packet', 'skb_priority', 'tnl', and
323 * Initializes 'packet' header pointers as follows:
325 * - packet->l2 to the start of the Ethernet header.
327 * - packet->l3 to just past the Ethernet header, or just past the
328 * vlan_header if one is present, to the first byte of the payload of the
331 * - packet->l4 to just past the IPv4 header, if one is present and has a
332 * correct length, and otherwise NULL.
334 * - packet->l7 to just past the TCP or UDP or ICMP header, if one is
335 * present and has a correct length, and otherwise NULL.
338 flow_extract(struct ofpbuf
*packet
, uint32_t skb_priority
,
339 const struct flow_tnl
*tnl
, uint16_t ofp_in_port
,
342 struct ofpbuf b
= *packet
;
343 struct eth_header
*eth
;
345 COVERAGE_INC(flow_extract
);
347 memset(flow
, 0, sizeof *flow
);
350 assert(tnl
!= &flow
->tunnel
);
353 flow
->in_port
= ofp_in_port
;
354 flow
->skb_priority
= skb_priority
;
361 if (b
.size
< sizeof *eth
) {
367 memcpy(flow
->dl_src
, eth
->eth_src
, ETH_ADDR_LEN
);
368 memcpy(flow
->dl_dst
, eth
->eth_dst
, ETH_ADDR_LEN
);
370 /* dl_type, vlan_tci. */
371 ofpbuf_pull(&b
, ETH_ADDR_LEN
* 2);
372 if (eth
->eth_type
== htons(ETH_TYPE_VLAN
)) {
373 parse_vlan(&b
, flow
);
375 flow
->dl_type
= parse_ethertype(&b
);
379 if (flow
->dl_type
== htons(ETH_TYPE_IP
)) {
380 const struct ip_header
*nh
= pull_ip(&b
);
384 flow
->nw_src
= get_unaligned_be32(&nh
->ip_src
);
385 flow
->nw_dst
= get_unaligned_be32(&nh
->ip_dst
);
386 flow
->nw_proto
= nh
->ip_proto
;
388 flow
->nw_tos
= nh
->ip_tos
;
389 if (IP_IS_FRAGMENT(nh
->ip_frag_off
)) {
390 flow
->nw_frag
= FLOW_NW_FRAG_ANY
;
391 if (nh
->ip_frag_off
& htons(IP_FRAG_OFF_MASK
)) {
392 flow
->nw_frag
|= FLOW_NW_FRAG_LATER
;
395 flow
->nw_ttl
= nh
->ip_ttl
;
397 if (!(nh
->ip_frag_off
& htons(IP_FRAG_OFF_MASK
))) {
398 if (flow
->nw_proto
== IPPROTO_TCP
) {
399 parse_tcp(packet
, &b
, flow
);
400 } else if (flow
->nw_proto
== IPPROTO_UDP
) {
401 parse_udp(packet
, &b
, flow
);
402 } else if (flow
->nw_proto
== IPPROTO_ICMP
) {
403 const struct icmp_header
*icmp
= pull_icmp(&b
);
405 flow
->tp_src
= htons(icmp
->icmp_type
);
406 flow
->tp_dst
= htons(icmp
->icmp_code
);
412 } else if (flow
->dl_type
== htons(ETH_TYPE_IPV6
)) {
413 if (parse_ipv6(&b
, flow
)) {
418 if (flow
->nw_proto
== IPPROTO_TCP
) {
419 parse_tcp(packet
, &b
, flow
);
420 } else if (flow
->nw_proto
== IPPROTO_UDP
) {
421 parse_udp(packet
, &b
, flow
);
422 } else if (flow
->nw_proto
== IPPROTO_ICMPV6
) {
423 if (parse_icmpv6(&b
, flow
)) {
