X-Git-Url: http://git.onelab.eu/?a=blobdiff_plain;f=lib%2Fflow.c;h=6c6978d6bdcb2758749421a4af53feacaffa2139;hb=8d25251929c8f325bed0fff24192d5a87034b32e;hp=a42fea1e3d0f3113b0959b6f80cae9e25312642a;hpb=7431e17196fdb1c3189d67e3aeed4adeab4cf479;p=sliver-openvswitch.git diff --git a/lib/flow.c b/lib/flow.c index a42fea1e3..6c6978d6b 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. @@ -35,14 +35,21 @@ #include "ofpbuf.h" #include "openflow/openflow.h" #include "packets.h" +#include "odp-util.h" +#include "random.h" #include "unaligned.h" -#include "vlog.h" - -VLOG_DEFINE_THIS_MODULE(flow); 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) { @@ -52,35 +59,16 @@ pull_arp(struct ofpbuf *packet) static struct ip_header * pull_ip(struct ofpbuf *packet) { - if (packet->size >= IP_HEADER_LEN) { - struct ip_header *ip = packet->data; + if (ofpbuf_size(packet) >= IP_HEADER_LEN) { + struct ip_header *ip = ofpbuf_data(packet); int ip_len = IP_IHL(ip->ip_ihl_ver) * 4; - if (ip_len >= IP_HEADER_LEN && packet->size >= ip_len) { + if (ip_len >= IP_HEADER_LEN && ofpbuf_size(packet) >= ip_len) { return ofpbuf_pull(packet, ip_len); } } 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 icmp_header * pull_icmp(struct ofpbuf *packet) { @@ -97,12 +85,14 @@ static void parse_mpls(struct ofpbuf *b, struct flow *flow) { struct mpls_hdr *mh; + int idx = 0; while ((mh = ofpbuf_try_pull(b, sizeof *mh))) { - if (flow->mpls_depth++ == 0) { - flow->mpls_lse = mh->mpls_lse; + ovs_be32 mpls_lse = get_16aligned_be32(&mh->mpls_lse); + if (idx < FLOW_MAX_MPLS_LABELS) { + flow->mpls_lse[idx++] = mpls_lse; } - if (mh->mpls_lse & htonl(MPLS_BOS_MASK)) { + if (mpls_lse & htonl(MPLS_BOS_MASK)) { break; } } @@ -116,7 +106,7 @@ parse_vlan(struct ofpbuf *b, struct flow *flow) ovs_be16 tci; }; - if (b->size >= sizeof(struct qtag_prefix) + sizeof(ovs_be16)) { + if (ofpbuf_size(b) >= sizeof(struct qtag_prefix) + sizeof(ovs_be16)) { struct qtag_prefix *qp = ofpbuf_pull(b, sizeof *qp); flow->vlan_tci = qp->tci | htons(VLAN_CFI); } @@ -133,11 +123,11 @@ parse_ethertype(struct ofpbuf *b) return proto; } - if (b->size < sizeof *llc) { + if (ofpbuf_size(b) < sizeof *llc) { return htons(FLOW_DL_TYPE_NONE); } - llc = b->data; + llc = ofpbuf_data(b); if (llc->llc.llc_dsap != LLC_DSAP_SNAP || llc->llc.llc_ssap != LLC_SSAP_SNAP || llc->llc.llc_cntl != LLC_CNTL_SNAP @@ -158,7 +148,7 @@ parse_ethertype(struct ofpbuf *b) static int parse_ipv6(struct ofpbuf *packet, struct flow *flow) { - const struct ip6_hdr *nh; + const struct ovs_16aligned_ip6_hdr *nh; ovs_be32 tc_flow; int nexthdr; @@ -169,10 +159,10 @@ parse_ipv6(struct ofpbuf *packet, struct flow *flow) nexthdr = nh->ip6_nxt; - flow->ipv6_src = nh->ip6_src; - flow->ipv6_dst = nh->ip6_dst; + memcpy(&flow->ipv6_src, &nh->ip6_src, sizeof flow->ipv6_src); + memcpy(&flow->ipv6_dst, &nh->ip6_dst, sizeof flow->ipv6_dst); - tc_flow = get_unaligned_be32(&nh->ip6_flow); + tc_flow = get_16aligned_be32(&nh->ip6_flow); flow->nw_tos = ntohl(tc_flow) >> 20; flow->ipv6_label = tc_flow & htonl(IPV6_LABEL_MASK); flow->nw_ttl = nh->ip6_hlim; @@ -195,7 +185,7 @@ parse_ipv6(struct ofpbuf *packet, struct flow *flow) * accesses within the extension header are within those first 8 * bytes. All extension headers are required to be at least 8 * bytes. */ - if (packet->size < 8) { + if (ofpbuf_size(packet) < 8) { return EINVAL; } @@ -204,7 +194,7 @@ parse_ipv6(struct ofpbuf *packet, struct flow *flow) || (nexthdr == IPPROTO_DSTOPTS)) { /* These headers, while different, have the fields we care about * in the same location and with the same interpretation. */ - const struct ip6_ext *ext_hdr = packet->data; + const struct ip6_ext *ext_hdr = ofpbuf_data(packet); nexthdr = ext_hdr->ip6e_nxt; if (!ofpbuf_try_pull(packet, (ext_hdr->ip6e_len + 1) * 8)) { return EINVAL; @@ -214,13 +204,13 @@ parse_ipv6(struct ofpbuf *packet, struct flow *flow) * we care about are in the same location as the generic * option header--only the header length is calculated * differently. */ - const struct ip6_ext *ext_hdr = packet->data; + const struct ip6_ext *ext_hdr = ofpbuf_data(packet); nexthdr = ext_hdr->ip6e_nxt; if (!ofpbuf_try_pull(packet, (ext_hdr->ip6e_len + 2) * 4)) { return EINVAL; } } else if (nexthdr == IPPROTO_FRAGMENT) { - const struct ip6_frag *frag_hdr = packet->data; + const struct ovs_16aligned_ip6_frag *frag_hdr = ofpbuf_data(packet); nexthdr = frag_hdr->ip6f_nxt; if (!ofpbuf_try_pull(packet, sizeof *frag_hdr)) { @@ -244,34 +234,46 @@ 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 (ofpbuf_size(b) >= TCP_HEADER_LEN) { + const struct tcp_header *tcp = ofpbuf_data(b); + flow->tp_src = tcp->tcp_src; flow->tp_dst = tcp->tcp_dst; - packet->l7 = b->data; + flow->tcp_flags = tcp->tcp_ctl & htons(0x0fff); } } 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 (ofpbuf_size(b) >= UDP_HEADER_LEN) { + const struct udp_header *udp = ofpbuf_data(b); + flow->tp_src = udp->udp_src; flow->tp_dst = udp->udp_dst; - packet->l7 = b->data; } } -static bool +static void +parse_sctp(struct ofpbuf *b, struct flow *flow) +{ + if (ofpbuf_size(b) >= SCTP_HEADER_LEN) { + const struct sctp_header *sctp = ofpbuf_data(b); + + flow->tp_src = sctp->sctp_src; + flow->tp_dst = sctp->sctp_dst; + } +} + +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 @@ -286,17 +288,17 @@ 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; - while (b->size >= 8) { + while (ofpbuf_size(b) >= 8) { /* The minimum size of an option is 8 bytes, which also is * the size of Ethernet link-layer options. */ - const struct nd_opt_hdr *nd_opt = b->data; + const struct nd_opt_hdr *nd_opt = ofpbuf_data(b); int opt_len = nd_opt->nd_opt_len * 8; - if (!opt_len || opt_len > b->size) { + if (!opt_len || opt_len > ofpbuf_size(b)) { goto invalid; } @@ -325,39 +327,35 @@ 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 'packet' header pointers as follows: +/* Initializes 'flow' members from 'packet' and 'md' * - * - packet->l2 to the start of the Ethernet header. + * Initializes 'packet' header l2 pointer to the start of the Ethernet + * header, and the layer offsets as follows: * - * - packet->l2_5 to the start of the MPLS shim header. + * - packet->l2_5_ofs to the start of the MPLS shim header, or UINT16_MAX + * when there is no MPLS shim header. * - * - packet->l3 to just past the Ethernet header, or just past the + * - packet->l3_ofs to just past the Ethernet header, or just past the * vlan_header if one is present, to the first byte of the payload of the - * Ethernet frame. + * Ethernet frame. UINT16_MAX if the frame is too short to contain an + * Ethernet header. * - * - 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 or UDP or ICMP header, if one is - * present and has a correct length, and otherwise NULL. + * - packet->l4_ofs to just past the IPv4 header, if