X-Git-Url: http://git.onelab.eu/?a=blobdiff_plain;f=lib%2Fflow.c;h=4e8cc0e0759756b31d4ae4c4077d8ef0d6a3b280;hb=7e36ac42e33a89d5d2c981ea7750939a1da6db89;hp=31fd07cbaf47b1713b56d0d70da6caeeab430916;hpb=dc235f7fbcff39b318b715f9d3d01f97b2634357;p=sliver-openvswitch.git diff --git a/lib/flow.c b/lib/flow.c index 31fd07cba..4e8cc0e07 100644 --- a/lib/flow.c +++ b/lib/flow.c @@ -1,5 +1,5 @@ /* - * Copyright (c) 2008, 2009, 2010, 2011, 2012, 2013 Nicira, Inc. + * Copyright (c) 2008, 2009, 2010, 2011, 2012, 2013, 2014 Nicira, Inc. * * Licensed under the Apache License, Version 2.0 (the "License"); * you may not use this file except in compliance with the License. @@ -35,12 +35,21 @@ #include "ofpbuf.h" #include "openflow/openflow.h" #include "packets.h" +#include "odp-util.h" #include "random.h" #include "unaligned.h" COVERAGE_DEFINE(flow_extract); COVERAGE_DEFINE(miniflow_malloc); +/* U32 indices for segmented flow classification. */ +const uint8_t flow_segment_u32s[4] = { + FLOW_SEGMENT_1_ENDS_AT / 4, + FLOW_SEGMENT_2_ENDS_AT / 4, + FLOW_SEGMENT_3_ENDS_AT / 4, + FLOW_U32S +}; + static struct arp_eth_header * pull_arp(struct ofpbuf *packet) { @@ -60,31 +69,6 @@ pull_ip(struct ofpbuf *packet) return NULL; } -static struct tcp_header * -pull_tcp(struct ofpbuf *packet) -{ - if (packet->size >= TCP_HEADER_LEN) { - struct tcp_header *tcp = packet->data; - int tcp_len = TCP_OFFSET(tcp->tcp_ctl) * 4; - if (tcp_len >= TCP_HEADER_LEN && packet->size >= tcp_len) { - return ofpbuf_pull(packet, tcp_len); - } - } - return NULL; -} - -static struct udp_header * -pull_udp(struct ofpbuf *packet) -{ - return ofpbuf_try_pull(packet, UDP_HEADER_LEN); -} - -static struct sctp_header * -pull_sctp(struct ofpbuf *packet) -{ - return ofpbuf_try_pull(packet, SCTP_HEADER_LEN); -} - static struct icmp_header * pull_icmp(struct ofpbuf *packet) { @@ -101,12 +85,11 @@ static void parse_mpls(struct ofpbuf *b, struct flow *flow) { struct mpls_hdr *mh; - bool top = true; + int idx = 0; while ((mh = ofpbuf_try_pull(b, sizeof *mh))) { - if (top) { - top = false; - flow->mpls_lse = mh->mpls_lse; + if (idx < FLOW_MAX_MPLS_LABELS) { + flow->mpls_lse[idx++] = mh->mpls_lse; } if (mh->mpls_lse & htonl(MPLS_BOS_MASK)) { break; @@ -250,46 +233,46 @@ parse_ipv6(struct ofpbuf *packet, struct flow *flow) } static void -parse_tcp(struct ofpbuf *packet, struct ofpbuf *b, struct flow *flow) +parse_tcp(struct ofpbuf *b, struct flow *flow) { - const struct tcp_header *tcp = pull_tcp(b); - if (tcp) { + if (b->size >= TCP_HEADER_LEN) { + const struct tcp_header *tcp = b->data; + flow->tp_src = tcp->tcp_src; flow->tp_dst = tcp->tcp_dst; flow->tcp_flags = tcp->tcp_ctl & htons(0x0fff); - packet->l7 = b->data; } } static void -parse_udp(struct ofpbuf *packet, struct ofpbuf *b, struct flow *flow) +parse_udp(struct ofpbuf *b, struct flow *flow) { - const struct udp_header *udp = pull_udp(b); - if (udp) { + if (b->size >= UDP_HEADER_LEN) { + const struct udp_header *udp = b->data; + flow->tp_src = udp->udp_src; flow->tp_dst = udp->udp_dst; - packet->l7 = b->data; } } static void -parse_sctp(struct ofpbuf *packet, struct ofpbuf *b, struct flow *flow) +parse_sctp(struct ofpbuf *b, struct flow *flow) { - const struct sctp_header *sctp = pull_sctp(b); - if (sctp) { + if (b->size >= SCTP_HEADER_LEN) { + const struct sctp_header *sctp = b->data; + flow->tp_src = sctp->sctp_src; flow->tp_dst = sctp->sctp_dst; - packet->l7 = b->data; } } -static bool +static void parse_icmpv6(struct ofpbuf *b, struct flow *flow) { const struct icmp6_hdr *icmp = pull_icmpv6(b); if (!icmp) { - return false; + return; } /* The ICMPv6 type and code fields use the 16-bit transport port @@ -304,7 +287,7 @@ parse_icmpv6(struct ofpbuf *b, struct flow *flow) nd_target = ofpbuf_try_pull(b, sizeof *nd_target); if (!nd_target) { - return false; + return; } flow->nd_target = *nd_target; @@ -343,19 +326,17 @@ parse_icmpv6(struct ofpbuf *b, struct flow *flow) } } - return true; + return; invalid: memset(&flow->nd_target, 0, sizeof(flow->nd_target)); memset(flow->arp_sha, 0, sizeof(flow->arp_sha)); memset(flow->arp_tha, 0, sizeof(flow->arp_tha)); - return false; - + return; } -/* Initializes 'flow' members from 'packet', 'skb_priority', 'tnl', and - * 'in_port'. +/* Initializes 'flow' members from 'packet' and 'md' * * Initializes 'packet' header pointers as follows: * @@ -369,13 +350,9 @@ invalid: * * - packet->l4 to just past the IPv4 header, if one is present and has a * correct length, and otherwise NULL. - * - * - packet->l7 to just past the TCP/UDP/SCTP/ICMP header, if one is - * present and has a correct length, and otherwise NULL. */ void -flow_extract(struct ofpbuf *packet, uint32_t skb_priority, uint32_t pkt_mark, - const struct flow_tnl *tnl, const union flow_in_port *in_port, +flow_extract(struct ofpbuf *packet, const struct pkt_metadata *md, struct flow *flow) { struct ofpbuf b = *packet; @@ -385,21 +362,17 @@ flow_extract(struct ofpbuf *packet, uint32_t skb_priority, uint32_t pkt_mark, memset(flow, 0, sizeof *flow); - if (tnl) { - ovs_assert(tnl != &flow->tunnel); - flow->tunnel = *tnl; - } - if (in_port) { - flow->in_port = *in_port; + if (md) { + flow->tunnel = md->tunnel; + flow->in_port = md->in_port; + flow->skb_priority = md->skb_priority; + flow->pkt_mark = md->pkt_mark; } - flow->skb_priority = skb_priority; - flow->pkt_mark = pkt_mark; packet->l2 = b.data; packet->l2_5 = NULL; packet->l3 = NULL; packet->l4 = NULL; - packet->l7 = NULL; if (b.size < sizeof *eth) { return; @@ -445,17 +418,16 @@ flow_extract(struct ofpbuf *packet, uint32_t skb_priority, uint32_t pkt_mark, if (!(nh->ip_frag_off & htons(IP_FRAG_OFF_MASK))) { if (flow->nw_proto == IPPROTO_TCP) { - parse_tcp(packet, &b, flow); + parse_tcp(&b, flow); } else if (flow->nw_proto == IPPROTO_UDP) { - parse_udp(packet, &b, flow); + parse_udp(&b, flow); } else if (flow->nw_proto == IPPROTO_SCTP) { - parse_sctp(packet, &b, flow); + parse_sctp(&b, flow); } else if (flow->nw_proto == IPPROTO_ICMP) { const struct icmp_header *icmp = pull_icmp(&b); if (icmp) { flow->tp_src = htons(icmp->icmp_type); flow->tp_dst = htons(icmp->icmp_code); - packet->l7 = b.data; } } } @@ -467,15 +439,13 @@ flow_extract(struct ofpbuf *packet, uint32_t skb_priority, uint32_t pkt_mark, packet->l4 = b.data; if (flow->nw_proto == IPPROTO_TCP) { - parse_tcp(packet, &b, flow); + parse_tcp(&b, flow); } else if (flow->nw_proto == IPPROTO_UDP) { - parse_udp(packet, &b, flow); + parse_udp(&b, flow); } else if (flow->nw_proto == IPPROTO_SCTP) { - parse_sctp(packet, &b, flow); + parse_sctp(&b, flow); } else if (flow->nw_proto == IPPROTO_ICMPV6) { - if (parse_icmpv6(&b, flow)) { - packet->l7 = b.data; - } + parse_icmpv6(&b, flow); } } else if (flow->dl_type == htons(ETH_TYPE_ARP) || flow->dl_type == htons(ETH_TYPE_RARP)) { @@ -511,12 +481,26 @@ flow_zero_wildcards(struct flow *flow, const struct flow_wildcards *wildcards) } } +void +flow_unwildcard_tp_ports(const struct flow *flow, struct flow_wildcards *wc) +{ + if (flow->nw_proto != IPPROTO_ICMP) { + memset(&wc->masks.tp_src, 0xff, sizeof wc->masks.tp_src); + memset(&wc->masks.tp_dst, 0xff, sizeof wc->masks.tp_dst); + } else { + wc->masks.tp_src = htons(0xff); + wc->masks.tp_dst = htons(0xff); + } +} + /* Initializes 'fmd' with the metadata found in 'flow'. */ void flow_get_metadata(const struct flow *flow, struct flow_metadata *fmd) { - BUILD_ASSERT_DECL(FLOW_WC_SEQ == 22); + BUILD_ASSERT_DECL(FLOW_WC_SEQ == 25); + fmd->dp_hash = flow->dp_hash; + fmd->recirc_id = flow->recirc_id; fmd->tun_id = flow->tunnel.tun_id; fmd->tun_src = flow->tunnel.ip_src; fmd->tun_dst = flow->tunnel.ip_dst; @@ -578,6 +562,24 @@ format_flags(struct ds *ds, const char *(*bit_to_string)(uint32_t), ds_chomp(ds, del); } +void +format_flags_masked(struct ds *ds, const char *name, + const char *(*bit_to_string)(uint32_t), uint32_t flags, + uint32_t mask) +{ + if (name) { + ds_put_format(ds, "%s=", name); + } + while (mask) { + uint32_t bit = rightmost_1bit(mask); + const char *s = bit_to_string(bit); + + ds_put_format(ds, "%s%s", (flags & bit) ? "+" : "-", + s ? s : "[Unknown]"); + mask &= ~bit; + } +} + void flow_format(struct ds *ds, const struct flow *flow) { @@ -604,6 +606,15 @@ flow_wildcards_init_catchall(struct flow_wildcards *wc) memset(&wc->masks, 0, sizeof wc->masks); } +/* Clear the metadata and register wildcard masks. They are not packet + * header fields. */ +void +flow_wildcards_clear_non_packet_fields(struct flow_wildcards *wc) +{ + memset(&wc->masks.metadata, 0, sizeof wc->masks.metadata); + memset(&wc->masks.regs, 0, sizeof