427 } else if (flow
->dl_type
== htons(ETH_TYPE_ARP
) ||
428 flow
->dl_type
== htons(ETH_TYPE_RARP
)) {
429 const struct arp_eth_header
*arp
= pull_arp(&b
);
430 if (arp
&& arp
->ar_hrd
== htons(1)
431 && arp
->ar_pro
== htons(ETH_TYPE_IP
)
432 && arp
->ar_hln
== ETH_ADDR_LEN
433 && arp
->ar_pln
== 4) {
434 /* We only match on the lower 8 bits of the opcode. */
435 if (ntohs(arp
->ar_op
) <= 0xff) {
436 flow
->nw_proto
= ntohs(arp
->ar_op
);
439 flow
->nw_src
= arp
->ar_spa
;
440 flow
->nw_dst
= arp
->ar_tpa
;
441 memcpy(flow
->arp_sha
, arp
->ar_sha
, ETH_ADDR_LEN
);
442 memcpy(flow
->arp_tha
, arp
->ar_tha
, ETH_ADDR_LEN
);
447 /* For every bit of a field that is wildcarded in 'wildcards', sets the
448 * corresponding bit in 'flow' to zero. */
450 flow_zero_wildcards(struct flow
*flow
, const struct flow_wildcards
*wildcards
)
452 uint32_t *flow_u32
= (uint32_t *) flow
;
453 const uint32_t *wc_u32
= (const uint32_t *) &wildcards
->masks
;
456 for (i
= 0; i
< FLOW_U32S
; i
++) {
457 flow_u32
[i
] &= wc_u32
[i
];
461 /* Initializes 'fmd' with the metadata found in 'flow'. */
463 flow_get_metadata(const struct flow
*flow
, struct flow_metadata
*fmd
)
465 BUILD_ASSERT_DECL(FLOW_WC_SEQ
== 17);
467 fmd
->tun_id
= flow
->tunnel
.tun_id
;
468 fmd
->metadata
= flow
->metadata
;
469 memcpy(fmd
->regs
, flow
->regs
, sizeof fmd
->regs
);
470 fmd
->in_port
= flow
->in_port
;
474 flow_to_string(const struct flow
*flow
)
476 struct ds ds
= DS_EMPTY_INITIALIZER
;
477 flow_format(&ds
, flow
);
482 flow_format(struct ds
*ds
, const struct flow
*flow
)
486 match_wc_init(&match
, flow
);
487 match_format(&match
, ds
, flow
->skb_priority
);
491 flow_print(FILE *stream
, const struct flow
*flow
)
493 char *s
= flow_to_string(flow
);
498 /* flow_wildcards functions. */
500 /* Initializes 'wc' as a set of wildcards that matches every packet. */
502 flow_wildcards_init_catchall(struct flow_wildcards
*wc
)
504 memset(&wc
->masks
, 0, sizeof wc
->masks
);
507 /* Initializes 'wc' as an exact-match set of wildcards; that is, 'wc' does not
508 * wildcard any bits or fields. */
510 flow_wildcards_init_exact(struct flow_wildcards
*wc
)
512 memset(&wc
->masks
, 0xff, sizeof wc
->masks
);
515 /* Returns true if 'wc' matches every packet, false if 'wc' fixes any bits or
518 flow_wildcards_is_catchall(const struct flow_wildcards
*wc
)
520 const uint32_t *wc_u32
= (const uint32_t *) &wc
->masks
;
523 for (i
= 0; i
< FLOW_U32S
; i
++) {
531 /* Initializes 'dst' as the combination of wildcards in 'src1' and 'src2'.
