2 * Copyright (c) 2008, 2009, 2010, 2011, 2012, 2013, 2014 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>
28 #include "byte-order.h"
31 #include "dynamic-string.h"
36 #include "openflow/openflow.h"
40 #include "unaligned.h"
42 COVERAGE_DEFINE(flow_extract
);
43 COVERAGE_DEFINE(miniflow_malloc
);
45 /* U32 indices for segmented flow classification. */
46 const uint8_t flow_segment_u32s
[4] = {
47 FLOW_SEGMENT_1_ENDS_AT
/ 4,
48 FLOW_SEGMENT_2_ENDS_AT
/ 4,
49 FLOW_SEGMENT_3_ENDS_AT
/ 4,
53 static struct arp_eth_header
*
54 pull_arp(struct ofpbuf
*packet
)
56 return ofpbuf_try_pull(packet
, ARP_ETH_HEADER_LEN
);
59 static struct ip_header
*
60 pull_ip(struct ofpbuf
*packet
)
62 if (ofpbuf_size(packet
) >= IP_HEADER_LEN
) {
63 struct ip_header
*ip
= ofpbuf_data(packet
);
64 int ip_len
= IP_IHL(ip
->ip_ihl_ver
) * 4;
65 if (ip_len
>= IP_HEADER_LEN
&& ofpbuf_size(packet
) >= ip_len
) {
66 return ofpbuf_pull(packet
, ip_len
);
72 static struct icmp_header
*
73 pull_icmp(struct ofpbuf
*packet
)
75 return ofpbuf_try_pull(packet
, ICMP_HEADER_LEN
);
78 static struct icmp6_hdr
*
79 pull_icmpv6(struct ofpbuf
*packet
)
81 return ofpbuf_try_pull(packet
, sizeof(struct icmp6_hdr
));
85 parse_mpls(struct ofpbuf
*b
, struct flow
*flow
)
90 while ((mh
= ofpbuf_try_pull(b
, sizeof *mh
))) {
91 ovs_be32 mpls_lse
= get_16aligned_be32(&mh
->mpls_lse
);
92 if (idx
< FLOW_MAX_MPLS_LABELS
) {
93 flow
->mpls_lse
[idx
++] = mpls_lse
;
95 if (mpls_lse
& htonl(MPLS_BOS_MASK
)) {
102 parse_vlan(struct ofpbuf
*b
, struct flow
*flow
)
105 ovs_be16 eth_type
; /* ETH_TYPE_VLAN */
109 if (ofpbuf_size(b
) >= sizeof(struct qtag_prefix
) + sizeof(ovs_be16
)) {
110 struct qtag_prefix
*qp
= ofpbuf_pull(b
, sizeof *qp
);
111 flow
->vlan_tci
= qp
->tci
| htons(VLAN_CFI
);
116 parse_ethertype(struct ofpbuf
*b
)
118 struct llc_snap_header
*llc
;
121 proto
= *(ovs_be16
*) ofpbuf_pull(b
, sizeof proto
);
122 if (ntohs(proto
) >= ETH_TYPE_MIN
) {
126 if (ofpbuf_size(b
) < sizeof *llc
) {
127 return htons(FLOW_DL_TYPE_NONE
);
130 llc
= ofpbuf_data(b
);
131 if (llc
->llc
.llc_dsap
!= LLC_DSAP_SNAP
132 || llc
->llc
.llc_ssap
!= LLC_SSAP_SNAP
133 || llc
->llc
.llc_cntl
!= LLC_CNTL_SNAP
134 || memcmp(llc
->snap
.snap_org
, SNAP_ORG_ETHERNET
,
135 sizeof llc
->snap
.snap_org
)) {
136 return htons(FLOW_DL_TYPE_NONE
);
139 ofpbuf_pull(b
, sizeof *llc
);
141 if (ntohs(llc
->snap
.snap_type
) >= ETH_TYPE_MIN
) {
142 return llc
->snap
.snap_type
;
145 return htons(FLOW_DL_TYPE_NONE
);
149 parse_ipv6(struct ofpbuf
*packet
, struct flow
*flow
)
151 const struct ovs_16aligned_ip6_hdr
*nh
;
155 nh
= ofpbuf_try_pull(packet
, sizeof *nh
);
160 nexthdr
= nh
->ip6_nxt
;
162 memcpy(&flow
->ipv6_src
, &nh
->ip6_src
, sizeof flow
->ipv6_src
);
163 memcpy(&flow
->ipv6_dst
, &nh
->ip6_dst
, sizeof flow
->ipv6_dst
);
165 tc_flow
= get_16aligned_be32(&nh
->ip6_flow
);
166 flow
->nw_tos
= ntohl(tc_flow
) >> 20;
167 flow
->ipv6_label
= tc_flow
& htonl(IPV6_LABEL_MASK
);
168 flow
->nw_ttl
= nh
->ip6_hlim
;
169 flow
->nw_proto
= IPPROTO_NONE
;
172 if ((nexthdr
!= IPPROTO_HOPOPTS
)
173 && (nexthdr
!= IPPROTO_ROUTING
)
174 && (nexthdr
!= IPPROTO_DSTOPTS
)
175 && (nexthdr
!= IPPROTO_AH
)
176 && (nexthdr
!= IPPROTO_FRAGMENT
)) {
177 /* It's either a terminal header (e.g., TCP, UDP) or one we
178 * don't understand. In either case, we're done with the
179 * packet, so use it to fill in 'nw_proto'. */
183 /* We only verify that at least 8 bytes of the next header are
184 * available, but many of these headers are longer. Ensure that
185 * accesses within the extension header are within those first 8
186 * bytes. All extension headers are required to be at least 8
188 if (ofpbuf_size(packet
) < 8) {
192 if ((nexthdr
== IPPROTO_HOPOPTS
)
193 || (nexthdr
== IPPROTO_ROUTING
)
194 || (nexthdr
== IPPROTO_DSTOPTS
)) {
195 /* These headers, while different, have the fields we care about
196 * in the same location and with the same interpretation. */
197 const struct ip6_ext
*ext_hdr
= ofpbuf_data(packet
);
198 nexthdr
= ext_hdr
->ip6e_nxt
;
199 if (!ofpbuf_try_pull(packet
, (ext_hdr
->ip6e_len
+ 1) * 8)) {
202 } else if (nexthdr
== IPPROTO_AH
) {
203 /* A standard AH definition isn't available, but the fields
204 * we care about are in the same location as the generic
205 * option header--only the header length is calculated
207 const struct ip6_ext
*ext_hdr
= ofpbuf_data(packet
);
208 nexthdr
= ext_hdr
->ip6e_nxt
;
209 if (!ofpbuf_try_pull(packet
, (ext_hdr
->ip6e_len
+ 2) * 4)) {
212 } else if (nexthdr
== IPPROTO_FRAGMENT
) {
213 const struct ovs_16aligned_ip6_frag
*frag_hdr
= ofpbuf_data(packet
);
215 nexthdr
= frag_hdr
->ip6f_nxt
;
216 if (!ofpbuf_try_pull(packet
, sizeof *frag_hdr
)) {
220 /* We only process the first fragment. */
221 if (frag_hdr
->ip6f_offlg
!= htons(0)) {
222 flow
->nw_frag
= FLOW_NW_FRAG_ANY
;
223 if ((frag_hdr
->ip6f_offlg
& IP6F_OFF_MASK
) != htons(0)) {
224 flow
->nw_frag
|= FLOW_NW_FRAG_LATER
;
225 nexthdr
= IPPROTO_FRAGMENT
;
232 flow
->nw_proto
= nexthdr
;
237 parse_tcp(struct ofpbuf
*b
, struct flow
*flow
)
239 if (ofpbuf_size(b
) >= TCP_HEADER_LEN
) {
240 const struct tcp_header
*tcp
= ofpbuf_data(b
);
242 flow
->tp_src
= tcp
->tcp_src
;
243 flow
->tp_dst
= tcp
->tcp_dst
;
244 flow
->tcp_flags
= tcp
->tcp_ctl
& htons(0x0fff);
249 parse_udp(struct ofpbuf
*b
, struct flow
*flow
)
251 if (ofpbuf_size(b
) >= UDP_HEADER_LEN
) {
252 const struct udp_header
*udp
= ofpbuf_data(b
);
254 flow
->tp_src
= udp
->udp_src
;
255 flow
->tp_dst
= udp
->udp_dst
;
260 parse_sctp(struct ofpbuf
*b
, struct flow
*flow
)
262 if (ofpbuf_size(b
) >= SCTP_HEADER_LEN
) {
263 const struct sctp_header
*sctp
= ofpbuf_data(b
);
265 flow
->tp_src
= sctp
->sctp_src
;
266 flow
->tp_dst
= sctp
->sctp_dst
;
271 parse_icmpv6(struct ofpbuf
*b
, struct flow
*flow
)
273 const struct icmp6_hdr
*icmp
= pull_icmpv6(b
);
279 /* The ICMPv6 type and code fields use the 16-bit transport port
280 * fields, so we need to store them in 16-bit network byte order. */
281 flow
->tp_src
= htons(icmp
->icmp6_type
);
282 flow
->tp_dst
= htons(icmp
->icmp6_code
);
284 if (icmp
