X-Git-Url: http://git.onelab.eu/?a=blobdiff_plain;f=lib%2Fflow.c;h=fc616106e56b541844ca86e8ded408896c386711;hb=e0edde6fee279cdbbf3c179f5f50adaf0c7c7f1e;hp=6842d3b5a17cc4433dac6e68cdfa0cfa9160cbda;hpb=ff55ea1f29360533699edb77747a919505282057;p=sliver-openvswitch.git diff --git a/lib/flow.c b/lib/flow.c index 6842d3b5a..fc616106e 100644 --- a/lib/flow.c +++ b/lib/flow.c @@ -1,5 +1,5 @@ /* - * Copyright (c) 2008, 2009, 2010, 2011 Nicira Networks. + * Copyright (c) 2008, 2009, 2010, 2011, 2012 Nicira, Inc. * * Licensed under the Apache License, Version 2.0 (the "License"); * you may not use this file except in compliance with the License. @@ -16,18 +16,20 @@ #include #include #include "flow.h" +#include +#include #include #include +#include +#include #include #include #include "byte-order.h" #include "coverage.h" -#include "dpif.h" #include "dynamic-string.h" #include "hash.h" #include "ofpbuf.h" #include "openflow/openflow.h" -#include "openvswitch/datapath-protocol.h" #include "packets.h" #include "unaligned.h" #include "vlog.h" @@ -80,6 +82,12 @@ pull_icmp(struct ofpbuf *packet) return ofpbuf_try_pull(packet, ICMP_HEADER_LEN); } +static struct icmp6_hdr * +pull_icmpv6(struct ofpbuf *packet) +{ + return ofpbuf_try_pull(packet, sizeof(struct icmp6_hdr)); +} + static void parse_vlan(struct ofpbuf *b, struct flow *flow) { @@ -122,7 +130,191 @@ parse_ethertype(struct ofpbuf *b) return llc->snap.snap_type; } -/* Initializes 'flow' members from 'packet', 'tun_id', and 'in_port. +static int +parse_ipv6(struct ofpbuf *packet, struct flow *flow) +{ + const struct ip6_hdr *nh; + ovs_be32 tc_flow; + int nexthdr; + + nh = ofpbuf_try_pull(packet, sizeof *nh); + if (!nh) { + return EINVAL; + } + + nexthdr = nh->ip6_nxt; + + flow->ipv6_src = nh->ip6_src; + flow->ipv6_dst = nh->ip6_dst; + + tc_flow = get_unaligned_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; + flow->nw_proto = IPPROTO_NONE; + + while (1) { + if ((nexthdr != IPPROTO_HOPOPTS) + && (nexthdr != IPPROTO_ROUTING) + && (nexthdr != IPPROTO_DSTOPTS) + && (nexthdr != IPPROTO_AH) + && (nexthdr != IPPROTO_FRAGMENT)) { + /* It's either a terminal header (e.g., TCP, UDP) or one we + * don't understand. In either case, we're done with the + * packet, so use it to fill in 'nw_proto'. */ + break; + } + + /* We only verify that at least 8 bytes of the next header are + * available, but many of these headers are longer. Ensure that + * 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) { + return EINVAL; + } + + if ((nexthdr == IPPROTO_HOPOPTS) + || (nexthdr == IPPROTO_ROUTING) + || (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 = (struct ip6_ext *)packet->data; + nexthdr = ext_hdr->ip6e_nxt; + if (!ofpbuf_try_pull(packet, (ext_hdr->ip6e_len + 1) * 8)) { + return EINVAL; + } + } else if (nexthdr == IPPROTO_AH) { + /* A standard AH definition isn't available, but the fields + * 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 = (struct ip6_ext *)packet->data; + 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 = (struct ip6_frag *)packet->data; + + nexthdr = frag_hdr->ip6f_nxt; + if (!ofpbuf_try_pull(packet, sizeof *frag_hdr)) { + return EINVAL; + } + + /* 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_LATER; + nexthdr = IPPROTO_FRAGMENT; + break; + } + } + } + } + + flow->nw_proto = nexthdr; + return 0; +} + +static void +parse_tcp(struct ofpbuf *packet, struct ofpbuf *b, struct flow *flow) +{ + const struct tcp_header *tcp = pull_tcp(b); + if (tcp) { + flow->tp_src = tcp->tcp_src; + flow->tp_dst = tcp->tcp_dst; + packet->l7 = b->data; + } +} + +static void +parse_udp(struct ofpbuf *packet, struct ofpbuf *b, struct flow *flow) +{ + const struct udp_header *udp = pull_udp(b); + if (udp) { + flow->tp_src = udp->udp_src; + flow->tp_dst = udp->udp_dst; + packet->l7 = b->data; + } +} + +static bool +parse_icmpv6(struct ofpbuf *b, struct flow *flow) +{ + const struct icmp6_hdr *icmp = pull_icmpv6(b); + + if (!icmp) { + return false; + } + + /* The ICMPv6 type and code fields use the 16-bit transport port + * fields, so we need to store them in 16-bit network byte order. */ + flow->tp_src = htons(icmp->icmp6_type); + flow->tp_dst = htons(icmp->icmp6_code); + + if (icmp->icmp6_code == 0 && + (icmp->icmp6_type == ND_NEIGHBOR_SOLICIT || + icmp->icmp6_type == ND_NEIGHBOR_ADVERT)) { + const struct in6_addr *nd_target; + + nd_target = ofpbuf_try_pull(b, sizeof *nd_target); + if (!nd_target) { + return false; + } + flow->nd_target = *nd_target; + + while (b->size >= 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; + int opt_len = nd_opt->nd_opt_len * 8; + + if (!opt_len || opt_len > b->size) { + goto invalid; + } + + /* Store the link layer address if the appropriate option is + * provided. It is considered an error if the same link + * layer option is specified twice. */ + if (nd_opt->nd_opt_type == ND_OPT_SOURCE_LINKADDR + && opt_len == 8) { + if (eth_addr_is_zero(flow->arp_sha)) { + memcpy(flow->arp_sha, nd_opt + 1, ETH_ADDR_LEN); + } else { + goto invalid; + } + } else if (nd_opt->nd_opt_type == ND_OPT_TARGET_LINKADDR + && opt_len == 8) { + if (eth_addr_is_zero(flow->arp_tha)) { + memcpy(flow->arp_tha, nd_opt + 1, ETH_ADDR_LEN); + } else { + goto invalid; + } + } + + if (!ofpbuf_try_pull(b, opt_len)) { + goto invalid; + } + } + } + + return true; + +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; + +} + +/* Initializes 'flow' members from 'packet', 'skb_priority', 'tun_id', and + * 'ofp_in_port'. + * * Initializes 'packet' header pointers as follows: * * - packet->l2 to the start of the Ethernet header. @@ -137,19 +329,19 @@ parse_ethertype(struct ofpbuf *b) * - packet->l7 to just past the TCP or UDP or ICMP header, if one is * present and has a correct length, and otherwise NULL. */ -int -flow_extract(struct ofpbuf *packet, ovs_be64 tun_id, uint16_t in_port, - struct flow *flow) +void +flow_extract(struct ofpbuf *packet, uint32_t skb_priority, ovs_be64 tun_id, + uint16_t ofp_in_port, struct flow *flow) { struct ofpbuf b = *packet; struct eth_header *eth; - int retval = 0; COVERAGE_INC(flow_extract); memset(flow, 0, sizeof *flow); flow->tun_id = tun_id; - flow->in_port = in_port; + flow->in_port = ofp_in_port; + flow->skb_priority = skb_priority; packet->l2 = b.data; packet->l3 = NULL; @@ -157,7 +349,7 @@ flow_extract(struct ofpbuf *packet, ovs_be64 tun_id, uint16_t in_port, packet->l7 = NULL; if (b.size < sizeof *eth) { - return 0; + return; } /* Link layer. */ @@ -177,36 +369,49 @@ flow_extract(struct ofpbuf *packet, ovs_be64 tun_id, uint16_t in_port, if (flow->dl_type == htons(ETH_TYPE_IP)) { const struct ip_header *nh = pull_ip(&b); 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_tos = nh->ip_tos & IP_DSCP_MASK; flow->nw_proto = nh->ip_proto; - packet->l4 = b.data; - if (!IP_IS_FRAGMENT(nh->ip_frag_off)) { - if (flow->nw_proto == IP_TYPE_TCP) { - const struct tcp_header *tcp = pull_tcp(&b); - if (tcp) { - flow->tp_src = tcp->tcp_src; - flow->tp_dst = tcp->tcp_dst; - packet->l7 = b.data; - } - } else if (flow->nw_proto == IP_TYPE_UDP) { - const struct udp_header *udp = pull_udp(&b); - if (udp) { - flow->tp_src = udp->udp_src; - flow->tp_dst = udp->udp_dst; - packet->l7 = b.data; - } - } else if (flow->nw_proto == IP_TYPE_ICMP) { + + flow->nw_tos = nh->ip_tos; + if (IP_IS_FRAGMENT(nh->ip_frag_off)) { + flow->nw_frag = FLOW_NW_FRAG_ANY; + if (nh->ip_frag_off & htons(IP_FRAG_OFF_MASK)) { + flow->nw_frag |= FLOW_NW_FRAG_LATER; + } + } + flow->nw_ttl = nh->ip_ttl; + + if (!(nh->ip_frag_off & htons(IP_FRAG_OFF_MASK))) { + if (flow->nw_proto == IPPROTO_TCP) { + parse_tcp(packet, &b, flow); + } else if (flow->nw_proto == IPPROTO_UDP) { + parse_udp(packet, &b, flow); + } else if (flow->nw_proto == IPPROTO_ICMP) { const struct icmp_header *icmp = pull_icmp(&b); if (icmp) { - flow->icmp_type = htons(icmp->icmp_type); - flow->icmp_code = htons(icmp->icmp_code); + flow->tp_src = htons(icmp->icmp_type); + flow->tp_dst = htons(icmp->icmp_code); packet->l7 = b.data; } } - } else { - retval = 1; + } + } + } else if (flow->dl_type == htons(ETH_TYPE_IPV6)) { + if (parse_ipv6(&b, flow)) { + return; + } + + packet->l4 = b.data; + if (flow->nw_proto == IPPROTO_TCP) { + parse_tcp(packet, &b, flow); + } else if (flow->nw_proto == IPPROTO_UDP) { + parse_udp(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)) { @@ -224,30 +429,92 @@ flow_extract(struct ofpbuf *packet, ovs_be64 tun_id, uint16_t in_port, || (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); } } } - return retval; } -/* Extracts the flow stats for a packet. The 'flow' and 'packet' - * arguments must have been initialized through a call to flow_extract(). - */ +/* For every bit of a field that is wildcarded in 'wildcards', sets the + * corresponding bit in 'flow' to zero. */ void -flow_extract_stats(const struct flow *flow, struct ofpbuf *packet, - struct dpif_flow_stats *stats) +flow_zero_wildcards(struct flow *flow, const struct flow_wildcards *wildcards) { - memset(stats, 0, sizeof(*stats)); + const flow_wildcards_t wc = wildcards->wildcards; + int i; - if ((flow->dl_type == htons(ETH_TYPE_IP)) && packet->l4) { - if ((flow->nw_proto == IP_TYPE_TCP) && packet->l7) { - struct tcp_header *tcp = packet->l4; - stats->tcp_flags = TCP_FLAGS(tcp->tcp_ctl); - } + BUILD_ASSERT_DECL(FLOW_WC_SEQ == 10); + + for (i = 0; i < FLOW_N_REGS; i++) { + flow->regs[i] &= wildcards->reg_masks[i]; + } + flow->tun_id &= wildcards->tun_id_mask; + flow->nw_src &= wildcards->nw_src_mask; + flow->nw_dst &= wildcards->nw_dst_mask; + if (wc & FWW_IN_PORT) { + flow->in_port = 0; + } + flow->vlan_tci &= wildcards->vlan_tci_mask; + if (wc & FWW_DL_TYPE) { + flow->dl_type = htons(0); + } + flow->tp_src &= wildcards->tp_src_mask; + flow->tp_dst &= wildcards->tp_dst_mask; + if (wc & FWW_DL_SRC) { + memset(flow->dl_src, 0, sizeof flow->dl_src); } + if (wc & FWW_DL_DST) { + flow->dl_dst[0] &= 0x01; + memset(&flow->dl_dst[1], 0, 5); + } + if (wc & FWW_ETH_MCAST) { + flow->dl_dst[0] &= 0xfe; + } + if (wc & FWW_NW_PROTO) { + flow->nw_proto = 0; + } + if (wc & FWW_IPV6_LABEL) { + flow->ipv6_label = htonl(0); + } + if (wc & FWW_NW_DSCP) { + flow->nw_tos &= ~IP_DSCP_MASK; + } + if (wc & FWW_NW_ECN) { + flow->nw_tos &= ~IP_ECN_MASK; + } + if (wc & FWW_NW_TTL) { + flow->nw_ttl = 0; + } + flow->nw_frag &= wildcards->nw_frag_mask; + if (wc & FWW_ARP_SHA) { + memset(flow->arp_sha, 0, sizeof flow->arp_sha); + } + if (wc & FWW_ARP_THA) { + memset(flow->arp_tha, 0, sizeof flow->arp_tha); + } + flow->ipv6_src = ipv6_addr_bitand(&flow->ipv6_src, + &wildcards->ipv6_src_mask); + flow->ipv6_dst = ipv6_addr_bitand(&flow->ipv6_dst, + &wildcards->ipv6_dst_mask); + flow->nd_target = ipv6_addr_bitand(&flow->nd_target, + &wildcards->nd_target_mask); + flow->skb_priority = 0; +} + +/* 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 == 10); + + fmd->tun_id = flow->tun_id; + fmd->tun_id_mask = htonll(UINT64_MAX); + + memcpy(fmd->regs, flow->regs, sizeof fmd->regs); + memset(fmd->reg_masks, 0xff, sizeof fmd->reg_masks); - stats->n_bytes = packet->size; - stats->n_packets = 1; + fmd->in_port = flow->in_port; } char * @@ -261,30 +528,57 @@ flow_to_string(const struct flow *flow) void flow_format(struct ds *ds, const struct flow *flow) { - ds_put_format(ds, "tunnel%#"PRIx64":in_port%04"PRIx16":tci(", - flow->tun_id, flow->in_port); + ds_put_format(ds, "priority:%"PRIu32 + ",tunnel:%#"PRIx64 + ",in_port:%04"PRIx16, + flow->skb_priority, + ntohll(flow->tun_id), + flow->in_port); + + ds_put_format(ds, ",tci("); if (flow->vlan_tci) { - ds_put_format(ds, "vlan%"PRIu16",pcp%d", + 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 - " proto%"PRIu8 - " tos%"PRIu8 - " ip"IP_FMT"->"IP_FMT - " port%"PRIu16"->%"PRIu16, + 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), - flow->nw_proto, - flow->nw_tos, - IP_ARGS(&flow->nw_src), - IP_ARGS(&flow->nw_dst), - ntohs(flow->tp_src), - ntohs(flow->tp_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 { + 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)); + } } void @@ -301,13 +595,21 @@ flow_print(FILE *stream, const struct flow *flow) void flow_wildcards_init_catchall(struct flow_wildcards *wc) { + BUILD_ASSERT_DECL(FLOW_WC_SEQ == 10); + wc->wildcards = FWW_ALL; wc->tun_id_mask = htonll(0); wc->nw_src_mask = htonl(0); wc->nw_dst_mask = htonl(0); + wc->ipv6_src_mask = in6addr_any; + wc->ipv6_dst_mask = in6addr_any; + wc->nd_target_mask = in6addr_any; memset(wc->reg_masks, 0, sizeof wc->reg_masks); wc->vlan_tci_mask = htons(0); - wc->zero = 0; + wc->nw_frag_mask = 0; + wc->tp_src_mask = htons(0); + wc->tp_dst_mask = htons(0); + memset(wc->zeros, 0, sizeof wc->zeros); } /* Initializes 'wc' as an exact-match set of wildcards; that is, 'wc' does not @@ -315,13 +617,21 @@ flow_wildcards_init_catchall(struct flow_wildcards *wc) void flow_wildcards_init_exact(struct flow_wildcards *wc) { + BUILD_ASSERT_DECL(FLOW_WC_SEQ == 10); + wc->wildcards = 0; wc->tun_id_mask = htonll(UINT64_MAX); wc->nw_src_mask = htonl(UINT32_MAX); wc->nw_dst_mask = htonl(UINT32_MAX); + wc->ipv6_src_mask = in6addr_exact; + wc->ipv6_dst_mask = in6addr_exact; + wc->nd_target_mask = in6addr_exact; memset(wc->reg_masks, 0xff, sizeof wc->reg_masks); wc->vlan_tci_mask = htons(UINT16_MAX); - wc->zero = 0; + wc->nw_frag_mask = UINT8_MAX; + wc->tp_src_mask = htons(UINT16_MAX); + wc->tp_dst_mask = htons(UINT16_MAX); + memset(wc->zeros, 0, sizeof wc->zeros); } /* Returns true if 'wc' is exact-match, false if 'wc' wildcards any bits or @@ -331,16 +641,56 @@ flow_wildcards_is_exact(const struct flow_wildcards *wc) { int i; + BUILD_ASSERT_DECL(FLOW_WC_SEQ == 10); + if (wc->wildcards || wc->tun_id_mask != htonll(UINT64_MAX) || wc->nw_src_mask != htonl(UINT32_MAX) || wc->nw_dst_mask != htonl(UINT32_MAX) - || wc->vlan_tci_mask != htons(UINT16_MAX)) { + || wc->tp_src_mask != htons(UINT16_MAX) + || wc->tp_dst_mask != htons(UINT16_MAX) + || wc->vlan_tci_mask != htons(UINT16_MAX) + || !ipv6_mask_is_exact(&wc->ipv6_src_mask) + || !ipv6_mask_is_exact(&wc->ipv6_dst_mask) + || !ipv6_mask_is_exact(&wc->nd_target_mask) + || wc->nw_frag_mask != UINT8_MAX) { + return false; + } + + for (i = 0; i < FLOW_N_REGS; i++) { + if (wc->reg_masks[i] != UINT32_MAX) { + return false; + } + } + + return true; +} + +/* Returns true if 'wc' matches every packet, false if 'wc' fixes any bits or + * fields. */ +bool +flow_wildcards_is_catchall(const struct flow_wildcards *wc) +{ + int i; + + BUILD_ASSERT_DECL(FLOW_WC_SEQ == 10); + + if (wc->wildcards != FWW_ALL + || wc->tun_id_mask != htonll(0) + || wc->nw_src_mask != htonl(0) + || wc->nw_dst_mask != htonl(0) + || wc->tp_src_mask != htons(0) + || wc->tp_dst_mask != htons(0) + || wc->vlan_tci_mask != htons(0) + || !ipv6_mask_is_any(&wc->ipv6_src_mask) + || !ipv6_mask_is_any(&wc->ipv6_dst_mask) + || !ipv6_mask_is_any(&wc->nd_target_mask) + || wc->nw_frag_mask != 0) { return false; } for (i = 0; i < FLOW_N_REGS; i++) { - if (wc->reg_masks[i] != htonl(UINT32_MAX)) { + if (wc->reg_masks[i] != 0) { return false; } } @@ -358,25 +708,35 @@ flow_wildcards_combine(struct flow_wildcards *dst, { int i; + BUILD_ASSERT_DECL(FLOW_WC_SEQ == 10); + dst->wildcards = src1->wildcards | src2->wildcards; dst->tun_id_mask = src1->tun_id_mask & src2->tun_id_mask; dst->nw_src_mask = src1->nw_src_mask & src2->nw_src_mask; dst->nw_dst_mask = src1->nw_dst_mask & src2->nw_dst_mask; + dst->ipv6_src_mask = ipv6_addr_bitand(&src1->ipv6_src_mask, + &src2->ipv6_src_mask); + dst->ipv6_dst_mask = ipv6_addr_bitand(&src1->ipv6_dst_mask, + &src2->ipv6_dst_mask); + dst->nd_target_mask = ipv6_addr_bitand(&src1->nd_target_mask, + &src2->nd_target_mask); for (i = 0; i < FLOW_N_REGS; i++) { dst->reg_masks[i] = src1->reg_masks[i] & src2->reg_masks[i]; } dst->vlan_tci_mask = src1->vlan_tci_mask & src2->vlan_tci_mask; + dst->tp_src_mask = src1->tp_src_mask & src2->tp_src_mask; + dst->tp_dst_mask = src1->tp_dst_mask & src2->tp_dst_mask; } /* Returns a hash of the wildcards in 'wc'. */ uint32_t -flow_wildcards_hash(const struct flow_wildcards *wc) +flow_wildcards_hash(const struct flow_wildcards *wc, uint32_t basis) { /* If you change struct flow_wildcards and thereby trigger this * assertion, please check that the new struct flow_wildcards has no holes * in it before you update the assertion. */ - BUILD_ASSERT_DECL(sizeof *wc == 24 + FLOW_N_REGS * 4); - return hash_bytes(wc, sizeof *wc, 0); + BUILD_ASSERT_DECL(sizeof *wc == 80 + FLOW_N_REGS * 4); + return hash_bytes(wc, sizeof *wc, basis); } /* Returns true if 'a' and 'b' represent the same wildcards, false if they are @@ -387,11 +747,18 @@ flow_wildcards_equal(const struct flow_wildcards *a, { int i; + BUILD_ASSERT_DECL(FLOW_WC_SEQ == 10); + if (a->wildcards != b->wildcards || a->tun_id_mask != b->tun_id_mask || a->nw_src_mask != b->nw_src_mask || a->nw_dst_mask != b->nw_dst_mask - || a->vlan_tci_mask != b->vlan_tci_mask) { + || a->vlan_tci_mask != b->vlan_tci_mask + || !ipv6_addr_equals(&a->ipv6_src_mask, &b->ipv6_src_mask) + || !ipv6_addr_equals(&a->ipv6_dst_mask, &b->ipv6_dst_mask) + || !ipv6_addr_equals(&a->nd_target_mask, &b->nd_target_mask) + || a->tp_src_mask != b->tp_src_mask + || a->tp_dst_mask != b->tp_dst_mask) { return false; } @@ -411,6 +778,9 @@ flow_wildcards_has_extra(const struct flow_wildcards *a, const struct flow_wildcards *b) { int i; + struct in6_addr ipv6_masked; + + BUILD_ASSERT_DECL(FLOW_WC_SEQ == 10); for (i = 0; i < FLOW_N_REGS; i++) { if ((a->reg_masks[i] & b->reg_masks[i]) != b->reg_masks[i]) { @@ -418,48 +788,113 @@ flow_wildcards_has_extra(const struct flow_wildcards *a, } } + ipv6_masked = ipv6_addr_bitand(&a->ipv6_src_mask, &b->ipv6_src_mask); + if (!ipv6_addr_equals(&ipv6_masked, &b->ipv6_src_mask)) { + return true; + } + + ipv6_masked = ipv6_addr_bitand(&a->ipv6_dst_mask, &b->ipv6_dst_mask); + if (!ipv6_addr_equals(&ipv6_masked, &b->ipv6_dst_mask)) { + return true; + } + + ipv6_masked = ipv6_addr_bitand(&a->nd_target_mask, &b->nd_target_mask); + if (!ipv6_addr_equals(&ipv6_masked, &b->nd_target_mask)) { + return true; + } + return (a->wildcards & ~b->wildcards || (a->tun_id_mask & b->tun_id_mask) != b->tun_id_mask || (a->nw_src_mask & b->nw_src_mask) != b->nw_src_mask || (a->nw_dst_mask & b->nw_dst_mask) != b->nw_dst_mask - || (a->vlan_tci_mask & b->vlan_tci_mask) != b->vlan_tci_mask); + || (a->vlan_tci_mask & b->vlan_tci_mask) != b->vlan_tci_mask + || (a->tp_src_mask & b->tp_src_mask) != b->tp_src_mask + || (a->tp_dst_mask & b->tp_dst_mask) != b->tp_dst_mask); } -static bool -set_nw_mask(ovs_be32 *maskp, ovs_be32 mask) +/* Sets the wildcard mask for register 'idx' in 'wc' to 'mask'. + * (A 0-bit indicates a wildcard bit.) */ +void +flow_wildcards_set_reg_mask(struct flow_wildcards *wc, int idx, uint32_t mask) { - if (ip_is_cidr(mask)) { - *maskp = mask; - return true; - } else { - return false; - } + wc->reg_masks[idx] = mask; } -/* Sets the IP (or ARP) source wildcard mask to CIDR 'mask' (consisting of N - * high-order 1-bit and 32-N low-order 0-bits). Returns true if successful, - * false if 'mask' is not a CIDR mask. */ -bool -flow_wildcards_set_nw_src_mask(struct flow_wildcards *wc, ovs_be32 mask) +/* Returns the wildcard bitmask for the Ethernet destination address + * that 'wc' specifies. The bitmask has a 0 in each bit that is wildcarded + * and a 1 in each bit that must match. */ +const uint8_t * +flow_wildcards_to_dl_dst_mask(flow_wildcards_t wc) { - return set_nw_mask(&wc->nw_src_mask, mask); + static const uint8_t no_wild[] = {0xff, 0xff, 0xff, 0xff, 0xff, 0xff}; + static const uint8_t addr_wild[] = {0x01, 0x00, 0x00, 0x00, 0x00, 0x00}; + static const uint8_t mcast_wild[] = {0xfe, 0xff, 0xff, 0xff, 0xff, 0xff}; + static const uint8_t all_wild[] = {0x00, 0x00, 0x00, 0x00, 0x00, 0x00}; + + switch (wc & (FWW_DL_DST | FWW_ETH_MCAST)) { + case 0: return no_wild; + case FWW_DL_DST: return addr_wild; + case FWW_ETH_MCAST: return mcast_wild; + case FWW_DL_DST | FWW_ETH_MCAST: return all_wild; + } + NOT_REACHED(); } -/* Sets the IP (or ARP) destination wildcard mask to CIDR 'mask' (consisting of - * N high-order 1-bit and 32-N low-order 0-bits). Returns true if successful, - * false if 'mask' is not a CIDR mask. */ +/* Returns true if 'mask' is a valid wildcard bitmask for the Ethernet + * destination address. Valid bitmasks are either all-bits-0 or all-bits-1, + * except that the multicast bit may differ from the rest of the bits. So, + * there are four possible valid