532 * That is, a bit or a field is wildcarded in 'dst' if it is wildcarded in
533 * 'src1' or 'src2' or both. */
535 flow_wildcards_combine(struct flow_wildcards
*dst
,
536 const struct flow_wildcards
*src1
,
537 const struct flow_wildcards
*src2
)
539 uint32_t *dst_u32
= (uint32_t *) &dst
->masks
;
540 const uint32_t *src1_u32
= (const uint32_t *) &src1
->masks
;
541 const uint32_t *src2_u32
= (const uint32_t *) &src2
->masks
;
544 for (i
= 0; i
< FLOW_U32S
; i
++) {
545 dst_u32
[i
] = src1_u32
[i
] & src2_u32
[i
];
549 /* Returns a hash of the wildcards in 'wc'. */
551 flow_wildcards_hash(const struct flow_wildcards
*wc
, uint32_t basis
)
553 return flow_hash(&wc
->masks
, basis
);;
556 /* Returns true if 'a' and 'b' represent the same wildcards, false if they are
559 flow_wildcards_equal(const struct flow_wildcards
*a
,
560 const struct flow_wildcards
*b
)
562 return flow_equal(&a
->masks
, &b
->masks
);
565 /* Returns true if at least one bit or field is wildcarded in 'a' but not in
566 * 'b', false otherwise. */
568 flow_wildcards_has_extra(const struct flow_wildcards
*a
,
569 const struct flow_wildcards
*b
)
571 const uint32_t *a_u32
= (const uint32_t *) &a
->masks
;
572 const uint32_t *b_u32
= (const uint32_t *) &b
->masks
;
575 for (i
= 0; i
< FLOW_U32S
; i
++) {
576 if ((a_u32
[i
] & b_u32
[i
]) != b_u32
[i
]) {
583 /* Returns true if 'a' and 'b' are equal, except that 0-bits (wildcarded bits)
584 * in 'wc' do not need to be equal in 'a' and 'b'. */
586 flow_equal_except(const struct flow
*a
, const struct flow
*b
,
587 const struct flow_wildcards
*wc
)
589 const uint32_t *a_u32
= (const uint32_t *) a
;
590 const uint32_t *b_u32
= (const uint32_t *) b
;
591 const uint32_t *wc_u32
= (const uint32_t *) &wc
->masks
;
594 for (i
= 0; i
< FLOW_U32S
; i
++) {
595 if ((a_u32
[i
] ^ b_u32
[i
]) & wc_u32
[i
]) {
602 /* Sets the wildcard mask for register 'idx' in 'wc' to 'mask'.
603 * (A 0-bit indicates a wildcard bit.) */
605 flow_wildcards_set_reg_mask(struct flow_wildcards
*wc
, int idx
, uint32_t mask
)
607 wc
->masks
.regs
[idx
] = mask
;
610 /* Hashes 'flow' based on its L2 through L4 protocol information. */
612 flow_hash_symmetric_l4(const struct flow
*flow
, uint32_t basis
)
617 struct in6_addr ipv6_addr
;
622 uint8_t eth_addr
[ETH_ADDR_LEN
];
628 memset(&fields
, 0, sizeof fields
);
629 for (i
= 0; i
< ETH_ADDR_LEN
; i
++) {
630 fields
.eth_addr
[i
] = flow
->dl_src
[i
] ^ flow
->dl_dst
[i
];
632 fields
.vlan_tci
= flow
->vlan_tci
& htons(VLAN_VID_MASK
);
633 fields
.eth_type
= flow
->dl_type
;
635 /* UDP source and destination port are not taken into account because they
636 * will not necessarily be symmetric in a bidirectional flow. */
637 if (fields
.eth_type
== htons(ETH_TYPE_IP
)) {
638 fields
.ipv4_addr
= flow
->nw_src
^ flow
->nw_dst
;
639 fields
.ip_proto
= flow
->nw_proto
;
640 if (fields
.ip_proto
== IPPROTO_TCP
) {
641 fields
.tp_port
= flow
->tp_src
^ flow
->tp_dst
;
643 } else if (fields
.eth_type
== htons(ETH_TYPE_IPV6
)) {
644 const uint8_t *a
= &flow
->ipv6_src
.s6_addr
[0];
645 const uint8_t *b
= &flow
->ipv6_dst
.s6_addr
[0];
646 uint8_t *ipv6_addr
= &fields
.ipv6_addr
.s6_addr
[0];
648 for (i
=0; i
<16; i
++) {
649 ipv6_addr
[i
] = a
[i
] ^ b
[i
];
651 fields
.ip_proto
= flow
->nw_proto
;
652 if (fields
.ip_proto
== IPPROTO_TCP
) {
653 fields
.tp_port
= flow
->tp_src
^ flow
->tp_dst
;
656 return hash_bytes(&fields
, sizeof fields
, basis
);
659 /* Hashes the portions of 'flow' designated by 'fields'. */
661 flow_hash_fields(const struct flow
*flow
, enum nx_hash_fields fields
,
666 case NX_HASH_FIELDS_ETH_SRC
:
667 return hash_bytes(flow
->dl_src
, sizeof flow
->dl_src
, basis
);
669 case NX_HASH_FIELDS_SYMMETRIC_L4
:
670 return flow_hash_symmetric_l4(flow
, basis
);
676 /* Returns a string representation of 'fields'. */
678 flow_hash_fields_to_str(enum nx_hash_fields fields
)
681 case NX_HASH_FIELDS_ETH_SRC
: return "eth_src";
682 case NX_HASH_FIELDS_SYMMETRIC_L4
: return "symmetric_l4";
683 default: return "<unknown>";
687 /* Returns true if the value of 'fields' is supported. Otherwise false. */
689 flow_hash_fields_valid(enum nx_hash_fields fields
)
691 return fields
== NX_HASH_FIELDS_ETH_SRC
692 || fields
== NX_HASH_FIELDS_SYMMETRIC_L4
;
695 /* Sets the VLAN VID that 'flow' matches to 'vid', which is interpreted as an
696 * OpenFlow 1.0 "dl_vlan" value:
698 * - If it is in the range 0...4095, 'flow->vlan_tci' is set to match
699 * that VLAN. Any existing PCP match is unchanged (it becomes 0 if
700 * 'flow' previously matched packets without a VLAN header).