->icmp6_code
== 0 &&
285 (icmp
->icmp6_type
== ND_NEIGHBOR_SOLICIT
||
286 icmp
->icmp6_type
== ND_NEIGHBOR_ADVERT
)) {
287 const struct in6_addr
*nd_target
;
289 nd_target
= ofpbuf_try_pull(b
, sizeof *nd_target
);
293 flow
->nd_target
= *nd_target
;
295 while (ofpbuf_size(b
) >= 8) {
296 /* The minimum size of an option is 8 bytes, which also is
297 * the size of Ethernet link-layer options. */
298 const struct nd_opt_hdr
*nd_opt
= ofpbuf_data(b
);
299 int opt_len
= nd_opt
->nd_opt_len
* 8;
301 if (!opt_len
|| opt_len
> ofpbuf_size(b
)) {
305 /* Store the link layer address if the appropriate option is
306 * provided. It is considered an error if the same link
307 * layer option is specified twice. */
308 if (nd_opt
->nd_opt_type
== ND_OPT_SOURCE_LINKADDR
310 if (eth_addr_is_zero(flow
->arp_sha
)) {
311 memcpy(flow
->arp_sha
, nd_opt
+ 1, ETH_ADDR_LEN
);
315 } else if (nd_opt
->nd_opt_type
== ND_OPT_TARGET_LINKADDR
317 if (eth_addr_is_zero(flow
->arp_tha
)) {
318 memcpy(flow
->arp_tha
, nd_opt
+ 1, ETH_ADDR_LEN
);
324 if (!ofpbuf_try_pull(b
, opt_len
)) {
333 memset(&flow
->nd_target
, 0, sizeof(flow
->nd_target
));
334 memset(flow
->arp_sha
, 0, sizeof(flow
->arp_sha
));
335 memset(flow
->arp_tha
, 0, sizeof(flow
->arp_tha
));
340 /* Initializes 'flow' members from 'packet' and 'md'
342 * Initializes 'packet' header l2 pointer to the start of the Ethernet
343 * header, and the layer offsets as follows:
345 * - packet->l2_5_ofs to the start of the MPLS shim header, or UINT16_MAX
346 * when there is no MPLS shim header.
348 * - packet->l3_ofs to just past the Ethernet header, or just past the
349 * vlan_header if one is present, to the first byte of the payload of the
350 * Ethernet frame. UINT16_MAX if the frame is too short to contain an
353 * - packet->l4_ofs to just past the IPv4 header, if one is present and
354 * has at least the content used for the fields of interest for the flow,
355 * otherwise UINT16_MAX.
358 flow_extract(struct ofpbuf
*packet
, const struct pkt_metadata
*md
,
361 struct ofpbuf b
= *packet
;
362 struct eth_header
*eth
;
364 COVERAGE_INC(flow_extract
);
366 memset(flow
, 0, sizeof *flow
);
369 flow
->tunnel
= md
->tunnel
;
370 flow
->in_port
= md
->in_port
;
371 flow
->skb_priority
= md
->skb_priority
;
372 flow
->pkt_mark
= md
->pkt_mark
;
375 ofpbuf_set_frame(packet
, ofpbuf_data(packet
));
377 if (ofpbuf_size(&b
) < sizeof *eth
) {
382 eth
= ofpbuf_data(&b
);
383 memcpy(flow
->dl_src
, eth
->eth_src
, ETH_ADDR_LEN
);
384 memcpy(flow
->dl_dst
, eth
->eth_dst
, ETH_ADDR_LEN
);
386 /* dl_type, vlan_tci. */
387 ofpbuf_pull(&b
, ETH_ADDR_LEN
* 2);
388 if (eth
->eth_type
== htons(ETH_TYPE_VLAN
)) {
389 parse_vlan(&b
, flow
);
391 flow
->dl_type
= parse_ethertype(&b
);
393 /* Parse mpls, copy l3 ttl. */
394 if (eth_type_mpls(flow
->dl_type
)) {
395 ofpbuf_set_l2_5(packet
, ofpbuf_data(&b
));
396 parse_mpls(&b
, flow
);
400 ofpbuf_set_l3(packet
, ofpbuf_data(&b
));
401 if (flow
->dl_type
== htons(ETH_TYPE_IP
)) {
402 const struct ip_header
*nh
= pull_ip(&b
);
404 ofpbuf_set_l4(packet
, ofpbuf_data(&b
));
406 flow
->nw_src
= get_16aligned_be32(&nh
->ip_src
);
407 flow
->nw_dst
= get_16aligned_be32(&nh
->ip_dst
);
408 flow
->nw_proto
= nh
->ip_proto
;
410 flow
->nw_tos
= nh
->ip_tos
;
411 if (IP_IS_FRAGMENT(nh
->ip_frag_off
)) {
412 flow
->nw_frag
= FLOW_NW_FRAG_ANY
;
413 if (nh
->ip_frag_off
& htons(IP_FRAG_OFF_MASK
)) {
414 flow
->nw_frag
|= FLOW_NW_FRAG_LATER
;
417 flow
->nw_ttl
= nh
->ip_ttl
;
419 if (!(nh
->ip_frag_off
& htons(IP_FRAG_OFF_MASK
))) {
420 if (flow
->nw_proto
== IPPROTO_TCP
) {
422 } else if (flow
->nw_proto
== IPPROTO_UDP
) {
424 } else if (flow
->nw_proto
== IPPROTO_SCTP
) {
425 parse_sctp(&b
, flow
);
426 } else if (flow
->nw_proto
== IPPROTO_ICMP
) {
427 const struct icmp_header
*icmp
= pull_icmp(&b
);
429 flow
->tp_src
= htons(icmp
->icmp_type
);
430 flow
->tp_dst
= htons(icmp
->icmp_code
);
435 } else if (flow
->dl_type
== htons(ETH_TYPE_IPV6
)) {
436 if (parse_ipv6(&b
, flow
)) {
440 ofpbuf_set_l4(packet
, ofpbuf_data(&b
));
441 if (flow
->nw_proto
== IPPROTO_TCP
) {
443 } else if (flow
->nw_proto
== IPPROTO_UDP
) {
445 } else if (flow
->nw_proto
== IPPROTO_SCTP
) {
446 parse_sctp(&b
, flow
);
447 } else if (flow
->nw_proto
== IPPROTO_ICMPV6
) {
448 parse_icmpv6(&b
, flow
);
450 } else if (flow
->dl_type
== htons(ETH_TYPE_ARP
) ||
451 flow
->dl_type
== htons(ETH_TYPE_RARP
)) {
452 const struct arp_eth_header
*arp
= pull_arp(&b
);
453 if (arp
&& arp
->ar_hrd
== htons(1)
454 && arp
->ar_pro
== htons(ETH_TYPE_IP
)
455 && arp
->ar_hln
== ETH_ADDR_LEN
456 && arp
->ar_pln
== 4) {
457 /* We only match on the lower 8 bits of the opcode. */
458 if (ntohs(arp
->ar_op
) <= 0xff) {
459 flow
->nw_proto
= ntohs(arp
->ar_op
);
462 flow
->nw_src
= get_16aligned_be32(&arp
->ar_spa
);
463 flow
->nw_dst
= get_16aligned_be32(&arp
->ar_tpa
);
464 memcpy(flow
->arp_sha
, arp
->ar_sha
, ETH_ADDR_LEN
);
465 memcpy(flow
->arp_tha
, arp
->ar_tha
, ETH_ADDR_LEN
);
470 /* For every bit of a field that is wildcarded in 'wildcards', sets the
471 * corresponding bit in 'flow' to zero. */
473 flow_zero_wildcards(struct flow
*flow
, const struct flow_wildcards
*wildcards
)
475 uint32_t *flow_u32
= (uint32_t *) flow
;
476 const uint32_t *wc_u32
= (const uint32_t *) &wildcards
->masks
;
479 for (i
= 0; i
< FLOW_U32S
; i
++) {
480 flow_u32
[i
] &= wc_u32
[i
];
485 flow_unwildcard_tp_ports(const struct flow
*flow
, struct flow_wildcards
*wc
)
487 if (flow
->nw_proto
!= IPPROTO_ICMP
) {
488 memset(&wc
->masks
.tp_src
, 0xff, sizeof wc
->masks
.tp_src
);
489 memset(&wc
->masks
.tp_dst
, 0xff, sizeof wc
->masks
.tp_dst
);
491 wc
->masks
.tp_src
= htons(0xff);
492 wc
->masks
.tp_dst
= htons(0xff);
496 /* Initializes 'fmd' with the metadata found in 'flow'. */
498 flow_get_metadata(const struct flow
*flow
, struct flow_metadata
*fmd
)
500 BUILD_ASSERT_DECL(FLOW_WC_SEQ
== 25);
502 fmd
->dp_hash
= flow
->dp_hash
;
503 fmd
->recirc_id
= flow
->recirc_id
;
504 fmd
->tun_id
= flow
->tunnel
.tun_id
;
505 fmd
->tun_src
= flow
->tunnel
.ip_src
;
506 fmd
->tun_dst
= flow
