X-Git-Url: http://git.onelab.eu/?a=blobdiff_plain;f=lib%2Fflow.c;h=c6683a515c12342bed44f049ad292ca7bde15807;hb=2cd0c475838d048edaec57d137c49bc552c45a08;hp=93bd9b25bd13cfda63acf847922ea482e2cb040e;hpb=296e07ace0ff2d5819ce760d6018ba82d14aa2a8;p=sliver-openvswitch.git diff --git a/lib/flow.c b/lib/flow.c index 93bd9b25b..c6683a515 100644 --- a/lib/flow.c +++ b/lib/flow.c @@ -1,5 +1,5 @@ /* - * Copyright (c) 2008, 2009, 2010, 2011, 2012 Nicira, Inc. + * Copyright (c) 2008, 2009, 2010, 2011, 2012, 2013 Nicira, Inc. * * Licensed under the Apache License, Version 2.0 (the "License"); * you may not use this file except in compliance with the License. @@ -16,7 +16,6 @@ #include #include #include "flow.h" -#include #include #include #include @@ -31,17 +30,25 @@ #include "csum.h" #include "dynamic-string.h" #include "hash.h" +#include "jhash.h" +#include "match.h" #include "ofpbuf.h" #include "openflow/openflow.h" #include "packets.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) { @@ -80,6 +87,12 @@ 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) { @@ -92,6 +105,23 @@ pull_icmpv6(struct ofpbuf *packet) return ofpbuf_try_pull(packet, sizeof(struct icmp6_hdr)); } +static void +parse_mpls(struct ofpbuf *b, struct flow *flow) +{ + struct mpls_hdr *mh; + bool top = true; + + while ((mh = ofpbuf_try_pull(b, sizeof *mh))) { + if (top) { + top = false; + flow->mpls_lse = mh->mpls_lse; + } + if (mh->mpls_lse & htonl(MPLS_BOS_MASK)) { + break; + } + } +} + static void parse_vlan(struct ofpbuf *b, struct flow *flow) { @@ -131,13 +161,18 @@ parse_ethertype(struct ofpbuf *b) } ofpbuf_pull(b, sizeof *llc); - return llc->snap.snap_type; + + if (ntohs(llc->snap.snap_type) >= ETH_TYPE_MIN) { + return llc->snap.snap_type; + } + + return htons(FLOW_DL_TYPE_NONE); } 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; @@ -148,10 +183,10 @@ 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; @@ -199,7 +234,7 @@ parse_ipv6(struct ofpbuf *packet, struct flow *flow) return EINVAL; } } else if (nexthdr == IPPROTO_FRAGMENT) { - const struct ip6_frag *frag_hdr = packet->data; + const struct ovs_16aligned_ip6_frag *frag_hdr = packet->data; nexthdr = frag_hdr->ip6f_nxt; if (!ofpbuf_try_pull(packet, sizeof *frag_hdr)) { @@ -208,9 +243,8 @@ parse_ipv6(struct ofpbuf *packet, struct flow *flow) /* We only process the first fragment. */ if (frag_hdr->ip6f_offlg != htons(0)) { - if ((frag_hdr->ip6f_offlg & IP6F_OFF_MASK) == htons(0)) { - flow->nw_frag = FLOW_NW_FRAG_ANY; - } else { + flow->nw_frag = FLOW_NW_FRAG_ANY; + if ((frag_hdr->ip6f_offlg & IP6F_OFF_MASK) != htons(0)) { flow->nw_frag |= FLOW_NW_FRAG_LATER; nexthdr = IPPROTO_FRAGMENT; break; @@ -230,6 +264,7 @@ parse_tcp(struct ofpbuf *packet, struct ofpbuf *b, struct flow *flow) if (tcp) { flow->tp_src = tcp->tcp_src; flow->tp_dst = tcp->tcp_dst; + flow->tcp_flags = tcp->tcp_ctl & htons(0x0fff); packet->l7 = b->data; } } @@ -245,6 +280,17 @@ parse_udp(struct ofpbuf *packet, struct ofpbuf *b, struct flow *flow) } } +static void +parse_sctp(struct ofpbuf *packet, struct ofpbuf *b, struct flow *flow) +{ + const struct sctp_header *sctp = pull_sctp(b); + if (sctp) { + flow->tp_src = sctp->sctp_src; + flow->tp_dst = sctp->sctp_dst; + packet->l7 = b->data; + } +} + static bool parse_icmpv6(struct ofpbuf *b, struct flow *flow) { @@ -317,12 +363,14 @@ invalid: } /* Initializes 'flow' members