wc->masks.regs); +} + /* Returns true if 'wc' matches every packet, false if 'wc' fixes any bits or * fields. */ bool @@ -662,16 +673,11 @@ static void flow_union_with_miniflow(struct flow *dst, const struct miniflow *src) { uint32_t *dst_u32 = (uint32_t *) dst; - int ofs; - int i; - - ofs = 0; - for (i = 0; i < MINI_N_MAPS; i++) { - uint32_t map; + const uint32_t *p = src->values; + uint64_t map; - for (map = src->map[i]; map; map = zero_rightmost_1bit(map)) { - dst_u32[raw_ctz(map) + i * 32] |= src->values[ofs++]; - } + for (map = src->map; map; map = zero_rightmost_1bit(map)) { + dst_u32[raw_ctz(map)] |= *p++; } } @@ -683,6 +689,43 @@ flow_wildcards_fold_minimask(struct flow_wildcards *wc, flow_union_with_miniflow(&wc->masks, &mask->masks); } +uint64_t +miniflow_get_map_in_range(const struct miniflow *miniflow, + uint8_t start, uint8_t end, unsigned int *offset) +{ + uint64_t map = miniflow->map; + *offset = 0; + + if (start > 0) { + uint64_t msk = (UINT64_C(1) << start) - 1; /* 'start' LSBs set */ + *offset = count_1bits(map & msk); + map &= ~msk; + } + if (end < FLOW_U32S) { + uint64_t msk = (UINT64_C(1) << end) - 1; /* 'end' LSBs set */ + map &= msk; + } + return map; +} + +/* Fold minimask 'mask''s wildcard mask into 'wc's wildcard mask + * in range [start, end). */ +void +flow_wildcards_fold_minimask_range(struct flow_wildcards *wc, + const struct minimask *mask, + uint8_t start, uint8_t end) +{ + uint32_t *dst_u32 = (uint32_t *)&wc->masks; + unsigned int offset; + uint64_t map = miniflow_get_map_in_range(&mask->masks, start, end, + &offset); + const uint32_t *p = mask->masks.values + offset; + + for (; map; map = zero_rightmost_1bit(map)) { + dst_u32[raw_ctz(map)] |= *p++; + } +} + /* Returns a hash of the wildcards in 'wc'. */ uint32_t flow_wildcards_hash(const struct flow_wildcards *wc, uint32_t basis) @@ -744,6 +787,25 @@ flow_wildcards_set_reg_mask(struct flow_wildcards *wc, int idx, uint32_t mask) wc->masks.regs[idx] = mask; } +/* Calculates the 5-tuple hash from the given flow. */ +uint32_t +flow_hash_5tuple(const struct flow *flow, uint32_t basis) +{ + uint32_t hash = 0; + + if (!flow) { + return 0; + } + + hash = mhash_add(basis, (OVS_FORCE uint32_t) flow->nw_src); + hash = mhash_add(hash, (OVS_FORCE uint32_t) flow->nw_dst); + hash = mhash_add(hash, ((OVS_FORCE uint32_t) flow->tp_src << 16) + | (OVS_FORCE uint32_t) flow->tp_dst); + hash = mhash_add(hash, flow->nw_proto); + + return mhash_finish(hash, 13); +} + /* Hashes 'flow' based on its L2 through L4 protocol information. */ uint32_t flow_hash_symmetric_l4(const struct flow *flow, uint32_t basis) @@ -855,14 +917,13 @@ flow_mask_hash_fields(const struct flow *flow, struct flow_wildcards *wc, } if (is_ip_any(flow)) { memset(&wc->masks.nw_proto, 0xff, sizeof wc->masks.nw_proto); - memset(&wc->masks.tp_src, 0xff, sizeof wc->masks.tp_src); - memset(&wc->masks.tp_dst, 0xff, sizeof wc->masks.tp_dst); + flow_unwildcard_tp_ports(flow, wc); } wc->masks.vlan_tci |= htons(VLAN_VID_MASK | VLAN_CFI); break; default: - NOT_REACHED(); + OVS_NOT_REACHED(); } } @@ -880,7 +941,7 @@ flow_hash_fields(const struct flow *flow, enum nx_hash_fields fields, return flow_hash_symmetric_l4(flow, basis); } - NOT_REACHED(); + OVS_NOT_REACHED(); } /* Returns a string representation of 'fields'. */ @@ -969,35 +1030,283 @@ flow_set_vlan_pcp(struct flow *flow, uint8_t pcp) flow->vlan_tci |= htons((pcp << VLAN_PCP_SHIFT) | VLAN_CFI); } +/* Returns the number of MPLS LSEs present in 'flow' + * + * Returns 0 if the 'dl_type' of 'flow' is not an MPLS ethernet type. + * Otherwise traverses 'flow''s MPLS label stack stopping at the + * first entry that has the BoS bit set. If no such entry exists then + * the maximum number of LSEs that can be stored in 'flow' is returned. + */ +int +flow_count_mpls_labels(const struct flow *flow, struct flow_wildcards *wc) +{ + if (wc) { + wc->masks.dl_type = OVS_BE16_MAX; + } + if (eth_type_mpls(flow->dl_type)) { + int i; + int len = FLOW_MAX_MPLS_LABELS; + + for (i = 0; i < len; i++) { + if (wc) { + wc->masks.mpls_lse[i] |= htonl(MPLS_BOS_MASK); + } + if (flow->mpls_lse[i] & htonl(MPLS_BOS_MASK)) { + return i + 1; + } + } + + return len; + } else { + return 0; + } +} + +/* Returns the number consecutive of MPLS LSEs, starting at the + * innermost