->tunnel
.ip_dst
;
507 fmd
->metadata
= flow
->metadata
;
508 memcpy(fmd
->regs
, flow
->regs
, sizeof fmd
->regs
);
509 fmd
->pkt_mark
= flow
->pkt_mark
;
510 fmd
->in_port
= flow
->in_port
.ofp_port
;
514 flow_to_string(const struct flow
*flow
)
516 struct ds ds
= DS_EMPTY_INITIALIZER
;
517 flow_format(&ds
, flow
);
522 flow_tun_flag_to_string(uint32_t flags
)
525 case FLOW_TNL_F_DONT_FRAGMENT
:
527 case FLOW_TNL_F_CSUM
:
537 format_flags(struct ds
*ds
, const char *(*bit_to_string
)(uint32_t),
538 uint32_t flags
, char del
)
546 uint32_t bit
= rightmost_1bit(flags
);
549 s
= bit_to_string(bit
);
551 ds_put_format(ds
, "%s%c", s
, del
);
560 ds_put_format(ds
, "0x%"PRIx32
"%c", bad
, del
);
566 format_flags_masked(struct ds
*ds
, const char *name
,
567 const char *(*bit_to_string
)(uint32_t), uint32_t flags
,
571 ds_put_format(ds
, "%s=", name
);
574 uint32_t bit
= rightmost_1bit(mask
);
575 const char *s
= bit_to_string(bit
);
577 ds_put_format(ds
, "%s%s", (flags
& bit
) ? "+" : "-",
578 s
? s
: "[Unknown]");
584 flow_format(struct ds
*ds
, const struct flow
*flow
)
588 match_wc_init(&match
, flow
);
589 match_format(&match
, ds
, OFP_DEFAULT_PRIORITY
);
593 flow_print(FILE *stream
, const struct flow
*flow
)
595 char *s
= flow_to_string(flow
);
600 /* flow_wildcards functions. */
602 /* Initializes 'wc' as a set of wildcards that matches every packet. */
604 flow_wildcards_init_catchall(struct flow_wildcards
*wc
)
606 memset(&wc
->masks
, 0, sizeof wc
->masks
);
609 /* Clear the metadata and register wildcard masks. They are not packet
612 flow_wildcards_clear_non_packet_fields(struct flow_wildcards
*wc
)
614 memset(&wc
->masks
.metadata
, 0, sizeof wc
->masks
.metadata
);
615 memset(&wc
->masks
.regs
, 0, sizeof wc
->masks
.regs
);
618 /* Returns true if 'wc' matches every packet, false if 'wc' fixes any bits or
621 flow_wildcards_is_catchall(const struct flow_wildcards
*wc
)
623 const uint32_t *wc_u32
= (const uint32_t *) &wc
->masks
;
626 for (i
= 0; i
< FLOW_U32S
; i
++) {
634 /* Sets 'dst' as the bitwise AND of wildcards in 'src1' and 'src2'.
635 * That is, a bit or a field is wildcarded in 'dst' if it is wildcarded
636 * in 'src1' or 'src2' or both. */
638 flow_wildcards_and(struct flow_wildcards
*dst
,
639 const struct flow_wildcards
*src1
,
640 const struct flow_wildcards
*src2
)
642 uint32_t *dst_u32
= (uint32_t *) &dst
->masks
;
643 const uint32_t *src1_u32
= (const uint32_t *) &src1
->masks
;
644 const uint32_t *src2_u32
= (const uint32_t *) &src2
->masks
;
647 for (i
= 0; i
< FLOW_U32S
; i
++) {
648 dst_u32
[i
] = src1_u32
[i
] & src2_u32
[i
];
652 /* Sets 'dst' as the bitwise OR of wildcards in 'src1' and 'src2'. That
653 * is, a bit or a field is wildcarded in 'dst' if it is neither
654 * wildcarded in 'src1' nor 'src2'. */
656 flow_wildcards_or(struct flow_wildcards
*dst
,
657 const struct flow_wildcards
*src1
,
658 const struct flow_wildcards
*src2
)
660 uint32_t *dst_u32
= (uint32_t *) &dst
->masks
;
661 const uint32_t *src1_u32
= (const uint32_t *) &src1
->masks
;
662 const uint32_t *src2_u32
= (const uint32_t *) &src2
->masks
;
665 for (i
= 0; i
< FLOW_U32S
; i
++) {
666 dst_u32
[i
] = src1_u32
[i
] | src2_u32
[i
];
670 /* Perform a bitwise OR of miniflow 'src' flow data with the equivalent
671 * fields in 'dst', storing the result in 'dst'. */
673 flow_union_with_miniflow(struct flow
*dst
, const struct miniflow
*src
)
675 uint32_t *dst_u32
= (uint32_t *) dst
;
676 const uint32_t *p
= src
->values
;
679 for (map
= src
->map
; map
; map
= zero_rightmost_1bit(map
)) {
680 dst_u32
[raw_ctz(map
)] |= *p
++;
684 /* Fold minimask 'mask''s wildcard mask into 'wc's wildcard mask. */
686 flow_wildcards_fold_minimask(struct flow_wildcards
*wc
,
687 const struct minimask
*mask
)
689 flow_union_with_miniflow(&wc
->masks
, &mask
->masks
);
693 miniflow_get_map_in_range(const struct miniflow
*miniflow
,
694 uint8_t start
, uint8_t end
, unsigned int *offset
)
696 uint64_t map
= miniflow
->map
;
700 uint64_t msk
= (UINT64_C(1) << start
) - 1; /* 'start' LSBs set */
701 *offset
= count_1bits(map
& msk
);
704 if (end
< FLOW_U32S
) {
705 uint64_t msk
= (UINT64_C(1) << end
) - 1; /* 'end' LSBs set */
711 /* Fold minimask 'mask''s wildcard mask into 'wc's wildcard mask
712 * in range [start, end). */
714 flow_wildcards_fold_minimask_range(struct flow_wildcards
*wc
,
715 const struct minimask
*mask
,
716 uint8_t start
, uint8_t end
)
718 uint32_t *dst_u32
= (uint32_t *)&wc
->masks
;
720 uint64_t map
= miniflow_get_map_in_range(&mask
->masks
, start
, end
,
722 const uint32_t *p
= mask
->masks
.values
+ offset
;
724 for (; map
; map
= zero_rightmost_1bit(map
)) {
725 dst_u32
[raw_ctz(map
)] |= *p
++;
729 /* Returns a hash of the wildcards in 'wc'. */
731 flow_wildcards_hash(const struct flow_wildcards
*wc
, uint32_t basis
)
733 return flow_hash(&wc
->masks
, basis
);
736 /* Returns true if 'a' and 'b' represent the same wildcards, false if they are
739 flow_wildcards_equal(const struct flow_wildcards
*a
,
740 const struct flow_wildcards
*b
)
742 return flow_equal(&a
->masks
, &b
->masks
);
745 /* Returns true if at least one bit or field is wildcarded in 'a' but not in
746 * 'b', false otherwise. */
748 flow_wildcards_has_extra(const struct flow_wildcards
*a
,
749 const struct flow_wildcards
*b
)
751 const uint32_t *a_u32
= (const uint32_t *) &a
->masks
;
752 const uint32_t *b_u32
= (const uint32_t *) &b
->masks
;
755 for (i
= 0; i
< FLOW_U32S
; i
++) {
756 if ((a_u32
[i
] & b_u32
[i
]) != b_u32
[i
]) {
763 /* Returns true if 'a' and 'b' are equal, except that 0-bits (wildcarded bits)
764 * in 'wc' do not need to be equal in 'a' and 'b'. */
766 flow_equal_except(const struct flow
*a
, const struct flow
*b
,
767 const struct flow_wildcards
*wc
)
769 const uint32_t *a_u32
= (const uint32_t *) a
;
770 const uint32_t *b_u32
= (const uint32_t *) b
;
771 const uint32_t *wc_u32
= (const uint32_t *) &wc
->masks
;
774 for (i
= 0; i
< FLOW_U32S
; i
++) {
775 if ((a_u32
[i
] ^ b_u32
[i
]) & wc_u32
[i
]) {
782 /* Sets the wildcard mask for register 'idx' in 'wc' to 'mask'.