bitmasks: + * + * - 00:00:00:00:00:00 + * - 01:00:00:00:00:00 + * - fe:ff:ff:ff:ff:ff + * - ff:ff:ff:ff:ff:ff + * + * All other bitmasks are invalid. */ bool -flow_wildcards_set_nw_dst_mask(struct flow_wildcards *wc, ovs_be32 mask) +flow_wildcards_is_dl_dst_mask_valid(const uint8_t mask[ETH_ADDR_LEN]) { - return set_nw_mask(&wc->nw_dst_mask, mask); + switch (mask[0]) { + case 0x00: + case 0x01: + return (mask[1] | mask[2] | mask[3] | mask[4] | mask[5]) == 0x00; + + case 0xfe: + case 0xff: + return (mask[1] & mask[2] & mask[3] & mask[4] & mask[5]) == 0xff; + + default: + return false; + } } -/* Sets the wildcard mask for register 'idx' in 'wc' to 'mask'. - * (A 0-bit indicates a wildcard bit.) */ -void -flow_wildcards_set_reg_mask(struct flow_wildcards *wc, int idx, uint32_t mask) +/* Returns 'wc' with the FWW_DL_DST and FWW_ETH_MCAST bits modified + * appropriately to match 'mask'. + * + * This function will assert-fail if 'mask' is invalid. Only 'mask' values + * accepted by flow_wildcards_is_dl_dst_mask_valid() are allowed. */ +flow_wildcards_t +flow_wildcards_set_dl_dst_mask(flow_wildcards_t wc, + const uint8_t mask[ETH_ADDR_LEN]) { - wc->reg_masks[idx] = mask; + assert(flow_wildcards_is_dl_dst_mask_valid(mask)); + + switch (mask[0]) { + case 0x00: + return wc | FWW_DL_DST | FWW_ETH_MCAST; + + case 0x01: + return (wc | FWW_DL_DST) & ~FWW_ETH_MCAST; + + case 0xfe: + return (wc & ~FWW_DL_DST) | FWW_ETH_MCAST; + + case 0xff: + return wc & ~(FWW_DL_DST | FWW_ETH_MCAST); + + default: + NOT_REACHED(); + } } /* Hashes 'flow' based on its L2 through L4 protocol information. */ @@ -467,10 +902,13 @@ uint32_t flow_hash_symmetric_l4(const struct flow *flow, uint32_t basis) { struct { - ovs_be32 ip_addr; + union { + ovs_be32 ipv4_addr; + struct in6_addr ipv6_addr; + }; ovs_be16 eth_type; ovs_be16 vlan_tci; - ovs_be16 tp_addr; + ovs_be16 tp_port; uint8_t eth_addr[ETH_ADDR_LEN]; uint8_t ip_proto; } fields; @@ -483,18 +921,185 @@ flow_hash_symmetric_l4(const struct flow *flow, uint32_t basis) } fields.vlan_tci = flow->vlan_tci & htons(VLAN_VID_MASK); fields.eth_type = flow->dl_type; + + /* UDP source and destination port are not taken into account because they + * will not necessarily be symmetric in a bidirectional flow. */ if (fields.eth_type == htons(ETH_TYPE_IP)) { - fields.ip_addr = flow->nw_src ^ flow->nw_dst; + fields.ipv4_addr = flow->nw_src ^ flow->nw_dst; fields.ip_proto = flow->nw_proto; - if (fields.ip_proto == IP_TYPE_TCP || fields.ip_proto == IP_TYPE_UDP) { - fields.tp_addr = flow->tp_src ^ flow->tp_dst; - } else { - fields.tp_addr = htons(0); + if (fields.ip_proto == IPPROTO_TCP) { + fields.tp_port = flow->tp_src ^ flow->tp_dst; + } + } else if (fields.eth_type == htons(ETH_TYPE_IPV6)) { + const uint8_t *a = &flow->ipv6_src.s6_addr[0]; + const uint8_t *b = &flow->ipv6_dst.s6_addr[0]; + uint8_t *ipv6_addr = &fields.ipv6_addr.s6_addr[0]; + + for (i=0; i<16; i++) { + ipv6_addr[i] = a[i] ^ b[i]; + } + fields.ip_proto = flow->nw_proto; + if (fields.ip_proto == IPPROTO_TCP) { + fields.tp_port = flow->tp_src ^ flow->tp_dst; } - } else { - fields.ip_addr = htonl(0); - fields.ip_proto = 0; - fields.tp_addr = htons(0); } return hash_bytes(&fields, sizeof fields, basis); } + +/* Hashes the portions of 'flow' designated by 'fields'. */ +uint32_t +flow_hash_fields(const struct flow *flow, enum nx_hash_fields fields, + uint16_t basis) +{ + switch (fields) { + + case NX_HASH_FIELDS_ETH_SRC: + return hash_bytes(flow->dl_src, sizeof flow->dl_src, basis); + + case NX_HASH_FIELDS_SYMMETRIC_L4: + return flow_hash_symmetric_l4(flow, basis); + } + + NOT_REACHED(); +} + +/* Returns a string representation of 'fields'. */ +const char * +flow_hash_fields_to_str(enum nx_hash_fields fields) +{ + switch (fields) { + case NX_HASH_FIELDS_ETH_SRC: return "eth_src"; + case NX_HASH_FIELDS_SYMMETRIC_L4: return "symmetric_l4"; + default: return ""; + } +} + +/* Returns true if the value of 'fields' is supported. Otherwise false. */ +bool +flow_hash_fields_valid(enum nx_hash_fields fields) +{ + return fields == NX_HASH_FIELDS_ETH_SRC + || fields == NX_HASH_FIELDS_SYMMETRIC_L4; +} + +/* Sets the VLAN VID that 'flow' matches to 'vid', which is interpreted as an + * OpenFlow 1.0 "dl_vlan" value: + * + * - If it is in the range 0...4095, 'flow->vlan_tci' is set to match + * that VLAN. Any existing PCP match is unchanged (it becomes 0 if + * 'flow' previously matched packets without a VLAN header). + * + * - If it is OFP_VLAN_NONE, 'flow->vlan_tci' is set to match a packet + * without a VLAN tag. + * + * - Other values of 'vid' should not be used. */ +void +flow_set_vlan_vid(struct flow *flow, ovs_be16 vid) +{ + if (vid == htons(OFP_VLAN_NONE)) { + flow->vlan_tci = htons(0); + } else { + vid &= htons(VLAN_VID_MASK); + flow->vlan_tci &= ~htons(VLAN_VID_MASK); + flow->vlan_tci |= htons(VLAN_CFI) | vid; + } +} + +/* Sets the VLAN PCP that 'flow' matches to 'pcp', which should be in the + * range 0...7. + * + * This function has no effect on the VLAN ID that 'flow' matches. + * + * After calling this function, 'flow' will not match packets without a VLAN + * header. */ +void +flow_set_vlan_pcp(struct flow *flow, uint8_t pcp) +{ + pcp &= 0x07; + flow->vlan_tci &= ~htons(VLAN_PCP_MASK); + flow->vlan_tci |= htons((pcp << VLAN_PCP_SHIFT) | VLAN_CFI); +} + +/* Puts into 'b' a packet that flow_extract() would parse as having the given + * 'flow'. + * + * (This is useful only for testing, obviously, and the packet isn't really + * valid. It hasn't got any checksums filled in, for one, and lots of fields + * are just zeroed.) */ +void +flow_compose(struct ofpbuf *b, const struct flow *flow) +{ + 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); + return; + } + + if (flow->vlan_tci & htons(VLAN_CFI)) { + eth_push_vlan(b, 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->ip_ihl_ver = IP_IHL_VER(5, 4); + ip->ip_tos = flow->nw_tos; + ip->ip_proto = flow->nw_proto; + ip->ip_src = flow->nw_src; + ip->ip_dst = flow->nw_dst; + + if (flow->nw_frag & FLOW_NW_FRAG_ANY) { + ip->ip_frag_off |= htons(IP_MORE_FRAGMENTS); + if (flow->nw_frag & FLOW_NW_FRAG_LATER) { + 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); + } + } + + ip->ip_tot_len = htons((uint8_t *) b->data + b->size + - (uint8_t *) b->l3); + } else if (flow->dl_type == htons(ETH_TYPE_IPV6)) { + /* XXX */ + } else if (flow->dl_type == htons(ETH_TYPE_ARP)) { + struct arp_eth_header *arp; + + b->l3 = arp = ofpbuf_put_zeros(b, sizeof *arp); + arp->ar_hrd = htons(1); + arp->ar_pro = htons(ETH_TYPE_IP); + arp->ar_hln = ETH_ADDR_LEN; + arp->ar_pln = 4; + arp->ar_op = htons(flow->nw_proto); + + 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; + memcpy(arp->ar_sha, flow->arp_sha, ETH_ADDR_LEN); + memcpy(arp->ar_tha, flow->arp_tha, ETH_ADDR_LEN); + } + } +}