LSE, that are common in 'a' and 'b'. + * + * 'an' must be flow_count_mpls_labels(a). + * 'bn' must be flow_count_mpls_labels(b). + */ +int +flow_count_common_mpls_labels(const struct flow *a, int an, + const struct flow *b, int bn, + struct flow_wildcards *wc) +{ + int min_n = MIN(an, bn); + if (min_n == 0) { + return 0; + } else { + int common_n = 0; + int a_last = an - 1; + int b_last = bn - 1; + int i; + + for (i = 0; i < min_n; i++) { + if (wc) { + wc->masks.mpls_lse[a_last - i] = OVS_BE32_MAX; + wc->masks.mpls_lse[b_last - i] = OVS_BE32_MAX; + } + if (a->mpls_lse[a_last - i] != b->mpls_lse[b_last - i]) { + break; + } else { + common_n++; + } + } + + return common_n; + } +} + +/* Adds a new outermost MPLS label to 'flow' and changes 'flow''s Ethernet type + * to 'mpls_eth_type', which must be an MPLS Ethertype. + * + * If the new label is the first MPLS label in 'flow', it is generated as; + * + * - label: 2, if 'flow' is IPv6, otherwise 0. + * + * - TTL: IPv4 or IPv6 TTL, if present and nonzero, otherwise 64. + * + * - TC: IPv4 or IPv6 TOS, if present, otherwise 0. + * + * - BoS: 1. + * + * If the new label is the second or label MPLS label in 'flow', it is + * generated as; + * + * - label: Copied from outer label. + * + * - TTL: Copied from outer label. + * + * - TC: Copied from outer label. + * + * - BoS: 0. + * + * 'n' must be flow_count_mpls_labels(flow). 'n' must be less than + * FLOW_MAX_MPLS_LABELS (because otherwise flow->mpls_lse[] would overflow). + */ +void +flow_push_mpls(struct flow *flow, int n, ovs_be16 mpls_eth_type, + struct flow_wildcards *wc) +{ + ovs_assert(eth_type_mpls(mpls_eth_type)); + ovs_assert(n < FLOW_MAX_MPLS_LABELS); + + memset(wc->masks.mpls_lse, 0xff, sizeof wc->masks.mpls_lse); + if (n) { + int i; + + for (i = n; i >= 1; i--) { + flow->mpls_lse[i] = flow->mpls_lse[i - 1]; + } + flow->mpls_lse[0] = (flow->mpls_lse[1] + & htonl(~MPLS_BOS_MASK)); + } else { + int label = 0; /* IPv4 Explicit Null. */ + int tc = 0; + int ttl = 64; + + if (flow->dl_type == htons(ETH_TYPE_IPV6)) { + label = 2; + } + + if (is_ip_any(flow)) { + tc = (flow->nw_tos & IP_DSCP_MASK) >> 2; + wc->masks.nw_tos |= IP_DSCP_MASK; + + if (flow->nw_ttl) { + ttl = flow->nw_ttl; + } + wc->masks.nw_ttl = 0xff; + } + + flow->mpls_lse[0] = set_mpls_lse_values(ttl, tc, 1, htonl(label)); + + /* Clear all L3 and L4 fields. */ + BUILD_ASSERT(FLOW_WC_SEQ == 25); + memset((char *) flow + FLOW_SEGMENT_2_ENDS_AT, 0, + sizeof(struct flow) - FLOW_SEGMENT_2_ENDS_AT); + } + flow->dl_type = mpls_eth_type; +} + +/* Tries to remove the outermost MPLS label from 'flow'. Returns true if + * successful, false otherwise. On success, sets 'flow''s Ethernet type to + * 'eth_type'. + * + * 'n' must be flow_count_mpls_labels(flow). */ +bool +flow_pop_mpls(struct flow *flow, int n, ovs_be16 eth_type, + struct flow_wildcards *wc) +{ + int i; + + if (n == 0) { + /* Nothing to pop. */ + return false; + } else if (n == FLOW_MAX_MPLS_LABELS + && !(flow->mpls_lse[n - 1] & htonl(MPLS_BOS_MASK))) { + /* Can't pop because we don't know what to fill in mpls_lse[n - 1]. */ + return false; + } + + memset(wc->masks.mpls_lse, 0xff, sizeof wc->masks.mpls_lse); + for (i = 1; i < n; i++) { + flow->mpls_lse[i - 1] = flow->mpls_lse[i]; + } + flow->mpls_lse[n - 1] = 0; + flow->dl_type = eth_type; + return true; +} + /* Sets the MPLS Label that 'flow' matches to 'label', which is interpreted * as an OpenFlow 1.1 "mpls_label" value. */ void -flow_set_mpls_label(struct flow *flow, ovs_be32 label) +flow_set_mpls_label(struct flow *flow, int idx, ovs_be32 label) { - set_mpls_lse_label(&flow->mpls_lse, label); + set_mpls_lse_label(&flow->mpls_lse[idx], label); } /* Sets the MPLS TTL that 'flow' matches to 'ttl', which should be in the * range 0...255. */ void -flow_set_mpls_ttl(struct flow *flow, uint8_t ttl) +flow_set_mpls_ttl(struct flow *flow, int idx, uint8_t ttl) { - set_mpls_lse_ttl(&flow->mpls_lse, ttl); + set_mpls_lse_ttl(&flow->mpls_lse[idx], ttl); } /* Sets the MPLS TC that 'flow' matches to 'tc', which should be in the * range 0...7. */ void -flow_set_mpls_tc(struct flow *flow, uint8_t tc) +flow_set_mpls_tc(struct flow *flow, int idx, uint8_t tc) { - set_mpls_lse_tc(&flow->mpls_lse, tc); + set_mpls_lse_tc(&flow->mpls_lse[idx], tc); } /* Sets the MPLS BOS bit that 'flow' matches to which should be 0 or 1. */ void -flow_set_mpls_bos(struct flow *flow, uint8_t bos) +flow_set_mpls_bos(struct flow *flow, int idx, uint8_t bos) +{ + set_mpls_lse_bos(&flow->mpls_lse[idx], bos); +} + +/* Sets the entire MPLS LSE. */ +void +flow_set_mpls_lse(struct flow *flow, int idx, ovs_be32 lse) { - set_mpls_lse_bos(&flow->mpls_lse, bos); + flow->mpls_lse[idx] = lse; +} + +static uint8_t * +flow_compose_l4(struct ofpbuf *b, const struct flow *flow) +{ + if (!