702 * - If it is OFP_VLAN_NONE, 'flow->vlan_tci' is set to match a packet
703 * without a VLAN tag.
705 * - Other values of 'vid' should not be used. */
707 flow_set_dl_vlan(struct flow
*flow
, ovs_be16 vid
)
709 if (vid
== htons(OFP10_VLAN_NONE
)) {
710 flow
->vlan_tci
= htons(0);
712 vid
&= htons(VLAN_VID_MASK
);
713 flow
->vlan_tci
&= ~htons(VLAN_VID_MASK
);
714 flow
->vlan_tci
|= htons(VLAN_CFI
) | vid
;
718 /* Sets the VLAN VID that 'flow' matches to 'vid', which is interpreted as an
719 * OpenFlow 1.2 "vlan_vid" value, that is, the low 13 bits of 'vlan_tci' (VID
722 flow_set_vlan_vid(struct flow
*flow
, ovs_be16 vid
)
724 ovs_be16 mask
= htons(VLAN_VID_MASK
| VLAN_CFI
);
725 flow
->vlan_tci
&= ~mask
;
726 flow
->vlan_tci
|= vid
& mask
;
729 /* Sets the VLAN PCP that 'flow' matches to 'pcp', which should be in the
732 * This function has no effect on the VLAN ID that 'flow' matches.
734 * After calling this function, 'flow' will not match packets without a VLAN
737 flow_set_vlan_pcp(struct flow
*flow
, uint8_t pcp
)
740 flow
->vlan_tci
&= ~htons(VLAN_PCP_MASK
);
741 flow
->vlan_tci
|= htons((pcp
<< VLAN_PCP_SHIFT
) | VLAN_CFI
);
744 /* Puts into 'b' a packet that flow_extract() would parse as having the given
747 * (This is useful only for testing, obviously, and the packet isn't really
748 * valid. It hasn't got some checksums filled in, for one, and lots of fields
749 * are just zeroed.) */
751 flow_compose(struct ofpbuf
*b
, const struct flow
*flow
)
753 eth_compose(b
, flow
->dl_dst
, flow
->dl_src
, ntohs(flow
->dl_type
), 0);
754 if (flow
->dl_type
== htons(FLOW_DL_TYPE_NONE
)) {
755 struct eth_header
*eth
= b
->l2
;
756 eth
->eth_type
= htons(b
->size
);
760 if (flow
->vlan_tci
& htons(VLAN_CFI
)) {
761 eth_push_vlan(b
, flow
->vlan_tci
);
764 if (flow
->dl_type
== htons(ETH_TYPE_IP
)) {
765 struct ip_header
*ip
;
767 b
->l3
= ip
= ofpbuf_put_zeros(b
, sizeof *ip
);
768 ip
->ip_ihl_ver
= IP_IHL_VER(5, 4);
769 ip
->ip_tos
= flow
->nw_tos
;
770 ip
->ip_ttl
= flow
->nw_ttl
;
771 ip
->ip_proto
= flow
->nw_proto
;
772 ip
->ip_src
= flow
->nw_src
;
773 ip
->ip_dst
= flow
->nw_dst
;
775 if (flow
->nw_frag
& FLOW_NW_FRAG_ANY
) {
776 ip
->ip_frag_off
|= htons(IP_MORE_FRAGMENTS
);
777 if (flow
->nw_frag
& FLOW_NW_FRAG_LATER
) {
778 ip
->ip_frag_off
|= htons(100);
781 if (!(flow
->nw_frag
& FLOW_NW_FRAG_ANY
)
782 || !(flow
->nw_frag
& FLOW_NW_FRAG_LATER
)) {
783 if (flow
->nw_proto
== IPPROTO_TCP
) {
784 struct tcp_header
*tcp
;
786 b
->l4
= tcp
= ofpbuf_put_zeros(b
, sizeof *tcp
);
787 tcp
->tcp_src
= flow
->tp_src
;
788 tcp
->tcp_dst
= flow
->tp_dst
;
789 tcp
->tcp_ctl
= TCP_CTL(0, 5);
790 } else if (flow
->nw_proto
== IPPROTO_UDP
) {
791 struct udp_header
*udp
;
793 b
->l4
= udp