783 * (A 0-bit indicates a wildcard bit.) */
785 flow_wildcards_set_reg_mask(struct flow_wildcards
*wc
, int idx
, uint32_t mask
)
787 wc
->masks
.regs
[idx
] = mask
;
790 /* Calculates the 5-tuple hash from the given flow. */
792 flow_hash_5tuple(const struct flow
*flow
, uint32_t basis
)
800 hash
= mhash_add(basis
, (OVS_FORCE
uint32_t) flow
->nw_src
);
801 hash
= mhash_add(hash
, (OVS_FORCE
uint32_t) flow
->nw_dst
);
802 hash
= mhash_add(hash
, ((OVS_FORCE
uint32_t) flow
->tp_src
<< 16)
803 | (OVS_FORCE
uint32_t) flow
->tp_dst
);
804 hash
= mhash_add(hash
, flow
->nw_proto
);
806 return mhash_finish(hash
, 13);
809 /* Hashes 'flow' based on its L2 through L4 protocol information. */
811 flow_hash_symmetric_l4(const struct flow
*flow
, uint32_t basis
)
816 struct in6_addr ipv6_addr
;
821 uint8_t eth_addr
[ETH_ADDR_LEN
];
827 memset(&fields
, 0, sizeof fields
);
828 for (i
= 0; i
< ETH_ADDR_LEN
; i
++) {
829 fields
.eth_addr
[i
] = flow
->dl_src
[i
] ^ flow
->dl_dst
[i
];
831 fields
.vlan_tci
= flow
->vlan_tci
& htons(VLAN_VID_MASK
);
832 fields
.eth_type
= flow
->dl_type
;
834 /* UDP source and destination port are not taken into account because they
835 * will not necessarily be symmetric in a bidirectional flow. */
836 if (fields
.eth_type
== htons(ETH_TYPE_IP
)) {
837 fields
.ipv4_addr
= flow
->nw_src
^ flow
->nw_dst
;
838 fields
.ip_proto
= flow
->nw_proto
;
839 if (fields
.ip_proto
== IPPROTO_TCP
|| fields
.ip_proto
== IPPROTO_SCTP
) {
840 fields
.tp_port
= flow
->tp_src
^ flow
->tp_dst
;
842 } else if (fields
.eth_type
== htons(ETH_TYPE_IPV6
)) {
843 const uint8_t *a
= &flow
->ipv6_src
.s6_addr
[0];
844 const uint8_t *b
= &flow
->ipv6_dst
.s6_addr
[0];
845 uint8_t *ipv6_addr
= &fields
.ipv6_addr
.s6_addr
[0];
847 for (i
=0; i
<16; i
++) {
848 ipv6_addr
[i
] = a
[i
] ^ b
[i
];
850 fields
.ip_proto
= flow
->nw_proto
;
851 if (fields
.ip_proto
== IPPROTO_TCP
|| fields
.ip_proto
== IPPROTO_SCTP
) {
852 fields
.tp_port
= flow
->tp_src
^ flow
->tp_dst
;
855 return jhash_bytes(&fields
, sizeof fields
, basis
);
858 /* Initialize a flow with random fields that matter for nx_hash_fields. */
860 flow_random_hash_fields(struct flow
*flow
)
862 uint16_t rnd
= random_uint16();
864 /* Initialize to all zeros. */
865 memset(flow
, 0, sizeof *flow
);
867 eth_addr_random(flow
->dl_src
);
868 eth_addr_random(flow
->dl_dst
);
870 flow
->vlan_tci
= (OVS_FORCE ovs_be16
) (random_uint16() & VLAN_VID_MASK
);
872 /* Make most of the random flows IPv4, some IPv6, and rest random. */
873 flow
->dl_type
= rnd
< 0x8000 ? htons(ETH_TYPE_IP
) :
874 rnd
< 0xc000 ? htons(ETH_TYPE_IPV6
) : (OVS_FORCE ovs_be16
)rnd
;
876 if (dl_type_is_ip_any(flow
->dl_type
)) {
877 if (flow
->dl_type
== htons(ETH_TYPE_IP
)) {
878 flow
->nw_src
= (OVS_FORCE ovs_be32
)random_uint32();
879 flow
->nw_dst
= (OVS_FORCE ovs_be32
)random_uint32();
881 random_bytes(&flow
->ipv6_src
, sizeof flow
->ipv6_src
);
882 random_bytes(&flow
->ipv6_dst
, sizeof flow
->ipv6_dst
);
884 /* Make most of IP flows TCP, some UDP or SCTP, and rest random. */
885 rnd
= random_uint16();
886 flow
->nw_proto
= rnd
< 0x8000 ? IPPROTO_TCP
:
887 rnd
< 0xc000 ? IPPROTO_UDP
:
888 rnd
< 0xd000 ? IPPROTO_SCTP
: (uint8_t)rnd
;
889 if (flow
->nw_proto
== IPPROTO_TCP
||
890 flow
->nw_proto
== IPPROTO_UDP
||
891 flow
->nw_proto
== IPPROTO_SCTP
) {
892 flow
->tp_src
= (OVS_FORCE ovs_be16
)random_uint16();
893 flow
->tp_dst
= (OVS_FORCE ovs_be16
)random_uint16();
898 /* Masks the fields in 'wc' that are used by the flow hash 'fields'. */
900 flow_mask_hash_fields(const struct flow
*flow
, struct flow_wildcards
*wc
,
901 enum nx_hash_fields fields
)
904 case NX_HASH_FIELDS_ETH_SRC
:
905 memset(&wc
->masks
.dl_src
, 0xff, sizeof wc
->masks
.dl_src
);
908 case NX_HASH_FIELDS_SYMMETRIC_L4
:
909 memset(&wc
->masks
.dl_src
, 0xff, sizeof wc
->masks
.dl_src
);
910 memset(&wc
->masks
.dl_dst
, 0xff, sizeof wc
->masks
.dl_dst
);
911 if (flow
->dl_type
== htons(ETH_TYPE_IP
)) {
912 memset(&wc
->masks
.nw_src
, 0xff, sizeof wc
->masks
.nw_src
);
913 memset(&wc
->masks
.nw_dst
, 0xff, sizeof wc
->masks
.nw_dst
);
914 } else if (flow
->dl_type
== htons(ETH_TYPE_IPV6
)) {
915 memset(&wc
->masks
.ipv6_src
, 0xff, sizeof wc
->masks
.ipv6_src
);
916 memset(&wc
->masks
.ipv6_dst
, 0xff, sizeof wc
->masks
.ipv6_dst
);
918 if (is_ip_any(flow
)) {
919 memset(&wc
->masks
.nw_proto
, 0xff, sizeof wc
->masks
.nw_proto
);
920 flow_unwildcard_tp_ports(flow
, wc
);
922 wc
->masks
.vlan_tci
|= htons(VLAN_VID_MASK
| VLAN_CFI
);
930 /* Hashes the portions of 'flow' designated by 'fields'. */
932 flow_hash_fields(const struct flow
*flow
, enum nx_hash_fields fields
,
937 case NX_HASH_FIELDS_ETH_SRC
:
938 return jhash_bytes(flow
->dl_src
, sizeof flow
->dl_src
, basis
);
940 case NX_HASH_FIELDS_SYMMETRIC_L4
:
941 return flow_hash_symmetric_l4(flow
, basis
);
947 /* Returns a string representation of 'fields'. */
949 flow_hash_fields_to_str(enum nx_hash_fields fields
)
952 case NX_HASH_FIELDS_ETH_SRC
: return "eth_src";
953 case NX_HASH_FIELDS_SYMMETRIC_L4
: return "symmetric_l4";
954 default: return "<unknown>";
958 /* Returns true if the value of 'fields' is supported. Otherwise false. */
960 flow_hash_fields_valid(enum nx_hash_fields fields
)
962 return fields
== NX_HASH_FIELDS_ETH_SRC
963 || fields
== NX_HASH_FIELDS_SYMMETRIC_L4
;
966 /* Returns a hash value for the bits of 'flow' that are active based on
967 * 'wc', given 'basis'. */
969 flow_hash_in_wildcards(const struct flow
*flow
,
970 const struct flow_wildcards
*wc
, uint32_t basis
)
972 const uint32_t *wc_u32
= (const uint32_t *) &wc
->masks
;
973 const uint32_t *flow_u32
= (const uint32_t *) flow
;
978 for (i
= 0; i
< FLOW_U32S
; i
++) {
979 hash
= mhash_add(hash
, flow_u32
[i
] & wc_u32
[i
]);
981 return mhash_finish(hash
, 4 * FLOW_U32S
);
984 /* Sets the VLAN VID that 'flow' matches to 'vid', which is interpreted as an
985 * OpenFlow 1.0 "dl_vlan" value:
987 * - If it is in the range 0...4095, 'flow->vlan_tci' is set to match
988 * that VLAN. Any existing PCP match is unchanged (it becomes 0 if
989 * 'flow' previously matched packets without a VLAN header).