(flow->nw_frag & FLOW_NW_FRAG_ANY) + || !(flow->nw_frag & FLOW_NW_FRAG_LATER)) { + if (flow->nw_proto == IPPROTO_TCP) { + struct tcp_header *tcp; + + tcp = ofpbuf_put_zeros(b, sizeof *tcp); + tcp->tcp_src = flow->tp_src; + tcp->tcp_dst = flow->tp_dst; + tcp->tcp_ctl = TCP_CTL(ntohs(flow->tcp_flags), 5); + return ofpbuf_tail(b); + } else if (flow->nw_proto == IPPROTO_UDP) { + struct udp_header *udp; + + udp = ofpbuf_put_zeros(b, sizeof *udp); + udp->udp_src = flow->tp_src; + udp->udp_dst = flow->tp_dst; + return ofpbuf_tail(b); + } else if (flow->nw_proto == IPPROTO_SCTP) { + struct sctp_header *sctp; + + sctp = ofpbuf_put_zeros(b, sizeof *sctp); + sctp->sctp_src = flow->tp_src; + sctp->sctp_dst = flow->tp_dst; + return ofpbuf_tail(b); + } else if (flow->nw_proto == IPPROTO_ICMP) { + struct icmp_header *icmp; + + icmp = ofpbuf_put_zeros(b, sizeof *icmp); + icmp->icmp_type = ntohs(flow->tp_src); + icmp->icmp_code = ntohs(flow->tp_dst); + icmp->icmp_csum = csum(icmp, ICMP_HEADER_LEN); + return ofpbuf_tail(b); + } else if (flow->nw_proto == IPPROTO_ICMPV6) { + struct icmp6_hdr *icmp; + + icmp = ofpbuf_put_zeros(b, sizeof *icmp); + icmp->icmp6_type = ntohs(flow->tp_src); + icmp->icmp6_code = ntohs(flow->tp_dst); + + if (icmp->icmp6_code == 0 && + (icmp->icmp6_type == ND_NEIGHBOR_SOLICIT || + icmp->icmp6_type == ND_NEIGHBOR_ADVERT)) { + struct in6_addr *nd_target; + struct nd_opt_hdr *nd_opt; + + nd_target = ofpbuf_put_zeros(b, sizeof *nd_target); + *nd_target = flow->nd_target; + + if (!eth_addr_is_zero(flow->arp_sha)) { + nd_opt = ofpbuf_put_zeros(b, 8); + nd_opt->nd_opt_len = 1; + nd_opt->nd_opt_type = ND_OPT_SOURCE_LINKADDR; + memcpy(nd_opt + 1, flow->arp_sha, ETH_ADDR_LEN); + } + if (!eth_addr_is_zero(flow->arp_tha)) { + nd_opt = ofpbuf_put_zeros(b, 8); + nd_opt->nd_opt_len = 1; + nd_opt->nd_opt_type = ND_OPT_TARGET_LINKADDR; + memcpy(nd_opt + 1, flow->arp_tha, ETH_ADDR_LEN); + } + } + icmp->icmp6_cksum = (OVS_FORCE uint16_t) + csum(icmp, (char *)ofpbuf_tail(b) - (char *)icmp); + return ofpbuf_tail(b); + } + } + return NULL; } /* Puts into 'b' a packet that flow_extract() would parse as having the given @@ -1009,6 +1318,7 @@ flow_set_mpls_bos(struct flow *flow, uint8_t bos) void flow_compose(struct ofpbuf *b, const struct flow *flow) { + /* eth_compose() sets l3 pointer and makes sure it is 32-bit aligned. */ eth_compose(b, flow->dl_dst, flow->dl_src, ntohs(flow->dl_type), 0); if (flow->dl_type == htons(FLOW_DL_TYPE_NONE)) { struct eth_header *eth = b->l2; @@ -1017,13 +1327,13 @@ flow_compose(struct ofpbuf *b, const struct flow *flow) } if (flow->vlan_tci & htons(VLAN_CFI)) { - eth_push_vlan(b, flow->vlan_tci); + eth_push_vlan(b, htons(ETH_TYPE_VLAN), flow->vlan_tci); } if (flow->dl_type == htons(ETH_TYPE_IP)) { struct ip_header *ip; - b->l3 = ip = ofpbuf_put_zeros(b, sizeof *ip); + ip = ofpbuf_put_zeros(b, sizeof *ip); ip->ip_ihl_ver = IP_IHL_VER(5, 4); ip->ip_tos = flow->nw_tos; ip->ip_ttl = flow->nw_ttl; @@ -1037,43 +1347,32 @@ flow_compose(struct ofpbuf *b, const struct flow *flow) ip->ip_frag_off |= htons(100); } } - if (!(flow->nw_frag & FLOW_NW_FRAG_ANY) - || !(flow->nw_frag & FLOW_NW_FRAG_LATER)) { - if (flow->nw_proto == IPPROTO_TCP) { - struct tcp_header *tcp; - - b->l4 = tcp = ofpbuf_put_zeros(b, sizeof *tcp); - tcp->tcp_src = flow->tp_src; - tcp->tcp_dst = flow->tp_dst; - tcp->tcp_ctl = TCP_CTL(ntohs(flow->tcp_flags), 5); - } else if (flow->nw_proto == IPPROTO_UDP) { - struct udp_header *udp; - - b->l4 = udp = ofpbuf_put_zeros(b, sizeof *udp); - udp->udp_src = flow->tp_src; - udp->udp_dst = flow->tp_dst; - } else if (flow->nw_proto == IPPROTO_SCTP) { - struct sctp_header *sctp; - - b->l4 = sctp = ofpbuf_put_zeros(b, sizeof *sctp); - sctp->sctp_src = flow->tp_src; - sctp->sctp_dst = flow->tp_dst; - } else