from 'packet', 'skb_priority', 'tnl', and - * 'ofp_in_port'. + * 'in_port'. * * Initializes 'packet' header pointers as follows: * * - packet->l2 to the start of the Ethernet header. * + * - packet->l2_5 to the start of the MPLS shim header. + * * - packet->l3 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. @@ -330,12 +378,12 @@ 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 or UDP or ICMP header, if one is + * - 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, - const struct flow_tnl *tnl, uint16_t ofp_in_port, +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, struct flow *flow) { struct ofpbuf b = *packet; @@ -346,15 +394,20 @@ flow_extract(struct ofpbuf *packet, uint32_t skb_priority, memset(flow, 0, sizeof *flow); if (tnl) { + ovs_assert(tnl != &flow->tunnel); flow->tunnel = *tnl; } - flow->in_port = ofp_in_port; + if (in_port) { + flow->in_port = *in_port; + } flow->skb_priority = skb_priority; + flow->pkt_mark = pkt_mark; - packet->l2 = b.data; - packet->l3 = NULL; - packet->l4 = NULL; - packet->l7 = NULL; + packet->l2 = b.data; + packet->l2_5 = NULL; + packet->l3 = NULL; + packet->l4 = NULL; + packet->l7 = NULL; if (b.size < sizeof *eth) { return; @@ -372,6 +425,12 @@ flow_extract(struct ofpbuf *packet, uint32_t skb_priority, } flow->dl_type = parse_ethertype(&b); + /* Parse mpls, copy l3 ttl. */ + if (eth_type_mpls(flow->dl_type)) { + packet->l2_5 = b.data; + parse_mpls(&b, flow); + } + /* Network layer. */ packet->l3 = b.data; if (flow->dl_type == htons(ETH_TYPE_IP)) { @@ -379,8 +438,8 @@ flow_extract(struct ofpbuf *packet, uint32_t skb_priority, if (nh) { packet->l4 = b.data; - 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; @@ -397,6 +456,8 @@ flow_extract(struct ofpbuf *packet, uint32_t skb_priority, parse_tcp(packet, &b, flow); } else if (flow->nw_proto == IPPROTO_UDP) { parse_udp(packet, &b, flow); + } else if (flow->nw_proto == IPPROTO_SCTP) { + parse_sctp(packet, &b, flow); } else if (flow->nw_proto == IPPROTO_ICMP) { const struct icmp_header *icmp = pull_icmp(&b); if (icmp) { @@ -417,12 +478,15 @@ flow_extract(struct ofpbuf *packet, uint32_t skb_priority, parse_tcp(packet, &b, flow); } else if (flow->nw_proto == IPPROTO_UDP) { parse_udp(packet, &b, flow); + } else if (flow->nw_proto == IPPROTO_SCTP) { + parse_sctp(packet, &b, flow); } else if (flow->nw_proto == IPPROTO_ICMPV6) { if (parse_icmpv6(&b, flow)) { packet->l7 = b.data; } } - } else if (flow->dl_type == htons(ETH_TYPE_ARP)) { + } else if (flow->dl_type == htons(ETH_TYPE_ARP) || + flow->dl_type == htons(ETH_TYPE_RARP)) { const struct arp_eth_header *arp = pull_arp(&b); if (arp && arp->ar_hrd == htons(1) && arp->ar_pro == htons(ETH_TYPE_IP) @@ -433,13 +497,10 @@ flow_extract(struct ofpbuf *packet, uint32_t skb_priority, flow->nw_proto = ntohs(arp->ar_op); } - if ((flow->nw_proto == ARP_OP_REQUEST) - || (flow->nw_proto == ARP_OP_REPLY)) { - flow->nw_src = arp->ar_spa; - flow->nw_dst = arp->ar_tpa; - memcpy(flow->arp_sha, arp->ar_sha, ETH_ADDR_LEN); - memcpy(flow->arp_tha, arp->ar_tha, ETH_ADDR_LEN); - } + 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); } } } @@ -462,12 +523,15 @@ flow_zero_wildcards(struct flow *flow, const struct flow_wildcards *wildcards) void flow_get_metadata(const struct flow *flow, struct flow_metadata *fmd) { - BUILD_ASSERT_DECL(FLOW_WC_SEQ == 17); + BUILD_ASSERT_DECL(FLOW_WC_SEQ == 23); 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->in_port = flow->in_port; + fmd->pkt_mark = flow->pkt_mark; + fmd->in_port = flow->in_port.ofp_port; } char * @@ -478,94 +542,57 @@ flow_to_string(const struct flow *flow) return ds_cstr(&ds); } -static void format_tunnel_flags(uint16_t flags, struct ds *ds) +const char * +flow_tun_flag_to_string(uint32_t flags) +{ + switch (flags) { + case FLOW_TNL_F_DONT_FRAGMENT: + return "df"; + case FLOW_TNL_F_CSUM: + return "csum"; + case FLOW_TNL_F_KEY: + return "key"; + default: + return NULL; + } +} + +void +format_flags(struct ds *ds, const char *(*bit_to_string)(uint32_t), + uint32_t flags, char del) { - flags &= ~FLOW_TNL_F_KEY; + uint32_t bad = 0; - if (flags & FLOW_TNL_F_DONT_FRAGMENT) { - ds_put_cstr(ds, ",df"); - flags &= ~FLOW_TNL_F_DONT_FRAGMENT; + if (!flags) { + return; } + while (flags) { + uint32_t bit = rightmost_1bit(flags); + const char *s; - if (flags & FLOW_TNL_F_CSUM) { - ds_put_cstr(ds, ",csum"); - flags &= ~FLOW_TNL_F_CSUM; + s = bit_to_string(bit); + if (s) { + ds_put_format(ds, "%s%c", s, del); + } else { + bad |= bit; + } + + flags &= ~bit; } - if (flags) { - ds_put_format(ds, ",flags:%#"PRIx16, flags); + if (bad) { + ds_put_format(ds, "0x%"PRIx32"%c", bad, del); } + ds_chomp(ds, del); } void flow_format(struct ds *ds, const struct flow *flow) { - ds_put_format(ds, "priority:%"PRIu32, flow->skb_priority); - - if (flow->tunnel.ip_dst || flow->tunnel.tun_id) { - ds_put_cstr(ds, ",tunnel("); - ds_put_format(ds, IP_FMT"->"IP_FMT, IP_ARGS(&flow->tunnel.ip_src), - IP_ARGS(&flow->tunnel.ip_dst)); - - if (flow->tunnel.flags & FLOW_TNL_F_KEY) { - ds_put_format(ds, ",key:%#"PRIx64, ntohll(flow->tunnel.tun_id)); - } - ds_put_format(ds, ",tos:%#"PRIx8",ttl:%"PRIu8, flow->tunnel.ip_tos, - flow->tunnel.ip_ttl); - format_tunnel_flags(flow->tunnel.flags, ds); - ds_put_char(ds, ')'); - } - - ds_put_format(ds, ",metadata:%#"PRIx64 - ",in_port:%04"PRIx16, - ntohll(flow->metadata), - flow->in_port); + struct match match; - ds_put_format(ds, ",tci("); - if (flow->vlan_tci) { - ds_put_format(ds, "vlan:%"PRIu16",pcp:%d", - vlan_tci_to_vid(flow->vlan_tci), - vlan_tci_to_pcp(flow->vlan_tci)); - } else { - ds_put_char(ds, '0'); - } - ds_put_format(ds, ") mac("ETH_ADDR_FMT"->"ETH_ADDR_FMT - ") type:%04"PRIx16, - ETH_ADDR_ARGS(flow->dl_src), - ETH_ADDR_ARGS(flow->dl_dst), - ntohs(flow->dl_type)); - - if (flow->dl_type == htons(ETH_TYPE_IPV6)) { - ds_put_format(ds, " label:%#"PRIx32" proto:%"PRIu8" tos:%#"PRIx8 - " ttl:%"PRIu8" ipv6(", - ntohl(flow->ipv6_label), flow->nw_proto, - flow->nw_tos, flow->nw_ttl); - print_ipv6_addr(ds, &flow->ipv6_src); - ds_put_cstr(ds, "->"); - print_ipv6_addr(ds, &flow->ipv6_dst); - ds_put_char(ds, ')'); - } else if (flow->dl_type == htons(ETH_TYPE_IP) || - flow->dl_type == htons(ETH_TYPE_ARP)) { - ds_put_format(ds, " proto:%"PRIu8" tos:%#"PRIx8" ttl:%"PRIu8 - " ip("IP_FMT"->"IP_FMT")", - flow->nw_proto, flow->nw_tos, flow->nw_ttl, - IP_ARGS(&flow->nw_src), IP_ARGS(&flow->nw_dst)); - } - if (flow->nw_frag) { - ds_put_format(ds, " frag(%s)", - flow->nw_frag == FLOW_NW_FRAG_ANY ? "first" - : flow->nw_frag == (FLOW_NW_FRAG_ANY | FLOW_NW_FRAG_LATER) - ? "later" : ""); - } - if (flow->tp_src || flow->tp_dst) { - ds_put_format(ds, " port(%"PRIu16"->%"PRIu16")", - ntohs(flow->tp_src), ntohs(flow->tp_dst)); - } - if (!eth_addr_is_zero(flow->arp_sha) || !eth_addr_is_zero(flow->arp_tha)) { - ds_put_format(ds, " arp_ha("ETH_ADDR_FMT"->"ETH_ADDR_FMT")", - ETH_ADDR_ARGS(flow->arp_sha), - ETH_ADDR_ARGS(flow->arp_tha)); - } + match_wc_init(&match, flow); + match_format(&match, ds, OFP_DEFAULT_PRIORITY); } void @@ -585,15 +612,6 @@ 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. */ -void -flow_wildcards_init_exact(struct flow_wildcards *wc) -{ - memset(&wc->masks, 0xff, sizeof wc->masks); - memset(wc->masks.zeros, 0, sizeof wc->masks.zeros); -} - /* Returns true if 'wc' matches every packet, false if 'wc' fixes any bits or * fields. */ bool @@ -610,13 +628,13 @@ flow_wildcards_is_catchall(const struct flow_wildcards *wc) return true; } -/* Initializes 'dst' as the combination of wildcards in 'src1' and 'src2'. - * That is, a bit or a field is wildcarded in 'dst' if it is wildcarded in - * 'src1' or 'src2' or both. */ +/* Sets 'dst' as the bitwise AND of wildcards in 'src1' and 'src2'. + * That is, a bit or a field is wildcarded in 'dst' if it is wildcarded + * in 'src1' or 'src2' or both. */ void -flow_wildcards_combine(struct flow_wildcards *dst, - const struct flow_wildcards *src1, - const struct flow_wildcards *src2) +flow_wildcards_and(struct flow_wildcards *dst, + const struct flow_wildcards *src1, + const struct flow_wildcards *src2) { uint32_t *dst_u32 = (uint32_t *) &dst->masks; const uint32_t *src1_u32 = (const uint32_t *) &src1->masks; @@ -628,11 +646,88 @@ flow_wildcards_combine(struct flow_wildcards *dst, } } +/* Sets 'dst' as the bitwise OR of wildcards in 'src1' and 'src2'. That + * is, a bit or a field is wildcarded in 'dst' if it is neither + * wildcarded in 'src1' nor 'src2'. */ +void +flow_wildcards_or(struct flow_wildcards *dst, + const struct flow_wildcards *src1, + const struct flow_wildcards *src2) +{ + uint32_t *dst_u32 = (uint32_t *) &dst->masks; + const uint32_t *src1_u32 = (const uint32_t *) &src1->masks; + const uint32_t *src2_u32 = (const uint32_t *) &src2->masks; + size_t i; + + for (i = 0; i < FLOW_U32S; i++) { + dst_u32[i] = src1_u32[i] | src2_u32[i]; + } +} + +/* Perform a bitwise OR of miniflow 'src' flow data with the equivalent + * fields in 'dst', storing the result in 'dst'. */ +static void +flow_union_with_miniflow(struct flow *dst, const struct miniflow *src) +{ + uint32_t *dst_u32 = (uint32_t *) dst; + const uint32_t *p = src->values; + uint64_t map; + + for (map = src->map; map; map = zero_rightmost_1bit(map)) { + dst_u32[raw_ctz(map)] |= *p++; + } +} + +/* Fold minimask 'mask''s wildcard mask into 'wc's wildcard mask. */ +void +flow_wildcards_fold_minimask(struct flow_wildcards *wc, + const struct minimask *mask) +{ + flow_union_with_miniflow(&wc->masks, &mask->masks); +} + +inline uint64_t +miniflow_get_map_in_range(const struct miniflow *miniflow, + uint8_t start, uint8_t end, const uint32_t **data) +{ + uint64_t map = miniflow->map; + uint32_t *p = miniflow->values; + + if (start > 0) { + uint64_t msk = (UINT64_C(1) << start) - 1; /* 'start' LSBs set */ + p += count_1bits(map & msk); /* Skip to start. */ + map &= ~msk; + } + if (end < FLOW_U32S) { + uint64_t msk = (UINT64_C(1) << end) - 1; /* 'end' LSBs set */ + map &= msk; + } + + *data = p; + 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; + const uint32_t *p; + uint64_t map = miniflow_get_map_in_range(&mask->masks, start, end, &p); + + 