= ofpbuf_put_zeros(b
, sizeof *udp
);
794 udp
->udp_src
= flow
->tp_src
;
795 udp
->udp_dst
= flow
->tp_dst
;
796 } else if (flow
->nw_proto
== IPPROTO_ICMP
) {
797 struct icmp_header
*icmp
;
799 b
->l4
= icmp
= ofpbuf_put_zeros(b
, sizeof *icmp
);
800 icmp
->icmp_type
= ntohs(flow
->tp_src
);
801 icmp
->icmp_code
= ntohs(flow
->tp_dst
);
802 icmp
->icmp_csum
= csum(icmp
, ICMP_HEADER_LEN
);
807 ip
->ip_tot_len
= htons((uint8_t *) b
->data
+ b
->size
808 - (uint8_t *) b
->l3
);
809 ip
->ip_csum
= csum(ip
, sizeof *ip
);
810 } else if (flow
->dl_type
== htons(ETH_TYPE_IPV6
)) {
812 } else if (flow
->dl_type
== htons(ETH_TYPE_ARP
) ||
813 flow
->dl_type
== htons(ETH_TYPE_RARP
)) {
814 struct arp_eth_header
*arp
;
816 b
->l3
= arp
= ofpbuf_put_zeros(b
, sizeof *arp
);
817 arp
->ar_hrd
= htons(1);
818 arp
->ar_pro
= htons(ETH_TYPE_IP
);
819 arp
->ar_hln
= ETH_ADDR_LEN
;
821 arp
->ar_op
= htons(flow
->nw_proto
);
823 if (flow
->nw_proto
== ARP_OP_REQUEST
||
824 flow
->nw_proto
== ARP_OP_REPLY
) {
825 arp
->ar_spa
= flow
->nw_src
;
826 arp
->ar_tpa
= flow
->nw_dst
;
827 memcpy(arp
->ar_sha
, flow
->arp_sha
, ETH_ADDR_LEN
);
828 memcpy(arp
->ar_tha
, flow
->arp_tha
, ETH_ADDR_LEN
);
833 /* Compressed flow. */
836 miniflow_n_values(const struct miniflow
*flow
)
841 for (i
= 0; i
< MINI_N_MAPS
; i
++) {
842 n
+= popcount(flow
->map
[i
]);
848 miniflow_alloc_values(struct miniflow
*flow
, int n
)
850 if (n
<= MINI_N_INLINE
) {
851 return flow
->inline_values
;
853 COVERAGE_INC(miniflow_malloc
);
854 return xmalloc(n
* sizeof *flow
->values
);
858 /* Initializes 'dst' as a copy of 'src'. The caller must eventually free 'dst'
859 * with miniflow_destroy(). */
861 miniflow_init(struct miniflow
*dst
, const struct flow
*src
)
863 const uint32_t *src_u32
= (const uint32_t *) src
;
868 /* Initialize dst->map, counting the number of nonzero elements. */
870 memset(dst
->map
, 0, sizeof dst
->map
);
871 for (i
= 0; i
< FLOW_U32S
; i
++) {
873 dst
->map
[i
/ 32] |= 1u << (i
% 32);
878 /* Initialize dst->values. */
879 dst
->values
= miniflow_alloc_values(dst
, n
);
881 for (i
= 0; i
< MINI_N_MAPS
; i
++) {
884 for (map
= dst
->map
[i
]; map
; map
= zero_rightmost_1bit(map
)) {
885 dst
->values
[ofs
++] = src_u32
[raw_ctz(map
) + i
* 32];
890 /* Initializes 'dst' as a copy of 'src'. The caller must eventually free 'dst'
891 * with miniflow_destroy(). */
893 miniflow_clone(struct miniflow
*dst
, const struct miniflow
*src
)
895 int n
= miniflow_n_values(src
);
896 memcpy(dst
->map
, src
->map
, sizeof dst
->map
);
897 dst
->values
= miniflow_alloc_values(dst
, n
);
898 memcpy(dst
->values
, src
->values
, n
* sizeof *dst
->values
);
901 /* Frees any memory owned by 'flow'. Does not free the storage in which 'flow'