991 * - If it is OFP_VLAN_NONE, 'flow->vlan_tci' is set to match a packet
992 * without a VLAN tag.
994 * - Other values of 'vid' should not be used. */
996 flow_set_dl_vlan(struct flow
*flow
, ovs_be16 vid
)
998 if (vid
== htons(OFP10_VLAN_NONE
)) {
999 flow
->vlan_tci
= htons(0);
1001 vid
&= htons(VLAN_VID_MASK
);
1002 flow
->vlan_tci
&= ~htons(VLAN_VID_MASK
);
1003 flow
->vlan_tci
|= htons(VLAN_CFI
) | vid
;
1007 /* Sets the VLAN VID that 'flow' matches to 'vid', which is interpreted as an
1008 * OpenFlow 1.2 "vlan_vid" value, that is, the low 13 bits of 'vlan_tci' (VID
1011 flow_set_vlan_vid(struct flow
*flow
, ovs_be16 vid
)
1013 ovs_be16 mask
= htons(VLAN_VID_MASK
| VLAN_CFI
);
1014 flow
->vlan_tci
&= ~mask
;
1015 flow
->vlan_tci
|= vid
& mask
;
1018 /* Sets the VLAN PCP that 'flow' matches to 'pcp', which should be in the
1021 * This function has no effect on the VLAN ID that 'flow' matches.
1023 * After calling this function, 'flow' will not match packets without a VLAN
1026 flow_set_vlan_pcp(struct flow
*flow
, uint8_t pcp
)
1029 flow
->vlan_tci
&= ~htons(VLAN_PCP_MASK
);
1030 flow
->vlan_tci
|= htons((pcp
<< VLAN_PCP_SHIFT
) | VLAN_CFI
);
1033 /* Returns the number of MPLS LSEs present in 'flow'
1035 * Returns 0 if the 'dl_type' of 'flow' is not an MPLS ethernet type.
1036 * Otherwise traverses 'flow''s MPLS label stack stopping at the
1037 * first entry that has the BoS bit set. If no such entry exists then
1038 * the maximum number of LSEs that can be stored in 'flow' is returned.
1041 flow_count_mpls_labels(const struct flow
*flow
, struct flow_wildcards
*wc
)
1044 wc
->masks
.dl_type
= OVS_BE16_MAX
;
1046 if (eth_type_mpls(flow
->dl_type
)) {
1048 int len
= FLOW_MAX_MPLS_LABELS
;
1050 for (i
= 0; i
< len
; i
++) {
1052 wc
->masks
.mpls_lse
[i
] |= htonl(MPLS_BOS_MASK
);
1054 if (flow
->mpls_lse
[i
] & htonl(MPLS_BOS_MASK
)) {
1065 /* Returns the number consecutive of MPLS LSEs, starting at the
1066 * innermost LSE, that are common in 'a' and 'b'.
1068 * 'an' must be flow_count_mpls_labels(a).
1069 * 'bn' must be flow_count_mpls_labels(b).
1072 flow_count_common_mpls_labels(const struct flow
*a
, int an
,
1073 const struct flow
*b
, int bn
,
1074 struct flow_wildcards
*wc
)
1076 int min_n
= MIN(an
, bn
);
1081 int a_last
= an
- 1;
1082 int b_last
= bn
- 1;
1085 for (i
= 0; i
< min_n
; i
++) {
1087 wc
->masks
.mpls_lse
[a_last
- i
] = OVS_BE32_MAX
;
1088 wc
->masks
.mpls_lse
[b_last
- i
] = OVS_BE32_MAX
;
1090 if (a
->mpls_lse
[a_last
- i
] != b
->mpls_lse
[b_last
- i
]) {
1101 /* Adds a new outermost MPLS label to 'flow' and changes 'flow''s Ethernet type
1102 * to 'mpls_eth_type', which must be an MPLS Ethertype.
1104 * If the new label is the first MPLS label in 'flow', it is generated as;
1106 * - label: 2, if 'flow' is IPv6, otherwise 0.
1108 * - TTL: IPv4 or IPv6 TTL, if present and nonzero, otherwise 64.
1110 * - TC: IPv4 or IPv6 TOS, if present, otherwise 0.
1114 * If the new label is the second or label MPLS label in 'flow', it is
1117 * - label: Copied from outer label.
1119 * - TTL: Copied from outer label.
1121 * - TC: Copied from outer label.
1125 * 'n' must be flow_count_mpls_labels(flow). 'n' must be less than
1126 * FLOW_MAX_MPLS_LABELS (because otherwise flow->mpls_lse[] would overflow).
1129 flow_push_mpls(struct flow
*flow
, int n
, ovs_be16 mpls_eth_type
,
1130 struct flow_wildcards
*wc
)
1132 ovs_assert(eth_type_mpls(mpls_eth_type
));
1133 ovs_assert(n
< FLOW_MAX_MPLS_LABELS
);
1135 memset(wc
->masks
.mpls_lse
, 0xff, sizeof wc
->masks
.mpls_lse
);
1139 for (i
= n
; i
>= 1; i
--) {
1140 flow
->mpls_lse
[i
] = flow
->mpls_lse
[i
- 1];
1142 flow
->mpls_lse
[0] = (flow
->mpls_lse
[1]
1143 & htonl(~MPLS_BOS_MASK
));
1145 int label
= 0; /* IPv4 Explicit Null. */
1149 if (flow
->dl_type
== htons(ETH_TYPE_IPV6
)) {
1153 if (is_ip_any(flow
)) {
1154 tc
= (flow
->nw_tos
& IP_DSCP_MASK
) >> 2;
1155 wc
->masks
.nw_tos
|= IP_DSCP_MASK
;
1160 wc
->masks
.nw_ttl
= 0xff;
1163 flow
->mpls_lse
[0] = set_mpls_lse_values(ttl
, tc
, 1, htonl(label
));
1165 /* Clear all L3 and L4 fields. */
1166 BUILD_ASSERT(FLOW_WC_SEQ
== 25);
1167 memset((char *) flow
+ FLOW_SEGMENT_2_ENDS_AT
, 0,
1168 sizeof(struct flow
) - FLOW_SEGMENT_2_ENDS_AT
);
1170 flow
->dl_type
= mpls_eth_type
;
1173 /* Tries to remove the outermost MPLS label from 'flow'. Returns true if
1174 * successful, false otherwise. On success, sets 'flow''s Ethernet type to
1177 * 'n' must be flow_count_mpls_labels(flow). */
1179 flow_pop_mpls(struct flow
*flow
, int n
, ovs_be16 eth_type
,
1180 struct flow_wildcards
*wc
)
1185 /* Nothing to pop. */
1187 } else if (n
== FLOW_MAX_MPLS_LABELS
1188 && !(flow
->mpls_lse
[n
- 1] & htonl(MPLS_BOS_MASK
))) {
1189 /* Can't pop because we don't know what to fill in mpls_lse[n - 1]. */
1193 memset(wc
->masks
.mpls_lse
, 0xff, sizeof wc
->masks
.mpls_lse
);
1194 for (i
= 1; i
< n
; i
++) {
1195 flow
->mpls_lse
[i
- 1] = flow
->mpls_lse
[i
];
1197 flow
->mpls_lse
[n
- 1] = 0;
1198 flow
->dl_type
= eth_type
;
1202 /* Sets the MPLS Label that 'flow' matches to 'label', which is interpreted
1203 * as an OpenFlow 1.1 "mpls_label" value. */
1205 flow_set_mpls_label(struct flow
*flow
, int idx
, ovs_be32 label
)
1207 set_mpls_lse_label(&flow
->mpls_lse
[idx
], label
);
1210 /* Sets the MPLS TTL that 'flow' matches to 'ttl', which should be in the
1213 flow_set_mpls_ttl(struct flow
*flow
, int idx
, uint8_t ttl
)
1215 set_mpls_lse_ttl(&flow
->mpls_lse
[idx
], ttl
);
1218 /* Sets the MPLS TC that 'flow' matches to 'tc', which should be in the
1221 flow_set_mpls_tc(struct flow
*flow
, int idx
, uint8_t tc
)
1223 set_mpls_lse_tc(&flow
->mpls_lse
[idx
], tc
);
1226 /* Sets the MPLS BOS bit that 'flow' matches to which should be 0 or 1. */
1228 flow_set_mpls_bos(struct flow
*flow
, int idx
, uint8_t bos
)
1230 set_mpls_lse_bos(&flow
->mpls_lse
[idx
], bos
);
1233 /* Sets the entire MPLS LSE. */
1235 flow_set_mpls_lse(struct flow
*flow
, int idx
, ovs_be32 lse
)
1237 flow
->mpls_lse
[idx
] = lse
;