if (flow->nw_proto == IPPROTO_ICMP) { - struct icmp_header *icmp; - - b->l4 = icmp = ofpbuf_put_zeros(b, sizeof *icmp); - icmp->icmp_type = ntohs(flow->tp_src); - icmp->icmp_code = ntohs(flow->tp_dst); - icmp->icmp_csum = csum(icmp, ICMP_HEADER_LEN); - } - } - ip = b->l3; + b->l4 = ofpbuf_tail(b); + + flow_compose_l4(b, flow); + ip->ip_tot_len = htons((uint8_t *) b->data + b->size - (uint8_t *) b->l3); ip->ip_csum = csum(ip, sizeof *ip); } else if (flow->dl_type == htons(ETH_TYPE_IPV6)) { - /* XXX */ + struct ovs_16aligned_ip6_hdr *nh; + uint8_t *l7; + + nh = ofpbuf_put_zeros(b, sizeof *nh); + put_16aligned_be32(&nh->ip6_flow, htonl(6 << 28) | + htonl(flow->nw_tos << 20) | flow->ipv6_label); + nh->ip6_hlim = flow->nw_ttl; + nh->ip6_nxt = flow->nw_proto; + + memcpy(&nh->ip6_src, &flow->ipv6_src, sizeof(nh->ip6_src)); + memcpy(&nh->ip6_dst, &flow->ipv6_dst, sizeof(nh->ip6_dst)); + + b->l4 = ofpbuf_tail(b); + + l7 = flow_compose_l4(b, flow); + + nh->ip6_plen = l7 ? htons(l7 - (uint8_t *) b->l4) : htons(0); } else if (flow->dl_type == htons(ETH_TYPE_ARP) || flow->dl_type == htons(ETH_TYPE_RARP)) { struct arp_eth_header *arp; @@ -1095,8 +1394,17 @@ flow_compose(struct ofpbuf *b, const struct flow *flow) } if (eth_type_mpls(flow->dl_type)) { + int n; + b->l2_5 = b->l3; - push_mpls(b, flow->dl_type, flow->mpls_lse); + for (n = 1; n < FLOW_MAX_MPLS_LABELS; n++) { + if (flow->mpls_lse[n - 1] & htonl(MPLS_BOS_MASK)) { + break; + } + } + while (n > 0) { + push_mpls(b, flow->dl_type, flow->mpls_lse[--n]); + } } } @@ -1105,13 +1413,7 @@ flow_compose(struct ofpbuf *b, const struct flow *flow) static int miniflow_n_values(const struct miniflow *flow) { - int n, i; - - n = 0; - for (i = 0; i < MINI_N_MAPS; i++) { - n += popcount(flow->map[i]); - } - return n; + return count_1bits(flow->map); } static uint32_t * @@ -1127,27 +1429,27 @@ miniflow_alloc_values(struct miniflow *flow, int n) /* Completes an initialization of 'dst' as a miniflow copy of 'src' begun by * the caller. The caller must have already initialized 'dst->map' properly - * to indicate the nonzero uint32_t elements of 'src'. 'n' must be the number - * of 1-bits in 'dst->map'. + * to indicate the significant uint32_t elements of 'src'. 'n' must be the + * number of 1-bits in 'dst->map'. + * + * Normally the significant elements are the ones that are non-zero. However, + * when a miniflow is initialized from a (mini)mask, the values can be zeroes, + * so that the flow and mask always have the same maps. * * This function initializes 'dst->values' (either inline if possible or with - * malloc() otherwise) and copies the nonzero uint32_t elements of 'src' into - * it. */ + * malloc() otherwise) and copies the uint32_t elements of 'src' indicated by + * 'dst->map' into it. */ static void miniflow_init__(struct miniflow *dst, const struct flow *src, int n) { const uint32_t *src_u32 = (const uint32_t *) src; unsigned int ofs; - int i; + uint64_t map; dst->values = miniflow_alloc_values(dst, n); ofs = 0; - for (i = 0; i < MINI_N_MAPS; i++) { - uint32_t map; - - for (map = dst->map[i]; map; map = zero_rightmost_1bit(map)) { - dst->values[ofs++] = src_u32[raw_ctz(map) + i * 32]; - } + for (map = dst->map; map; map = zero_rightmost_1bit(map)) { + dst->values[ofs++] = src_u32[raw_ctz(map)]; } } @@ -1162,10 +1464,11 @@ miniflow_init(struct miniflow *dst, const struct flow *src) /* Initialize dst->map, counting the number of nonzero elements. */ n = 0; - memset(dst->map, 0, sizeof dst->map); + dst->map = 0; + for (i = 0; i < FLOW_U32S; i++) { if (src_u32[i]) { - dst->map[i / 32] |= 1u << (i % 32); + dst->map |= UINT64_C(1) << i; n++; } } @@ -1179,7 +1482,7 @@ void miniflow_init_with_minimask(struct miniflow *dst, const struct flow *src, const struct minimask *mask) { - memcpy(dst->map, mask->masks.map, sizeof dst->map); + dst->map = mask->masks.map; miniflow_init__(dst, src, miniflow_n_values(dst)); } @@ -1189,7 +1492,7 @@ void miniflow_clone(struct miniflow *dst, const struct miniflow *src) { int n = miniflow_n_values(src); - memcpy(dst->map, src->map, sizeof dst->map); + dst->map = src->map; dst->values = miniflow_alloc_values(dst, n); memcpy(dst->values, src->values, n * sizeof *dst->values); } @@ -1206,7 +1509,7 @@ miniflow_move(struct miniflow *dst, struct miniflow *src) } else { dst->values = src->values; } - memcpy(dst->map, src->map, sizeof dst->map); + dst->map = src->map; } /* Frees any memory owned by 'flow'. Does not free the storage in which 'flow' @@ -1230,21 +1533,12 @@ miniflow_expand(const struct miniflow *src, struct flow *dst) static const uint32_t * miniflow_get__(const struct miniflow *flow, unsigned int u32_ofs) { - if (!