902 * itself resides; the caller is responsible for that. */
904 miniflow_destroy(struct miniflow
*flow
)
906 if (flow
->values
!= flow
->inline_values
) {
911 /* Initializes 'dst' as a copy of 'src'. */
913 miniflow_expand(const struct miniflow
*src
, struct flow
*dst
)
915 uint32_t *dst_u32
= (uint32_t *) dst
;
919 memset(dst_u32
, 0, sizeof *dst
);
922 for (i
= 0; i
< MINI_N_MAPS
; i
++) {
925 for (map
= src
->map
[i
]; map
; map
= zero_rightmost_1bit(map
)) {
926 dst_u32
[raw_ctz(map
) + i
* 32] = src
->values
[ofs
++];
931 static const uint32_t *
932 miniflow_get__(const struct miniflow
*flow
, unsigned int u32_ofs
)
934 if (!(flow
->map
[u32_ofs
/ 32] & (1u << (u32_ofs
% 32)))) {
935 static const uint32_t zero
= 0;
938 const uint32_t *p
= flow
->values
;
940 BUILD_ASSERT(MINI_N_MAPS
== 2);
942 p
+= popcount(flow
->map
[0] & ((1u << u32_ofs
) - 1));
944 p
+= popcount(flow
->map
[0]);
945 p
+= popcount(flow
->map
[1] & ((1u << (u32_ofs
- 32)) - 1));
951 /* Returns the uint32_t that would be at byte offset '4 * u32_ofs' if 'flow'
952 * were expanded into a "struct flow". */
954 miniflow_get(const struct miniflow
*flow
, unsigned int u32_ofs
)
956 return *miniflow_get__(flow
, u32_ofs
);
959 /* Returns the ovs_be16 that would be at byte offset 'u8_ofs' if 'flow' were
960 * expanded into a "struct flow". */
962 miniflow_get_be16(const struct miniflow
*flow
, unsigned int u8_ofs
)
964 const uint32_t *u32p
= miniflow_get__(flow
, u8_ofs
/ 4);
965 const ovs_be16
*be16p
= (const ovs_be16
*) u32p
;
966 return be16p
[u8_ofs
% 4 != 0];
969 /* Returns the VID within the vlan_tci member of the "struct flow" represented
972 miniflow_get_vid(const struct miniflow
*flow
)
974 ovs_be16 tci
= miniflow_get_be16(flow
, offsetof(struct flow
, vlan_tci
));
975 return vlan_tci_to_vid(tci
);
978 /* Returns true if 'a' and 'b' are the same flow, false otherwise. */
980 miniflow_equal(const struct miniflow
*a
, const struct miniflow
*b
)
984 for (i
= 0; i
< MINI_N_MAPS
; i
++) {
985 if (a
->map
[i
] != b
->map
[i
]) {
990 return !memcmp(a
->values
, b
->values
,
991 miniflow_n_values(a
) * sizeof *a
->values
);
994 /* Returns true if 'a' and 'b' are equal at the places where there are 1-bits
995 * in 'mask', false if they differ. */
997 miniflow_equal_in_minimask(const struct miniflow
*a
, const struct miniflow
*b
,
998 const struct minimask
*mask
)
1003 p
= mask
->masks
.values
;
1004 for (i
= 0; i
< MINI_N_MAPS
; i
++) {
1007 for (map
= mask
->masks
.map
[i
]; map
; map
= zero_rightmost_1bit(map
)) {
1008 int ofs
= raw_ctz(map
) + i
* 32;
1010 if ((miniflow_get(a
, ofs
) ^ miniflow_get(b
, ofs
)) & *p
) {
1020 /* Returns true if 'a' and 'b' are equal at the places where there are 1-bits