one is present and + * has at least the content used for the fields of interest for the flow, + * otherwise UINT16_MAX. */ void -flow_extract(struct ofpbuf *packet, uint32_t skb_priority, uint32_t skb_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; @@ -367,28 +365,23 @@ flow_extract(struct ofpbuf *packet, uint32_t skb_priority, uint32_t skb_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->recirc_id = md->recirc_id; + flow->dp_hash = md->dp_hash; } - flow->skb_priority = skb_priority; - flow->skb_mark = skb_mark; - packet->l2 = b.data; - packet->l2_5 = NULL; - packet->l3 = NULL; - packet->l4 = NULL; - packet->l7 = NULL; + ofpbuf_set_frame(packet, ofpbuf_data(packet)); - if (b.size < sizeof *eth) { + if (ofpbuf_size(&b) < sizeof *eth) { return; } /* Link layer. */ - eth = b.data; + eth = ofpbuf_data(&b); memcpy(flow->dl_src, eth->eth_src, ETH_ADDR_LEN); memcpy(flow->dl_dst, eth->eth_dst, ETH_ADDR_LEN); @@ -401,19 +394,19 @@ flow_extract(struct ofpbuf *packet, uint32_t skb_priority, uint32_t skb_mark, /* Parse mpls, copy l3 ttl. */ if (eth_type_mpls(flow->dl_type)) { - packet->l2_5 = b.data; + ofpbuf_set_l2_5(packet, ofpbuf_data(&b)); parse_mpls(&b, flow); } /* Network layer. */ - packet->l3 = b.data; + ofpbuf_set_l3(packet, ofpbuf_data(&b)); if (flow->dl_type == htons(ETH_TYPE_IP)) { const struct ip_header *nh = pull_ip(&b); if (nh) { - packet->l4 = b.data; + ofpbuf_set_l4(packet, ofpbuf_data(&b)); - flow->nw_src = get_unaligned_be32(&nh->ip_src); - flow->nw_dst = get_unaligned_be32(&nh->ip_dst); + flow->nw_src = get_16aligned_be32(&nh->ip_src); + flow->nw_dst = get_16aligned_be32(&nh->ip_dst); flow->nw_proto = nh->ip_proto; flow->nw_tos = nh->ip_tos; @@ -427,15 +420,16 @@ flow_extract(struct ofpbuf *packet, uint32_t skb_priority, uint32_t skb_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(&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; } } } @@ -445,15 +439,15 @@ flow_extract(struct ofpbuf *packet, uint32_t skb_priority, uint32_t skb_mark, return; } - packet->l4 = b.data; + ofpbuf_set_l4(packet, ofpbuf_data(&b)); 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(&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)) { @@ -467,8 +461,8 @@ flow_extract(struct ofpbuf *packet, uint32_t skb_priority, uint32_t skb_mark, flow->nw_proto = ntohs(arp->ar_op); } - flow->nw_src = arp->ar_spa; - flow->nw_dst = arp->ar_tpa; + flow->nw_src = get_16aligned_be32(&arp->ar_spa); + flow->nw_dst = get_16aligned_be32(&arp->ar_tpa); memcpy(flow->arp_sha, arp->ar_sha, ETH_ADDR_LEN); memcpy(flow->arp_tha, arp->ar_tha, ETH_ADDR_LEN); } @@ -489,17 +483,32 @@ 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 == 20); + 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; fmd->metadata = flow->metadata; memcpy(fmd->regs, flow->regs, sizeof fmd->regs); + fmd->pkt_mark = flow->pkt_mark; fmd->in_port = flow->in_port.ofp_port; } @@ -555,6 +564,24 @@ 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) { @@ -581,13 +608,13 @@ flow_wildcards_init_catchall(struct flow_wildcards *wc) memset(&wc->masks, 0, sizeof wc->masks); } -/* Initializes 'wc' as an exact-match set of wildcards; that is, 'wc' does not - * wildcard any bits or fields. */ +/* Clear the metadata and register wildcard masks. They are not packet + * header fields. */ void -flow_wildcards_init_exact(struct flow_wildcards *wc) +flow_wildcards_clear_non_packet_fields(struct flow_wildcards *wc) { - memset(&wc->masks, 0xff, sizeof wc->masks); - memset(wc->masks.zeros, 0, sizeof wc->masks.zeros); + 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 @@ -648,16 +675,11 @@ static void flow_union_with_miniflow(struct flow *dst, const struct miniflow *src) { uint32_t *dst_u32 = (uint32_t *) dst; - int ofs; - int i; + const uint32_t *p = src->values; + uint64_t map; - ofs = 0; - for (i = 0; i < MINI_N_MAPS; i++) { - uint32_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++; } } @@ -669,6 +691,43 @@ 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) @@ -730,6 +789,25 @@ 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) @@ -760,7 +838,7 @@ flow_hash_symmetric_l4(const struct flow *flow, uint32_t basis) if (fields.eth_type == htons(ETH_TYPE_IP)) { fields.ipv4_addr = flow->nw_src ^ flow->nw_dst; fields.ip_proto = flow->nw_proto; - if (fields.ip_proto == IPPROTO_TCP) { + if (fields.ip_proto == IPPROTO_TCP || fields.ip_proto == IPPROTO_SCTP) { fields.tp_port = flow->tp_src ^ flow->tp_dst; } } else if (fields.eth_type == htons(ETH_TYPE_IPV6)) { @@ -772,16 +850,57 @@ flow_hash_symmetric_l4(const struct flow *flow, uint32_t basis) ipv6_addr[i] = a[i] ^ b[i]; } fields.ip_proto = flow->nw_proto; - if (fields.ip_proto == IPPROTO_TCP) { + if (fields.ip_proto == IPPROTO_TCP || fields.ip_proto == IPPROTO_SCTP) { fields.tp_port = flow->tp_src ^ flow->tp_dst; } } return jhash_bytes(&fields, sizeof fields, basis); } +/* Initialize a flow with random fields that matter for nx_hash_fields. */ +void +flow_random_hash_fields(struct flow *flow) +{ + uint16_t rnd = random_uint16(); + + /* Initialize to all zeros. */ + memset(flow, 0, sizeof *flow); + + eth_addr_random(flow->dl_src); + eth_addr_random(flow->dl_dst); + + flow->vlan_tci = (OVS_FORCE ovs_be16) (random_uint16() & VLAN_VID_MASK); + + /* Make most of the random flows IPv4, some IPv6, and rest random. */ + flow->dl_type = rnd < 0x8000 ? htons(ETH_TYPE_IP) : + rnd < 0xc000 ? htons(ETH_TYPE_IPV6) : (OVS_FORCE ovs_be16)rnd; + + if (dl_type_is_ip_any(flow->dl_type)) { + if (flow->dl_type == htons(ETH_TYPE_IP)) { + flow->nw_src = (OVS_FORCE ovs_be32)random_uint32(); + flow->nw_dst = (OVS_FORCE ovs_be32)random_uint32(); + } else { + random_bytes(&flow->ipv6_src, sizeof flow->ipv6_src); + random_bytes(&flow->ipv6_dst, sizeof flow->ipv6_dst); + } + /* Make most of IP flows TCP, some UDP or SCTP, and rest random. */ + rnd = random_uint16(); + flow->nw_proto = rnd < 0x8000 ? IPPROTO_TCP : + rnd < 0xc000 ? IPPROTO_UDP : + rnd < 0xd000 ? IPPROTO_SCTP : (uint8_t)rnd; + if (flow->nw_proto == IPPROTO_TCP || + flow->nw_proto == IPPROTO_UDP || + flow->nw_proto == IPPROTO_SCTP) { + flow->tp_src = (OVS_FORCE ovs_be16)random_uint16(); + flow->tp_dst = (OVS_FORCE ovs_be16)random_uint16(); + } + } +} + /* Masks the fields in 'wc' that are used by the flow hash 'fields'. */ void -flow_mask_hash_fields(struct flow_wildcards *wc, enum nx_hash_fields fields) +flow_mask_hash_fields(const struct flow *flow, struct flow_wildcards *wc, + enum nx_hash_fields fields) { switch (fields) { case NX_HASH_FIELDS_ETH_SRC: @@ -791,16 +910,22 @@ flow_mask_hash_fields(struct flow_wildcards *wc, enum nx_hash_fields fields) case NX_HASH_FIELDS_SYMMETRIC_L4: memset(&wc->masks.dl_src, 0xff, sizeof wc->masks.dl_src); memset(&wc->masks.dl_dst, 0xff, sizeof wc->masks.dl_dst); - memset(&wc->masks.nw_proto, 0xff, sizeof