(flow->map[u32_ofs / 32] & (1u << (u32_ofs % 32)))) { + if (!(flow->map & (UINT64_C(1) << u32_ofs))) { static const uint32_t zero = 0; return &zero; - } else { - const uint32_t *p = flow->values; - - BUILD_ASSERT(MINI_N_MAPS == 2); - if (u32_ofs < 32) { - p += popcount(flow->map[0] & ((1u << u32_ofs) - 1)); - } else { - p += popcount(flow->map[0]); - p += popcount(flow->map[1] & ((1u << (u32_ofs - 32)) - 1)); - } - return p; } + return flow->values + + count_1bits(flow->map & ((UINT64_C(1) << u32_ofs) - 1)); } /* Returns the uint32_t that would be at byte offset '4 * u32_ofs' if 'flow' @@ -1280,28 +1574,24 @@ miniflow_equal(const struct miniflow *a, const struct miniflow *b) { const uint32_t *ap = a->values; const uint32_t *bp = b->values; - int i; + const uint64_t a_map = a->map; + const uint64_t b_map = b->map; + uint64_t map; - for (i = 0; i < MINI_N_MAPS; i++) { - const uint32_t a_map = a->map[i]; - const uint32_t b_map = b->map[i]; - uint32_t map; - - if (a_map == b_map) { - for (map = a_map; map; map = zero_rightmost_1bit(map)) { - if (*ap++ != *bp++) { - return false; - } + if (a_map == b_map) { + for (map = a_map; map; map = zero_rightmost_1bit(map)) { + if (*ap++ != *bp++) { + return false; } - } else { - for (map = a_map | b_map; map; map = zero_rightmost_1bit(map)) { - uint32_t bit = rightmost_1bit(map); - uint32_t a_value = a_map & bit ? *ap++ : 0; - uint32_t b_value = b_map & bit ? *bp++ : 0; + } + } else { + for (map = a_map | b_map; map; map = zero_rightmost_1bit(map)) { + uint64_t bit = rightmost_1bit(map); + uint64_t a_value = a_map & bit ? *ap++ : 0; + uint64_t b_value = b_map & bit ? *bp++ : 0; - if (a_value != b_value) { - return false; - } + if (a_value != b_value) { + return false; } } } @@ -1316,20 +1606,17 @@ miniflow_equal_in_minimask(const struct miniflow *a, const struct miniflow *b, const struct minimask *mask) { const uint32_t *p; - int i; + uint64_t map; p = mask->masks.values; - for (i = 0; i < MINI_N_MAPS; i++) { - uint32_t map; - for (map = mask->masks.map[i]; map; map = zero_rightmost_1bit(map)) { - int ofs = raw_ctz(map) + i * 32; + for (map = mask->masks.map; map; map = zero_rightmost_1bit(map)) { + int ofs = raw_ctz(map); - if ((miniflow_get(a, ofs) ^ miniflow_get(b, ofs)) & *p) { - return false; - } - p++; + if ((miniflow_get(a, ofs) ^ miniflow_get(b, ofs)) & *p) { + return false; } + p++; } return true; @@ -1343,20 +1630,17 @@ miniflow_equal_flow_in_minimask(const struct miniflow *a, const struct flow *b, { const uint32_t *b_u32 = (const uint32_t *) b; const uint32_t *p; - int i; + uint64_t map; p = mask->masks.values; - for (i = 0; i < MINI_N_MAPS; i++) { - uint32_t map; - for (map = mask->masks.map[i]; map; map = zero_rightmost_1bit(map)) { - int ofs = raw_ctz(map) + i * 32; + for (map = mask->masks.map; map; map = zero_rightmost_1bit(map)) { + int ofs = raw_ctz(map); - if ((miniflow_get(a, ofs) ^ b_u32[ofs]) & *p) { - return false; - } - p++; + if ((miniflow_get(a, ofs) ^ b_u32[ofs]) & *p) { + return false; } + p++; } return true; @@ -1368,21 +1652,19 @@ miniflow_hash(const struct miniflow *flow, uint32_t basis) { const uint32_t *p = flow->values; uint32_t hash = basis; - int i; - - for (i = 0; i < MINI_N_MAPS; i++) { - uint32_t hash_map = 0; - uint32_t map; + uint64_t hash_map = 0; + uint64_t map; - for (map = flow->map[i]; map; map = zero_rightmost_1bit(map)) { - if (*p) { - hash = mhash_add(hash, *p); - hash_map |= rightmost_1bit(map); - } - p++; + for (map = flow->map; map; map = zero_rightmost_1bit(map)) { + if (*p) { + hash = mhash_add(hash, *p); + hash_map |= rightmost_1bit(map); } - hash = mhash_add(hash, hash_map); + p++; } + hash = mhash_add(hash, hash_map); + hash = mhash_add(hash, hash_map >> 32); + return mhash_finish(hash, p - flow->values); } @@ -1397,19 +1679,12 @@ miniflow_hash_in_minimask(const struct miniflow *flow, { const uint32_t *p = mask->masks.values; uint32_t