1241 flow_compose_l4(struct ofpbuf
*b
, const struct flow
*flow
)
1245 if (!(flow
->nw_frag
& FLOW_NW_FRAG_ANY
)
1246 || !(flow
->nw_frag
& FLOW_NW_FRAG_LATER
)) {
1247 if (flow
->nw_proto
== IPPROTO_TCP
) {
1248 struct tcp_header
*tcp
;
1250 l4_len
= sizeof *tcp
;
1251 tcp
= ofpbuf_put_zeros(b
, l4_len
);
1252 tcp
->tcp_src
= flow
->tp_src
;
1253 tcp
->tcp_dst
= flow
->tp_dst
;
1254 tcp
->tcp_ctl
= TCP_CTL(ntohs(flow
->tcp_flags
), 5);
1255 } else if (flow
->nw_proto
== IPPROTO_UDP
) {
1256 struct udp_header
*udp
;
1258 l4_len
= sizeof *udp
;
1259 udp
= ofpbuf_put_zeros(b
, l4_len
);
1260 udp
->udp_src
= flow
->tp_src
;
1261 udp
->udp_dst
= flow
->tp_dst
;
1262 } else if (flow
->nw_proto
== IPPROTO_SCTP
) {
1263 struct sctp_header
*sctp
;
1265 l4_len
= sizeof *sctp
;
1266 sctp
= ofpbuf_put_zeros(b
, l4_len
);
1267 sctp
->sctp_src
= flow
->tp_src
;
1268 sctp
->sctp_dst
= flow
->tp_dst
;
1269 } else if (flow
->nw_proto
== IPPROTO_ICMP
) {
1270 struct icmp_header
*icmp
;
1272 l4_len
= sizeof *icmp
;
1273 icmp
= ofpbuf_put_zeros(b
, l4_len
);
1274 icmp
->icmp_type
= ntohs(flow
->tp_src
);
1275 icmp
->icmp_code
= ntohs(flow
->tp_dst
);
1276 icmp
->icmp_csum
= csum(icmp
, ICMP_HEADER_LEN
);
1277 } else if (flow
->nw_proto
== IPPROTO_ICMPV6
) {
1278 struct icmp6_hdr
*icmp
;
1280 l4_len
= sizeof *icmp
;
1281 icmp
= ofpbuf_put_zeros(b
, l4_len
);
1282 icmp
->icmp6_type
= ntohs(flow
->tp_src
);
1283 icmp
->icmp6_code
= ntohs(flow
->tp_dst
);
1285 if (icmp
->icmp6_code
== 0 &&
1286 (icmp
->icmp6_type
== ND_NEIGHBOR_SOLICIT
||
1287 icmp
->icmp6_type
== ND_NEIGHBOR_ADVERT
)) {
1288 struct in6_addr
*nd_target
;
1289 struct nd_opt_hdr
*nd_opt
;
1291 l4_len
+= sizeof *nd_target
;
1292 nd_target
= ofpbuf_put_zeros(b
, sizeof *nd_target
);
1293 *nd_target
= flow
->nd_target
;
1295 if (!eth_addr_is_zero(flow
->arp_sha
)) {
1297 nd_opt
= ofpbuf_put_zeros(b
, 8);
1298 nd_opt
->nd_opt_len
= 1;
1299 nd_opt
->nd_opt_type
= ND_OPT_SOURCE_LINKADDR
;
1300 memcpy(nd_opt
+ 1, flow
->arp_sha
, ETH_ADDR_LEN
);
1302 if (!eth_addr_is_zero(flow
->arp_tha
)) {
1304 nd_opt
= ofpbuf_put_zeros(b
, 8);
1305 nd_opt
->nd_opt_len
= 1;
1306 nd_opt
->nd_opt_type
= ND_OPT_TARGET_LINKADDR
;
1307 memcpy(nd_opt
+ 1, flow
->arp_tha
, ETH_ADDR_LEN
);
1310 icmp
->icmp6_cksum
= (OVS_FORCE
uint16_t)
1311 csum(icmp
, (char *)ofpbuf_tail(b
) - (char *)icmp
);
1317 /* Puts into 'b' a packet that flow_extract() would parse as having the given
1320 * (This is useful only for testing, obviously, and the packet isn't really
1321 * valid. It hasn't got some checksums filled in, for one, and lots of fields
1322 * are just zeroed.) */
1324 flow_compose(struct ofpbuf
*b
, const struct flow
*flow
)
1328 /* eth_compose() sets l3 pointer and makes sure it is 32-bit aligned. */
1329 eth_compose(b
, flow
->dl_dst
, flow
->dl_src
, ntohs(flow
->dl_type
), 0);
1330 if (flow
->dl_type
== htons(FLOW_DL_TYPE_NONE
)) {
1331 struct eth_header
*eth
= ofpbuf_l2(b
);
1332 eth
->eth_type
= htons(ofpbuf_size(b
));
1336 if (flow
->vlan_tci
& htons(VLAN_CFI
)) {
1337 eth_push_vlan(b
, htons(ETH_TYPE_VLAN
), flow
->vlan_tci
);
1340 if (flow
->dl_type
== htons(ETH_TYPE_IP
)) {
1341 struct ip_header
*ip
;
1343 ip
= ofpbuf_put_zeros(b
, sizeof *ip
);
1344 ip
->ip_ihl_ver
= IP_IHL_VER(5, 4);
1345 ip
->ip_tos
= flow
->nw_tos
;
1346 ip
->ip_ttl
= flow
->nw_ttl
;
1347 ip
->ip_proto
= flow
->nw_proto
;
1348 put_16aligned_be32(&ip
->ip_src
, flow
->nw_src
);
1349 put_16aligned_be32(&ip
->ip_dst
, flow
->nw_dst
);
1351 if (flow
->nw_frag
& FLOW_NW_FRAG_ANY
) {
1352 ip
->ip_frag_off
|= htons(IP_MORE_FRAGMENTS
);
1353 if (flow
->nw_frag
& FLOW_NW_FRAG_LATER
) {
1354 ip
->ip_frag_off
|= htons(100);
1358 ofpbuf_set_l4(b
, ofpbuf_tail(b
));
1360 l4_len
= flow_compose_l4(b
, flow
);
1362 ip
->ip_tot_len
= htons(b
->l4_ofs
- b
->l3_ofs
+ l4_len
);
1363 ip
->ip_csum
= csum(ip
, sizeof *ip
);
1364 } else if (flow
->dl_type
== htons(ETH_TYPE_IPV6
)) {
1365 struct ovs_16aligned_ip6_hdr
*nh
;
1367 nh
= ofpbuf_put_zeros(b
, sizeof *nh
);
1368 put_16aligned_be32(&nh
->ip6_flow
, htonl(6 << 28) |
1369 htonl(flow
->nw_tos
<< 20) | flow
->ipv6_label
);
1370 nh
->ip6_hlim
= flow
->nw_ttl
;
1371 nh
->ip6_nxt
= flow
->nw_proto
;
1373 memcpy(&nh
->ip6_src
, &flow
->ipv6_src
, sizeof(nh
->ip6_src
));
1374 memcpy(&nh
->ip6_dst
, &flow
->ipv6_dst
, sizeof(nh
->ip6_dst
));
1376 ofpbuf_set_l4(b
, ofpbuf_tail(b
));
1378 l4_len
= flow_compose_l4(b
, flow
);
1380 nh
->ip6_plen
= htons(l4_len
);
1381 } else if (flow
->dl_type
== htons(ETH_TYPE_ARP
) ||
1382 flow
->dl_type
== htons(ETH_TYPE_RARP
)) {
1383 struct arp_eth_header
*arp
;
1385 arp
= ofpbuf_put_zeros(b
, sizeof *arp
);
1386 ofpbuf_set_l3(b
, arp
);
1387 arp
->ar_hrd
= htons(1);
1388 arp
->ar_pro
= htons(ETH_TYPE_IP
);
1389 arp
->ar_hln
= ETH_ADDR_LEN
;
1391 arp
->ar_op
= htons(flow
->nw_proto
);
1393 if (flow
->nw_proto
== ARP_OP_REQUEST
||
1394 flow
->nw_proto
== ARP_OP_REPLY
) {
1395 put_16aligned_be32(&arp
->ar_spa
, flow
->nw_src
);
1396 put_16aligned_be32(&arp
->ar_tpa
, flow
->nw_dst
);
1397 memcpy(arp
->ar_sha
, flow
->arp_sha
, ETH_ADDR_LEN
);
1398 memcpy(arp
->ar_tha
, flow
->arp_tha
, ETH_ADDR_LEN
);
1402 if (eth_type_mpls(flow
->dl_type
)) {
1405 b
->l2_5_ofs
= b
->l3_ofs
;
1406 for (n
= 1; n
< FLOW_MAX_MPLS_LABELS
; n
++) {
1407 if (flow
->mpls_lse
[n
- 1] & htonl(MPLS_BOS_MASK
)) {
1412 push_mpls(b
, flow
->dl_type
, flow
->mpls_lse
[--n
]);
1417 /* Compressed flow. */
1420 miniflow_n_values(const struct miniflow
*flow
)
1422 return count_1bits(flow
->map
);
1426 miniflow_alloc_values(struct miniflow
*flow
, int n
)
1428 if (n
<= MINI_N_INLINE
) {
1429 return flow
->inline_values
;
1431 COVERAGE_INC(miniflow_malloc
);
1432 return xmalloc(n
* sizeof *flow
->values
);
1436 /* Completes an initialization of 'dst' as a miniflow copy of 'src' begun by
1437 * the caller. The caller must have already initialized 'dst->map' properly
1438 * to indicate the significant uint32_t elements of 'src'. 'n' must be the
1439 * number of 1-bits in 'dst->map'.