1021 * in 'mask', false if they differ. */
1023 miniflow_equal_flow_in_minimask(const struct miniflow
*a
, const struct flow
*b
,
1024 const struct minimask
*mask
)
1026 const uint32_t *b_u32
= (const uint32_t *) b
;
1030 p
= mask
->masks
.values
;
1031 for (i
= 0; i
< MINI_N_MAPS
; i
++) {
1034 for (map
= mask
->masks
.map
[i
]; map
; map
= zero_rightmost_1bit(map
)) {
1035 int ofs
= raw_ctz(map
) + i
* 32;
1037 if ((miniflow_get(a
, ofs
) ^ b_u32
[ofs
]) & *p
) {
1047 /* Returns a hash value for 'flow', given 'basis'. */
1049 miniflow_hash(const struct miniflow
*flow
, uint32_t basis
)
1051 BUILD_ASSERT_DECL(MINI_N_MAPS
== 2);
1052 return hash_3words(flow
->map
[0], flow
->map
[1],
1053 hash_words(flow
->values
, miniflow_n_values(flow
),
1057 /* Returns a hash value for the bits of 'flow' where there are 1-bits in
1058 * 'mask', given 'basis'.
1060 * The hash values returned by this function are the same as those returned by
1061 * flow_hash_in_minimask(), only the form of the arguments differ. */
1063 miniflow_hash_in_minimask(const struct miniflow
*flow
,
1064 const struct minimask
*mask
, uint32_t basis
)
1066 const uint32_t *p
= mask
->masks
.values
;
1071 for (i
= 0; i
< MINI_N_MAPS
; i
++) {
1074 for (map
= mask
->masks
.map
[i
]; map
; map
= zero_rightmost_1bit(map
)) {
1075 int ofs
= raw_ctz(map
) + i
* 32;
1077 hash
= mhash_add(hash
, miniflow_get(flow
, ofs
) & *p
);
1082 return mhash_finish(hash
, p
- mask
->masks
.values
);
1085 /* Returns a hash value for the bits of 'flow' where there are 1-bits in
1086 * 'mask', given 'basis'.
1088 * The hash values returned by this function are the same as those returned by
1089 * miniflow_hash_in_minimask(), only the form of the arguments differ. */
1091 flow_hash_in_minimask(const struct flow
*flow
, const struct minimask
*mask
,
1094 const uint32_t *flow_u32
= (const uint32_t *) flow
;
1095 const uint32_t *p
= mask
->masks
.values
;
1100 for (i
= 0; i
< MINI_N_MAPS
; i
++) {
1103 for (map
= mask
->masks
.map
[i
]; map
; map
= zero_rightmost_1bit(map
)) {
1104 int ofs
= raw_ctz(map
) + i
* 32;
1106 hash
= mhash_add(hash
, flow_u32
[ofs
] & *p
);
1111 return mhash_finish(hash
, p
- mask
->masks
.values
);
1114 /* Initializes 'dst' as a copy of 'src'. The caller must eventually free 'dst'
1115 * with minimask_destroy(). */
1117 minimask_init(struct minimask
*mask
, const struct flow_wildcards
*wc
)
1119 miniflow_init(&mask
->masks
, &wc
->masks
);
1122 /* Initializes 'dst' as a copy of 'src'. The caller must eventually free 'dst'
1123 * with minimask_destroy(). */
1125 minimask_clone(struct minimask
*dst
, const struct minimask
*src
)
1127 miniflow_clone(&dst
->masks
, &src
->masks
);
1130 /* Initializes 'dst_' as the bit-wise "and" of 'a_' and 'b_'.