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) { - return flow_hash(&wc->masks, basis);; + return flow_hash(&wc->masks, basis); } /* Returns true if 'a' and 'b' represent the same wildcards, false if they are @@ -719,7 +814,7 @@ 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)) { @@ -731,11 +826,84 @@ 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 hash_bytes(&fields, sizeof fields, basis); + 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(const struct flow *flow, struct flow_wildcards *wc, + enum nx_hash_fields fields) +{ + switch (fields) { + case NX_HASH_FIELDS_ETH_SRC: + memset(&wc->masks.dl_src, 0xff, sizeof wc->masks.dl_src); + break; + + 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); + 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); + memset(&wc->masks.tp_src, 0xff, sizeof wc->masks.tp_src); + memset(&wc->masks.tp_dst, 0xff, sizeof wc->masks.tp_dst); + } + wc->masks.vlan_tci |= htons(VLAN_VID_MASK | VLAN_CFI); + break; + + default: + NOT_REACHED(); + } } /* Hashes the portions of 'flow' designated by 'fields'. */ @@ -746,7 +914,7 @@ flow_hash_fields(const struct flow *flow, enum nx_hash_fields fields, switch (fields) { case NX_HASH_FIELDS_ETH_SRC: - return hash_bytes(flow->dl_src, sizeof flow->dl_src, basis); + return jhash_bytes(flow->dl_src, sizeof flow->dl_src, basis); case NX_HASH_FIELDS_SYMMETRIC_L4: return flow_hash_symmetric_l4(flow, basis); @@ -774,6 +942,24 @@ flow_hash_fields_valid(enum nx_hash_fields fields) || fields == NX_HASH_FIELDS_SYMMETRIC_L4; } +/* Returns a hash value for the bits of 'flow' that are active based on + * 'wc', given 'basis'. */ +uint32_t +flow_hash_in_wildcards(const struct flow *flow, + const struct flow_wildcards *wc, uint32_t basis) +{ + const uint32_t *wc_u32 = (const uint32_t *) &wc->masks; + const uint32_t *flow_u32 = (const uint32_t *) flow; + uint32_t hash; + size_t i; + + hash = basis; + for (i = 0; i < FLOW_U32S; i++) { + hash = mhash_add(hash, flow_u32[i] & wc_u32[i]); + } + return mhash_finish(hash, 4 * FLOW_U32S); +} + /* Sets the VLAN VID that 'flow' matches to 'vid', which is interpreted as an * OpenFlow 1.0 "dl_vlan" value: * @@ -823,6 +1009,37 @@ flow_set_vlan_pcp(struct flow *flow, uint8_t pcp) flow->vlan_tci |= htons((pcp << VLAN_PCP_SHIFT) | VLAN_CFI); } +/* 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) +{ + set_mpls_lse_label(&flow->mpls_lse, 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) +{ + set_mpls_lse_ttl(&flow->mpls_lse, 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) +{ + set_mpls_lse_tc(&flow->mpls_lse, 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) +{ + set_mpls_lse_bos(&flow->mpls_lse, bos); +} + /* Puts into 'b' a packet that flow_extract() would parse as having the given * 'flow'. * @@ -849,9 +1066,10 @@ flow_compose(struct ofpbuf *b, const struct flow *flow) b->l3 = 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); @@ -867,13 +1085,19 @@ flow_compose(struct ofpbuf *b, const struct flow *flow) 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); + 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; @@ -882,6 +1106,7 @@ flow_compose(struct ofpbuf *b, const struct flow *flow) icmp->icmp_code = ntohs(flow->tp_dst); icmp->icmp_csum = csum(icmp, ICMP_HEADER_LEN); } + b->l7 = ofpbuf_tail(b); } ip = b->l3; @@ -890,7 +1115,8 @@ flow_compose(struct ofpbuf *b, const struct flow *flow) ip->ip_csum = csum(ip, sizeof *ip); } else if (flow->dl_type == htons(ETH_TYPE_IPV6)) { /* XXX */ - } else if (flow->dl_type == htons(ETH_TYPE_ARP)) { + } 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); @@ -902,12 +1128,17 @@ 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); + } } /* Compressed flow. */ @@ -915,13 +1146,7 @@ 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 * @@ -935,36 +1160,59 @@ 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'. + * + * This function initializes 'dst->values' (either inline if possible or with + * malloc() otherwise) and copies the nonzero uint32_t elements of 'src' 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' @@ -973,11 +1221,26 @@ 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 @@ -992,40 +1255,19 @@ miniflow_destroy(struct miniflow *flow) void miniflow_expand(const struct miniflow *src, struct flow *dst) { - uint32_t *dst_u32 = (uint32_t *) dst; - int ofs; - int i; - - memset(dst_u32, 0, sizeof *dst); - - 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++]; - } - } + memset(dst, 0, sizeof *dst); + flow_union_with_miniflow(dst, src); } 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' @@ -1059,16 +1301,31 @@ 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; + + 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; - for (i = 0; i < MINI_N_MAPS; i++) { - if (a->map[i] != b->map[i]) { - return false; + 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 @@ -1078,20 +1335,17 @@ 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; @@ -1105,20 +1359,17 @@ 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; @@ -1128,10 +1379,22 @@ 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 @@ -1145,21 +1408,19 @@ 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; + for (map = mask->masks.map; map; map = zero_rightmost_1bit(map)) { + if (*p) { + int ofs = raw_ctz(map); hash = mhash_add(hash, miniflow_get(flow, ofs) & *p); - p++; } + p++; } - return mhash_finish(hash, p - mask->masks.values); + return mhash_finish(hash, (p - mask->masks.values) * 4); } /* Returns a hash value for the bits of 'flow' where there are 1-bits in @@ -1171,25 +1432,48 @@ 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)) { + if (*p) { + hash = mhash_add(hash, flow_u32[raw_ctz(map)] & *p); + } + 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; + const uint32_t *p; + uint64_t map = miniflow_get_map_in_range(&mask->masks, start, end, &p); + uint32_t hash = *basis; + + for (; map; map = zero_rightmost_1bit(map)) { + if (*p) { + hash = mhash_add(hash, flow_u32[raw_ctz(map)] & *p); } + p++; } - return mhash_finish(hash, p - mask->masks.values); + *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(). */ @@ -1207,6 +1491,14 @@ 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 @@ -1219,23 +1511,19 @@ 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; } } } @@ -1292,20 +1580,15 @@ 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; + uint64_t map; - for (i = 0; i < MINI_N_MAPS; i++) { - uint32_t map; + 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); - 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); - - if ((a_u32 & b_u32) != b_u32) { - return true; - } + if ((a_u32 & b_u32) != b_u32) { + return true; } } @@ -1318,7 +1601,13 @@ 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; }