1441 * Normally the significant elements are the ones that are non-zero. However,
1442 * when a miniflow is initialized from a (mini)mask, the values can be zeroes,
1443 * so that the flow and mask always have the same maps.
1445 * This function initializes 'dst->values' (either inline if possible or with
1446 * malloc() otherwise) and copies the uint32_t elements of 'src' indicated by
1447 * 'dst->map' into it. */
1449 miniflow_init__(struct miniflow
*dst
, const struct flow
*src
, int n
)
1451 const uint32_t *src_u32
= (const uint32_t *) src
;
1455 dst
->values
= miniflow_alloc_values(dst
, n
);
1457 for (map
= dst
->map
; map
; map
= zero_rightmost_1bit(map
)) {
1458 dst
->values
[ofs
++] = src_u32
[raw_ctz(map
)];
1462 /* Initializes 'dst' as a copy of 'src'. The caller must eventually free 'dst'
1463 * with miniflow_destroy(). */
1465 miniflow_init(struct miniflow
*dst
, const struct flow
*src
)
1467 const uint32_t *src_u32
= (const uint32_t *) src
;
1471 /* Initialize dst->map, counting the number of nonzero elements. */
1475 for (i
= 0; i
< FLOW_U32S
; i
++) {
1477 dst
->map
|= UINT64_C(1) << i
;
1482 miniflow_init__(dst
, src
, n
);
1485 /* Initializes 'dst' as a copy of 'src', using 'mask->map' as 'dst''s map. The
1486 * caller must eventually free 'dst' with miniflow_destroy(). */
1488 miniflow_init_with_minimask(struct miniflow
*dst
, const struct flow
*src
,
1489 const struct minimask
*mask
)
1491 dst
->map
= mask
->masks
.map
;
1492 miniflow_init__(dst
, src
, miniflow_n_values(dst
));
1495 /* Initializes 'dst' as a copy of 'src'. The caller must eventually free 'dst'
1496 * with miniflow_destroy(). */
1498 miniflow_clone(struct miniflow
*dst
, const struct miniflow
*src
)
1500 int n
= miniflow_n_values(src
);
1501 dst
->map
= src
->map
;
1502 dst
->values
= miniflow_alloc_values(dst
, n
);
1503 memcpy(dst
->values
, src
->values
, n
* sizeof *dst
->values
);
1506 /* Initializes 'dst' with the data in 'src', destroying 'src'.
1507 * The caller must eventually free 'dst' with miniflow_destroy(). */
1509 miniflow_move(struct miniflow
*dst
, struct miniflow
*src
)
1511 if (src
->values
== src
->inline_values
) {
1512 dst
->values
= dst
->inline_values
;
1513 memcpy(dst
->values
, src
->values
,
1514 miniflow_n_values(src
) * sizeof *dst
->values
);
1516 dst
->values
= src
->values
;
1518 dst
->map
= src
->map
;
1521 /* Frees any memory owned by 'flow'. Does not free the storage in which 'flow'
1522 * itself resides; the caller is responsible for that. */
1524 miniflow_destroy(struct miniflow
*flow
)
1526 if (flow
->values
!= flow
->inline_values
) {
1531 /* Initializes 'dst' as a copy of 'src'. */
1533 miniflow_expand(const struct miniflow
*src
, struct flow
*dst
)
1535 memset(dst
, 0, sizeof *dst
);
1536 flow_union_with_miniflow(dst
, src
);
1539 static const uint32_t *
1540 miniflow_get__(const struct miniflow
*flow
, unsigned int u32_ofs
)
1542 if (!(flow
->map
& (UINT64_C(1) << u32_ofs
))) {
1543 static const uint32_t zero
= 0;
1546 return flow
->values
+
1547 count_1bits(flow
->map
& ((UINT64_C(1) << u32_ofs
) - 1));
1550 /* Returns the uint32_t that would be at byte offset '4 * u32_ofs' if 'flow'
1551 * were expanded into a "struct flow". */
1553 miniflow_get(const struct miniflow
*flow
, unsigned int u32_ofs
)
1555 return *miniflow_get__(flow
, u32_ofs
);
1558 /* Returns the ovs_be16 that would be at byte offset 'u8_ofs' if 'flow' were
1559 * expanded into a "struct flow". */
1561 miniflow_get_be16(const struct miniflow
*flow
, unsigned int u8_ofs
)
1563 const uint32_t *u32p
= miniflow_get__(flow
, u8_ofs
/ 4);
1564 const ovs_be16
*be16p
= (const ovs_be16
*) u32p
;
1565 return be16p
[u8_ofs
% 4 != 0];
1568 /* Returns the VID within the vlan_tci member of the "struct flow" represented
1571 miniflow_get_vid(const struct miniflow
*flow
)
1573 ovs_be16 tci
= miniflow_get_be16(flow
, offsetof(struct flow
, vlan_tci
));
1574 return vlan_tci_to_vid(tci
);
1577 /* Returns true if 'a' and 'b' are the same flow, false otherwise. */
1579 miniflow_equal(const struct miniflow
*a
, const struct miniflow
*b
)
1581 const uint32_t *ap
= a
->values
;
1582 const uint32_t *bp
= b
->values
;
1583 const uint64_t a_map
= a
->map
;
1584 const uint64_t b_map
= b
->map
;
1587 if (a_map
== b_map
) {
1588 for (map
= a_map
; map
; map
= zero_rightmost_1bit(map
)) {
1589 if (*ap
++ != *bp
++) {
1594 for (map
= a_map
| b_map
; map
; map
= zero_rightmost_1bit(map
)) {
1595 uint64_t bit
= rightmost_1bit(map
);
1596 uint64_t a_value
= a_map
& bit
? *ap
++ : 0;
1597 uint64_t b_value
= b_map
& bit
? *bp
++ : 0;
1599 if (a_value
!= b_value
) {
1608 /* Returns true if 'a' and 'b' are equal at the places where there are 1-bits
1609 * in 'mask', false if they differ. */
1611 miniflow_equal_in_minimask(const struct miniflow
*a
, const struct miniflow
*b
,
1612 const struct minimask
*mask
)
1617 p
= mask
->masks
.values
;
1619 for (map
= mask
->masks
.map
; map
; map
= zero_rightmost_1bit(map
)) {
1620 int ofs
= raw_ctz(map
);
1622 if ((miniflow_get(a
, ofs
) ^ miniflow_get(b
, ofs
)) & *p
) {
1631 /* Returns true if 'a' and 'b' are equal at the places where there are 1-bits
1632 * in 'mask', false if they differ. */
1634 miniflow_equal_flow_in_minimask(const struct miniflow
*a
, const struct flow
*b
,
1635 const struct minimask
*mask
)
1637 const uint32_t *b_u32
= (const uint32_t *) b
;
1641 p
= mask
->masks
.values
;
1643 for (map
= mask
->masks
.map
; map
; map
= zero_rightmost_1bit(map
)) {
1644 int ofs
= raw_ctz(map
);
1646 if ((miniflow_get(a
, ofs
) ^ b_u32
[ofs
]) & *p
) {
1655 /* Returns a hash value for 'flow', given 'basis'. */
1657 miniflow_hash(const struct miniflow
*flow
, uint32_t basis
)
1659 const uint32_t *p
= flow
->values
;
1660 uint32_t hash
= basis
;
1661 uint64_t hash_map
= 0;
1664 for (map
= flow
->map
; map
; map
= zero_rightmost_1bit(map
)) {
1666 hash
= mhash_add(hash
, *p
);
1667 hash_map
|= rightmost_1bit(map
);
1671 hash
= mhash_add(hash
, hash_map
);
1672 hash
= mhash_add(hash
, hash_map
>> 32);
1674 return mhash_finish(hash
, p
- flow
->values
);
1677 /* Returns a hash value for the bits of 'flow' where there are 1-bits in
1678 * 'mask', given 'basis'.