1132 * The caller must provide room for FLOW_U32S "uint32_t"s in 'storage', for use
1133 * by 'dst_'. The caller must *not* free 'dst_' with minimask_destroy(). */
1135 minimask_combine(struct minimask
*dst_
,
1136 const struct minimask
*a_
, const struct minimask
*b_
,
1137 uint32_t storage
[FLOW_U32S
])
1139 struct miniflow
*dst
= &dst_
->masks
;
1140 const struct miniflow
*a
= &a_
->masks
;
1141 const struct miniflow
*b
= &b_
->masks
;
1145 dst
->values
= storage
;
1146 for (i
= 0; i
< MINI_N_MAPS
; i
++) {
1150 for (map
= a
->map
[i
] & b
->map
[i
]; map
;
1151 map
= zero_rightmost_1bit(map
)) {
1152 int ofs
= raw_ctz(map
) + i
* 32;
1153 uint32_t mask
= miniflow_get(a
, ofs
) & miniflow_get(b
, ofs
);
1156 dst
->map
[i
] |= rightmost_1bit(map
);
1157 dst
->values
[n
++] = mask
;
1163 /* Frees any memory owned by 'mask'. Does not free the storage in which 'mask'
1164 * itself resides; the caller is responsible for that. */
1166 minimask_destroy(struct minimask
*mask
)
1168 miniflow_destroy(&mask
->masks
);
1171 /* Initializes 'dst' as a copy of 'src'. */
1173 minimask_expand(const struct minimask
*mask
, struct flow_wildcards
*wc
)
1175 miniflow_expand(&mask
->masks
, &wc
->masks
);
1178 /* Returns the uint32_t that would be at byte offset '4 * u32_ofs' if 'mask'
1179 * were expanded into a "struct flow_wildcards". */
1181 minimask_get(const struct minimask
*mask
, unsigned int u32_ofs
)
1183 return miniflow_get(&mask
->masks
, u32_ofs
);
1186 /* Returns the VID mask within the vlan_tci member of the "struct
1187 * flow_wildcards" represented by 'mask'. */
1189 minimask_get_vid_mask(const struct minimask
*mask
)
1191 return miniflow_get_vid(&mask
->masks
);
1194 /* Returns true if 'a' and 'b' are the same flow mask, false otherwise. */
1196 minimask_equal(const struct minimask
*a
, const struct minimask
*b
)
1198 return miniflow_equal(&a
->masks
, &b
->masks
);
1201 /* Returns a hash value for 'mask', given 'basis'. */
1203 minimask_hash(const struct minimask
*mask
, uint32_t basis
)
1205 return miniflow_hash(&mask
->masks
, basis
);
1208 /* Returns true if at least one bit is wildcarded in 'a_' but not in 'b_',
1209 * false otherwise. */
1211 minimask_has_extra(const struct minimask
*a_
, const struct minimask
*b_
)
1213 const struct miniflow
*a
= &a_
->masks
;
1214 const struct miniflow
*b
= &b_
->masks
;
1217 for (i
= 0; i
< MINI_N_MAPS
; i
++) {
1220 for (map
= a
->map
[i
] | b
->map
[i
]; map
;
1221 map
= zero_rightmost_1bit(map
)) {
1222 int ofs
= raw_ctz(map
) + i
* 32;
1223 uint32_t a_u32
= miniflow_get(a
, ofs
);
1224 uint32_t b_u32
= miniflow_get(b
, ofs
);
1226 if ((a_u32
& b_u32
) != b_u32
) {
1235 /* Returns true if 'mask' matches every packet, false if 'mask' fixes any bits
1238 minimask_is_catchall(const struct minimask
*mask_
)
1240 const struct miniflow
*mask
= &mask_
->masks
;
1242 BUILD_ASSERT(MINI_N_MAPS
== 2);
1243 return !(mask
->map
[0] | mask
->map
[1]);