hash; - int i; + uint64_t map; hash = basis; - for (i = 0; i < MINI_N_MAPS; i++) { - uint32_t map; - for (map = mask->masks.map[i]; map; map = zero_rightmost_1bit(map)) { - if (*p) { - int ofs = raw_ctz(map) + i * 32; - hash = mhash_add(hash, miniflow_get(flow, ofs) & *p); - } - p++; - } + for (map = mask->masks.map; map; map = zero_rightmost_1bit(map)) { + hash = mhash_add(hash, miniflow_get(flow, raw_ctz(map)) & *p++); } return mhash_finish(hash, (p - mask->masks.values) * 4); @@ -1424,27 +1699,44 @@ uint32_t flow_hash_in_minimask(const struct flow *flow, const struct minimask *mask, uint32_t basis) { - const uint32_t *flow_u32; + const uint32_t *flow_u32 = (const uint32_t *)flow; const uint32_t *p = mask->masks.values; uint32_t hash; - int i; + uint64_t map; hash = basis; - flow_u32 = (const uint32_t *) flow; - for (i = 0; i < MINI_N_MAPS; i++) { - uint32_t map; + for (map = mask->masks.map; map; map = zero_rightmost_1bit(map)) { + hash = mhash_add(hash, flow_u32[raw_ctz(map)] & *p++); + } - for (map = mask->masks.map[i]; map; map = zero_rightmost_1bit(map)) { - if (*p) { - hash = mhash_add(hash, flow_u32[raw_ctz(map)] & *p); - } - p++; - } - flow_u32 += 32; + return mhash_finish(hash, (p - mask->masks.values) * 4); +} + +/* Returns a hash value for the bits of range [start, end) in 'flow', + * where there are 1-bits in 'mask', given 'hash'. + * + * The hash values returned by this function are the same as those returned by + * minimatch_hash_range(), only the form of the arguments differ. */ +uint32_t +flow_hash_in_minimask_range(const struct flow *flow, + const struct minimask *mask, + uint8_t start, uint8_t end, uint32_t *basis) +{ + const uint32_t *flow_u32 = (const uint32_t *)flow; + unsigned int offset; + uint64_t map = miniflow_get_map_in_range(&mask->masks, start, end, + &offset); + const uint32_t *p = mask->masks.values + offset; + uint32_t hash = *basis; + + for (; map; map = zero_rightmost_1bit(map)) { + hash = mhash_add(hash, flow_u32[raw_ctz(map)] & *p++); } + *basis = hash; /* Allow continuation from the unfinished value. */ return mhash_finish(hash, (p - mask->masks.values) * 4); } + /* Initializes 'dst' as a copy of 'src'. The caller must eventually free 'dst' * with minimask_destroy(). */ @@ -1482,23 +1774,19 @@ minimask_combine(struct minimask *dst_, struct miniflow *dst = &dst_->masks; const struct miniflow *a = &a_->masks; const struct miniflow *b = &b_->masks; - int i, n; + uint64_t map; + int n = 0; - n = 0; dst->values = storage; - for (i = 0; i < MINI_N_MAPS; i++) { - uint32_t map; - - dst->map[i] = 0; - for (map = a->map[i] & b->map[i]; map; - map = zero_rightmost_1bit(map)) { - int ofs = raw_ctz(map) + i * 32; - uint32_t mask = miniflow_get(a, ofs) & miniflow_get(b, ofs); - - if (mask) { - dst->map[i] |= rightmost_1bit(map); - dst->values[n++] = mask; - } + + dst->map = 0; + for (map = a->map & b->map; map; map = zero_rightmost_1bit(map)) { + int ofs = raw_ctz(map); + uint32_t mask = miniflow_get(a, ofs) & miniflow_get(b, ofs); + + if (mask) { + dst->map |= rightmost_1bit(map); + dst->values[n++] = mask; } } } @@ -1555,20 +1843,15 @@ minimask_has_extra(const struct minimask *a_, const struct minimask *b_) { const struct miniflow *a = &a_->masks; const struct miniflow *b = &b_->masks; - int i; - - for (i = 0; i < MINI_N_MAPS; i++) { - uint32_t map; + uint64_t map; - for (map = a->map[i] | b->map[i]; map; - map = zero_rightmost_1bit(map)) { - int ofs = raw_ctz(map) + i * 32; - uint32_t a_u32 = miniflow_get(a, ofs); - uint32_t b_u32 = miniflow_get(b, ofs); + for (map = a->map | b->map; map; map = zero_rightmost_1bit(map)) { + int ofs = raw_ctz(map); + uint32_t a_u32 = miniflow_get(a, ofs); + uint32_t b_u32 = miniflow_get(b, ofs); - if ((a_u32 & b_u32) != b_u32) { - return true; - } + if ((a_u32 & b_u32) != b_u32) { + return true; } } @@ -1582,15 +1865,11 @@ minimask_is_catchall(const struct minimask *mask_) { const struct miniflow *mask = &mask_->masks; const uint32_t *p = mask->values; - int i; - - for (i = 0; i < MINI_N_MAPS; i++) { - uint32_t map; + uint64_t map; - for (map = mask->map[i]; map; map = zero_rightmost_1bit(map)) { - if (*p++) { - return false; - } + for (map = mask->map; map; map = zero_rightmost_1bit(map)) { + if (*p++) { + return false; } } return true;