1680 * The hash values returned by this function are the same as those returned by
1681 * flow_hash_in_minimask(), only the form of the arguments differ. */
1683 miniflow_hash_in_minimask(const struct miniflow
*flow
,
1684 const struct minimask
*mask
, uint32_t basis
)
1686 const uint32_t *p
= mask
->masks
.values
;
1692 for (map
= mask
->masks
.map
; map
; map
= zero_rightmost_1bit(map
)) {
1693 hash
= mhash_add(hash
, miniflow_get(flow
, raw_ctz(map
)) & *p
++);
1696 return mhash_finish(hash
, (p
- mask
->masks
.values
) * 4);
1699 /* Returns a hash value for the bits of 'flow' where there are 1-bits in
1700 * 'mask', given 'basis'.
1702 * The hash values returned by this function are the same as those returned by
1703 * miniflow_hash_in_minimask(), only the form of the arguments differ. */
1705 flow_hash_in_minimask(const struct flow
*flow
, const struct minimask
*mask
,
1708 const uint32_t *flow_u32
= (const uint32_t *)flow
;
1709 const uint32_t *p
= mask
->masks
.values
;
1714 for (map
= mask
->masks
.map
; map
; map
= zero_rightmost_1bit(map
)) {
1715 hash
= mhash_add(hash
, flow_u32
[raw_ctz(map
)] & *p
++);
1718 return mhash_finish(hash
, (p
- mask
->masks
.values
) * 4);
1721 /* Returns a hash value for the bits of range [start, end) in 'flow',
1722 * where there are 1-bits in 'mask', given 'hash'.
1724 * The hash values returned by this function are the same as those returned by
1725 * minimatch_hash_range(), only the form of the arguments differ. */
1727 flow_hash_in_minimask_range(const struct flow
*flow
,
1728 const struct minimask
*mask
,
1729 uint8_t start
, uint8_t end
, uint32_t *basis
)
1731 const uint32_t *flow_u32
= (const uint32_t *)flow
;
1732 unsigned int offset
;
1733 uint64_t map
= miniflow_get_map_in_range(&mask
->masks
, start
, end
,
1735 const uint32_t *p
= mask
->masks
.values
+ offset
;
1736 uint32_t hash
= *basis
;
1738 for (; map
; map
= zero_rightmost_1bit(map
)) {
1739 hash
= mhash_add(hash
, flow_u32
[raw_ctz(map
)] & *p
++);
1742 *basis
= hash
; /* Allow continuation from the unfinished value. */
1743 return mhash_finish(hash
, (p
- mask
->masks
.values
) * 4);
1747 /* Initializes 'dst' as a copy of 'src'. The caller must eventually free 'dst'
1748 * with minimask_destroy(). */
1750 minimask_init(struct minimask
*mask
, const struct flow_wildcards
*wc
)
1752 miniflow_init(&mask
->masks
, &wc
->masks
);
1755 /* Initializes 'dst' as a copy of 'src'. The caller must eventually free 'dst'
1756 * with minimask_destroy(). */
1758 minimask_clone(struct minimask
*dst
, const struct minimask
*src
)
1760 miniflow_clone(&dst
->masks
, &src
->masks
);
1763 /* Initializes 'dst' with the data in 'src', destroying 'src'.
1764 * The caller must eventually free 'dst' with minimask_destroy(). */
1766 minimask_move(struct minimask
*dst
, struct minimask
*src
)
1768 miniflow_move(&dst
->masks
, &src
->masks
);
1771 /* Initializes 'dst_' as the bit-wise "and" of 'a_' and 'b_'.
1773 * The caller must provide room for FLOW_U32S "uint32_t"s in 'storage', for use
1774 * by 'dst_'. The caller must *not* free 'dst_' with minimask_destroy(). */
1776 minimask_combine(struct minimask
*dst_
,
1777 const struct minimask
*a_
, const struct minimask
*b_
,
1778 uint32_t storage
[FLOW_U32S
])
1780 struct miniflow
*dst
= &dst_
->masks
;
1781 const struct miniflow
*a
= &a_
->masks
;
1782 const struct miniflow
*b
= &b_
->masks
;
1786 dst
->values
= storage
;
1789 for (map
= a
->map
& b
->map
; map
; map
= zero_rightmost_1bit(map
)) {
1790 int ofs
= raw_ctz(map
);
1791 uint32_t mask
= miniflow_get(a
, ofs
) & miniflow_get(b
, ofs
);
1794 dst
->map
|= rightmost_1bit(map
);
1795 dst
->values
[n
++] = mask
;
1800 /* Frees any memory owned by 'mask'. Does not free the storage in which 'mask'
1801 * itself resides; the caller is responsible for that. */
1803 minimask_destroy(struct minimask
*mask
)
1805 miniflow_destroy(&mask
->masks
);
1808 /* Initializes 'dst' as a copy of 'src'. */
1810 minimask_expand(const struct minimask
*mask
, struct flow_wildcards
*wc
)
1812 miniflow_expand(&mask
->masks
, &wc
->masks
);
1815 /* Returns the uint32_t that would be at byte offset '4 * u32_ofs' if 'mask'
1816 * were expanded into a "struct flow_wildcards". */
1818 minimask_get(const struct minimask
*mask
, unsigned int u32_ofs
)
1820 return miniflow_get(&mask
->masks
, u32_ofs
);
1823 /* Returns the VID mask within the vlan_tci member of the "struct
1824 * flow_wildcards" represented by 'mask'. */
1826 minimask_get_vid_mask(const struct minimask
*mask
)
1828 return miniflow_get_vid(&mask
->masks
);
1831 /* Returns true if 'a' and 'b' are the same flow mask, false otherwise. */
1833 minimask_equal(const struct minimask
*a
, const struct minimask
*b
)
1835 return miniflow_equal(&a
->masks
, &b
->masks
);
1838 /* Returns a hash value for 'mask', given 'basis'. */
1840 minimask_hash(const struct minimask
*mask
, uint32_t basis
)
1842 return miniflow_hash(&mask
->masks
, basis
);
1845 /* Returns true if at least one bit is wildcarded in 'a_' but not in 'b_',
1846 * false otherwise. */
1848 minimask_has_extra(const struct minimask
*a_
, const struct minimask
*b_
)
1850 const struct miniflow
*a
= &a_
->masks
;
1851 const struct miniflow
*b
= &b_
->masks
;
1854 for (map
= a
->map
| b
->map
; map
; map
= zero_rightmost_1bit(map
)) {
1855 int ofs
= raw_ctz(map
);
1856 uint32_t a_u32
= miniflow_get(a
, ofs
);
1857 uint32_t b_u32
= miniflow_get(b
, ofs
);
1859 if ((a_u32
& b_u32
) != b_u32
) {
1867 /* Returns true if 'mask' matches every packet, false if 'mask' fixes any bits
1870 minimask_is_catchall(const struct minimask
*mask_
)
1872 const struct miniflow
*mask
= &mask_
->masks
;
1873 const uint32_t *p
= mask
->values
;
1876 for (map
= mask
->map
; map
; map
= zero_rightmost_1bit(map
)) {