wc->masks.nw_proto); - memset(&wc->masks.nw_src, 0xff, sizeof wc->masks.nw_src); - memset(&wc->masks.nw_dst, 0xff, sizeof wc->masks.nw_dst); - memset(&wc->masks.tp_src, 0xff, sizeof wc->masks.tp_src); - memset(&wc->masks.tp_dst, 0xff, sizeof wc->masks.tp_dst); + if (flow->dl_type == htons(ETH_TYPE_IP)) { + memset(&wc->masks.nw_src, 0xff, sizeof wc->masks.nw_src); + memset(&wc->masks.nw_dst, 0xff, sizeof wc->masks.nw_dst); + } else if (flow->dl_type == htons(ETH_TYPE_IPV6)) { + memset(&wc->masks.ipv6_src, 0xff, sizeof wc->masks.ipv6_src); + memset(&wc->masks.ipv6_dst, 0xff, sizeof wc->masks.ipv6_dst); + } + if (is_ip_any(flow)) { + memset(&wc->masks.nw_proto, 0xff, sizeof wc->masks.nw_proto); + flow_unwildcard_tp_ports(flow, wc); + } wc->masks.vlan_tci |= htons(VLAN_VID_MASK | VLAN_CFI); break; default: - NOT_REACHED(); + OVS_NOT_REACHED(); } } @@ -818,7 +943,7 @@ 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'. */ @@ -907,35 +1032,288 @@ 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 size_t +flow_compose_l4(struct ofpbuf *b, const struct flow *flow) +{ + size_t l4_len = 0; + + 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; + + l4_len = sizeof *tcp; + tcp = ofpbuf_put_zeros(b, l4_len); + 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; + + l4_len = sizeof *udp; + udp = ofpbuf_put_zeros(b, l4_len); + udp->udp_src = flow->tp_src; + udp->udp_dst = flow->tp_dst; + } else if (flow->nw_proto == IPPROTO_SCTP) { + struct sctp_header *sctp; + + l4_len = sizeof *sctp; + sctp = ofpbuf_put_zeros(b, l4_len); + sctp->sctp_src = flow->tp_src; + sctp->sctp_dst = flow->tp_dst; + } else if (flow->nw_proto == IPPROTO_ICMP) { + struct icmp_header *icmp; + + l4_len = sizeof *icmp; + icmp = ofpbuf_put_zeros(b, l4_len); + icmp->icmp_type = ntohs(flow->tp_src); + icmp->icmp_code = ntohs(flow->tp_dst); + icmp->icmp_csum = csum(icmp, ICMP_HEADER_LEN); + } else if (flow->nw_proto == IPPROTO_ICMPV6) { + struct icmp6_hdr *icmp; + + l4_len = sizeof *icmp; + icmp = ofpbuf_put_zeros(b, l4_len); + 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; + + l4_len += sizeof *nd_target; + nd_target = ofpbuf_put_zeros(b, sizeof *nd_target); + *nd_target = flow->nd_target; + + if (!eth_addr_is_zero(flow->arp_sha)) { + l4_len += 8; + 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)) { + l4_len += 8; + 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 l4_len; } /* Puts into 'b' a packet that flow_extract() would parse as having the given @@ -947,27 +1325,30 @@ flow_set_mpls_bos(struct flow *flow, uint8_t bos) void flow_compose(struct ofpbuf *b, const struct flow *flow) { + size_t l4_len; + + /* 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; - eth->eth_type = htons(b->size); + struct eth_header *eth = ofpbuf_l2(b); + eth->eth_type = htons(ofpbuf_size(b)); return; } 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; ip->ip_proto = flow->nw_proto; - ip->ip_src = flow->nw_src; - ip->ip_dst = flow->nw_dst; + put_16aligned_be32(&ip->ip_src, flow->nw_src); + put_16aligned_be32(&ip->ip_dst, flow->nw_dst); if (flow->nw_frag & FLOW_NW_FRAG_ANY) { ip->ip_frag_off |= htons(IP_MORE_FRAGMENTS); @@ -975,42 +1356,36 @@ 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(0, 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_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; - ip->ip_tot_len = htons((uint8_t *) b->data + b->size - - (uint8_t *) b->l3); + ofpbuf_set_l4(b, ofpbuf_tail(b)); + + l4_len = flow_compose_l4(b, flow); + + ip->ip_tot_len = htons(b->l4_ofs - b->l3_ofs + l4_len); ip->ip_csum = csum(ip, sizeof *ip); } else if (flow->dl_type == htons(ETH_TYPE_IPV6)) { - /* XXX */ + struct ovs_16aligned_ip6_hdr *nh; + + 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)); + + ofpbuf_set_l4(b, ofpbuf_tail(b)); + + l4_len = flow_compose_l4(b, flow); + + nh->ip6_plen = htons(l4_len); } else if (flow->dl_type == htons(ETH_TYPE_ARP) || flow->dl_type == htons(ETH_TYPE_RARP)) { struct arp_eth_header *arp; - b->l3 = arp = ofpbuf_put_zeros(b, sizeof *arp); + arp = ofpbuf_put_zeros(b, sizeof *arp); + ofpbuf_set_l3(b, arp); arp->ar_hrd = htons(1); arp->ar_pro = htons(ETH_TYPE_IP); arp->ar_hln = ETH_ADDR_LEN; @@ -1019,16 +1394,25 @@ flow_compose(struct ofpbuf *b, const struct flow *flow) if (flow->nw_proto == ARP_OP_REQUEST || flow->nw_proto == ARP_OP_REPLY) { - arp->ar_spa = flow->nw_src; - arp->ar_tpa = flow->nw_dst; + put_16aligned_be32(&arp->ar_spa, flow->nw_src); + put_16aligned_be32(&arp->ar_tpa, flow->nw_dst); memcpy(arp->ar_sha, flow->arp_sha, ETH_ADDR_LEN); memcpy(arp->ar_tha, flow->arp_tha, ETH_ADDR_LEN); } } if (eth_type_mpls(flow->dl_type)) { - b->l2_5 = b->l3; - push_mpls(b, flow->dl_type, flow->mpls_lse); + int n; + + b->l2_5_ofs = b->l3_ofs; + 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]); + } } } @@ -1037,13 +1421,7 @@ 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 * @@ -1057,36 +1435,63 @@ 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 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 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; + uint64_t map; + + dst->values = miniflow_alloc_values(dst, n); + ofs = 0; + for (map = dst->map; map; map = zero_rightmost_1bit(map)) { + dst->values[ofs++] = src_u32[raw_ctz(map)]; + } +} + /* Initializes 'dst' as a copy of 'src'. The caller must eventually free 'dst' * with miniflow_destroy(). */ void miniflow_init(struct miniflow *dst, const struct flow *src) { const uint32_t *src_u32 = (const uint32_t *) src; - unsigned int ofs; unsigned int i; int n; /* 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++; } } - /* Initialize dst->values. */ - dst->values = miniflow_alloc_values(dst, n); - ofs = 0; - for (i = 0; i < MINI_N_MAPS; i++) { - uint32_t map; + miniflow_init__(dst, src, n); +} - for (map = dst->map[i]; map; map = zero_rightmost_1bit(map)) { - dst->values[ofs++] = src_u32[raw_ctz(map) + i * 32]; - } - } +/* Initializes 'dst' as a copy of 'src', using 'mask->map' as 'dst''s map. The + * caller must eventually free 'dst' with miniflow_destroy(). */ +void +miniflow_init_with_minimask(struct miniflow *dst, const struct flow *src, + const struct minimask *mask) +{ + dst->map = mask->masks.map; + miniflow_init__(dst, src, miniflow_n_values(dst)); } /* Initializes 'dst' as a copy of 'src'. The caller must eventually free 'dst' @@ -1095,11 +1500,26 @@ 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); } +/* Initializes 'dst' with the data in 'src', destroying 'src'. + * The caller must eventually free 'dst' with miniflow_destroy(). */ +void +miniflow_move(struct miniflow *dst, struct miniflow *src) +{ + if (src->values == src->inline_values) { + dst->values = dst->inline_values; + memcpy(dst->values, src->values, + miniflow_n_values(src) * sizeof *dst->values); + } else { + dst->values = src->values; + } + dst->map = src->map; +} + /* Frees any memory owned by 'flow'. Does not free the storage in which 'flow' * itself resides; the caller is responsible for that. */ void @@ -1121,21 +1541,12 @@ 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' @@ -1169,16 +1580,31 @@ miniflow_get_vid(const struct miniflow *flow) bool miniflow_equal(const struct miniflow *a, const struct miniflow *b) { - int i; + const uint32_t *ap = a->values; + const uint32_t *bp = b->values; + 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++) { - if (a->map[i] != b->map[i]) { - 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)) { + 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; + } } } - return !memcmp(a->values, b->values, - miniflow_n_values(a) * sizeof *a->values); + return true; } /* Returns true if 'a' and 'b' are equal at the places where there are 1-bits @@ -1188,20 +1614,17 @@ 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; @@ -1215,20 +1638,17 @@ 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; @@ -1238,10 +1658,22 @@ miniflow_equal_flow_in_minimask(const struct miniflow *a, const struct flow *b, uint32_t miniflow_hash(const struct miniflow *flow, uint32_t basis) { - BUILD_ASSERT_DECL(MINI_N_MAPS == 2); - return hash_3words(flow->map[0], flow->map[1], - hash_words(flow->values, miniflow_n_values(flow), - basis)); + const uint32_t *p = flow->values; + uint32_t hash = basis; + uint64_t hash_map = 0; + uint64_t map; + + for (map = flow->map; map; map = zero_rightmost_1bit(map)) { + if (*p) { + hash = mhash_add(hash, *p); + hash_map |= rightmost_1bit(map); + } + p++; + } + hash = mhash_add(hash, hash_map); + hash = mhash_add(hash, hash_map >> 32); + + return mhash_finish(hash, p - flow->values); } /* Returns a hash value for the bits of 'flow' where there are 1-bits in @@ -1255,18 +1687,12 @@ 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)) { - 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); @@ -1281,25 +1707,44 @@ 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; + 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; - 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)) { - int ofs = raw_ctz(map) + i * 32; + return mhash_finish(hash, (p - mask->masks.values) * 4); +} - hash = mhash_add(hash, flow_u32[ofs] & *p); - p++; - } +/* 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(). */ @@ -1317,6 +1762,14 @@ minimask_clone(struct minimask *dst, const struct minimask *src) miniflow_clone(&dst->masks, &src->masks); } +/* Initializes 'dst' with the data in 'src', destroying 'src'. + * The caller must eventually free 'dst' with minimask_destroy(). */ +void +minimask_move(struct minimask *dst, struct minimask *src) +{ + miniflow_move(&dst->masks, &src->masks); +} + /* Initializes 'dst_' as the bit-wise "and" of 'a_' and 'b_'. * * The caller must provide room for FLOW_U32S "uint32_t"s in 'storage', for use @@ -1329,23 +1782,19 @@ 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; } } } @@ -1402,20 +1851,15 @@ 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; } } @@ -1428,7 +1872,13 @@ bool minimask_is_catchall(const struct minimask *mask_) { const struct miniflow *mask = &mask_->masks; + const uint32_t *p = mask->values; + uint64_t map; - BUILD_ASSERT(MINI_N_MAPS == 2); - return !(mask->map[0] | mask->map[1]); + for (map = mask->map; map; map = zero_rightmost_1bit(map)) { + if (*p++) { + return false; + } + } + return true; }