X-Git-Url: http://git.onelab.eu/?a=blobdiff_plain;f=lib%2Fflow.c;h=a13519ef5fb7701fee85d23b31a75487dccf736d;hb=85b20fd6ee585f462e012fbcc7f966a81edab2ed;hp=754c0de0ba19cf8d9b1f12cf6101a677816baa22;hpb=19997ceee06cec2ce1f36ebfc225fa379b6b1ed0;p=sliver-openvswitch.git diff --git a/lib/flow.c b/lib/flow.c index 754c0de0b..a13519ef5 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. @@ -16,21 +16,24 @@ #include #include #include "flow.h" +#include #include #include +#include #include #include #include +#include #include #include #include "byte-order.h" #include "coverage.h" -#include "dpif.h" +#include "csum.h" #include "dynamic-string.h" #include "hash.h" +#include "match.h" #include "ofpbuf.h" #include "openflow/openflow.h" -#include "openvswitch/datapath-protocol.h" #include "packets.h" #include "unaligned.h" #include "vlog.h" @@ -38,6 +41,7 @@ VLOG_DEFINE_THIS_MODULE(flow); COVERAGE_DEFINE(flow_extract); +COVERAGE_DEFINE(miniflow_malloc); static struct arp_eth_header * pull_arp(struct ofpbuf *packet) @@ -149,7 +153,9 @@ parse_ipv6(struct ofpbuf *packet, struct flow *flow) flow->ipv6_dst = nh->ip6_dst; tc_flow = get_unaligned_be32(&nh->ip6_flow); - flow->nw_tos = (ntohl(tc_flow) >> 4) & IP_DSCP_MASK; + 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) { @@ -178,7 +184,7 @@ 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 = (struct ip6_ext *)packet->data; + const struct ip6_ext *ext_hdr = packet->data; nexthdr = ext_hdr->ip6e_nxt; if (!ofpbuf_try_pull(packet, (ext_hdr->ip6e_len + 1) * 8)) { return EINVAL; @@ -188,13 +194,13 @@ 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 = (struct ip6_ext *)packet->data; + const struct ip6_ext *ext_hdr = 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; + const struct ip6_frag *frag_hdr = packet->data; nexthdr = frag_hdr->ip6f_nxt; if (!ofpbuf_try_pull(packet, sizeof *frag_hdr)) { @@ -202,9 +208,14 @@ parse_ipv6(struct ofpbuf *packet, struct flow *flow) } /* We only process the first fragment. */ - if ((frag_hdr->ip6f_offlg & IP6F_OFF_MASK) != htons(0)) { - nexthdr = IPPROTO_FRAGMENT; - break; + 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; + } } } } @@ -246,8 +257,8 @@ parse_icmpv6(struct ofpbuf *b, struct flow *flow) /* 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->icmp_type = htons(icmp->icmp6_type); - flow->icmp_code = htons(icmp->icmp6_code); + 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 || @@ -306,7 +317,9 @@ invalid: } -/* Initializes 'flow' members from 'packet', 'tun_id', and 'ofp_in_port'. +/* Initializes 'flow' members from 'packet', 'skb_priority', 'tnl', and + * 'ofp_in_port'. + * * Initializes 'packet' header pointers as follows: * * - packet->l2 to the start of the Ethernet header. @@ -321,19 +334,25 @@ invalid: * - 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 ofp_in_port, +void +flow_extract(struct ofpbuf *packet, uint32_t skb_priority, uint32_t skb_mark, + const struct flow_tnl *tnl, 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; + + if (tnl) { + assert(tnl != &flow->tunnel); + flow->tunnel = *tnl; + } flow->in_port = ofp_in_port; + flow->skb_priority = skb_priority; + flow->skb_mark = skb_mark; packet->l2 = b.data; packet->l3 = NULL; @@ -341,7 +360,7 @@ flow_extract(struct ofpbuf *packet, ovs_be64 tun_id, uint16_t ofp_in_port, packet->l7 = NULL; if (b.size < sizeof *eth) { - return 0; + return; } /* Link layer. */ @@ -356,17 +375,51 @@ flow_extract(struct ofpbuf *packet, ovs_be64 tun_id, uint16_t ofp_in_port, } flow->dl_type = parse_ethertype(&b); - /* Network layer. */ packet->l3 = b.data; - if (flow->dl_type == htons(ETH_TYPE_IP)) { + flow_extract_l3_onwards(packet, flow, flow->dl_type); +} + +/* Initializes l3 and higher 'flow' members from 'packet' + * + * This should be called by or after flow_extract() + * + * Initializes 'packet' header pointers as follows: + * + * - 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. + */ +void +flow_extract_l3_onwards(struct ofpbuf *packet, struct flow *flow, + ovs_be16 dl_type) +{ + struct ofpbuf b; + + ofpbuf_use_const(&b, packet->l3, packet->size - + (size_t)((char *)packet->l3 - (char *)packet->l2)); + + /* Network layer. */ + if (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)) { + + 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) { @@ -374,20 +427,16 @@ flow_extract(struct ofpbuf *packet, ovs_be64 tun_id, uint16_t ofp_in_port, } 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)) { - - retval = parse_ipv6(&b, flow); - if (retval) { - return 0; + } else if (dl_type == htons(ETH_TYPE_IPV6)) { + if (parse_ipv6(&b, flow)) { + return; } packet->l4 = b.data; @@ -400,7 +449,8 @@ flow_extract(struct ofpbuf *packet, ovs_be64 tun_id, uint16_t ofp_in_port, packet->l7 = b.data; } } - } else if (flow->dl_type == htons(ETH_TYPE_ARP)) { + } else if (dl_type == htons(ETH_TYPE_ARP) || + 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) @@ -411,37 +461,38 @@ flow_extract(struct ofpbuf *packet, ovs_be64 tun_id, uint16_t ofp_in_port, 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 = 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)); + uint32_t *flow_u32 = (uint32_t *) flow; + const uint32_t *wc_u32 = (const uint32_t *) &wildcards->masks; + size_t i; - if ((flow->dl_type == htons(ETH_TYPE_IP)) && packet->l4) { - if ((flow->nw_proto == IPPROTO_TCP) && packet->l7) { - struct tcp_header *tcp = packet->l4; - stats->tcp_flags = TCP_FLAGS(tcp->tcp_ctl); - } + for (i = 0; i < FLOW_U32S; i++) { + flow_u32[i] &= wc_u32[i]; } +} - stats->n_bytes = packet->size; - stats->n_packets = 1; +/* 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 == 18); + + fmd->tun_id = flow->tunnel.tun_id; + fmd->metadata = flow->metadata; + memcpy(fmd->regs, flow->regs, sizeof fmd->regs); + fmd->in_port = flow->in_port; } char * @@ -452,49 +503,57 @@ flow_to_string(const struct flow *flow) return ds_cstr(&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 -flow_format(struct ds *ds, const struct flow *flow) +format_flags(struct ds *ds, const char *(*bit_to_string)(uint32_t), + uint32_t flags, char del) { - ds_put_format(ds, "tunnel%#"PRIx64":in_port%04"PRIx16":tci(", - ntohll(flow->tun_id), flow->in_port); - 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, " proto%"PRIu8" tos%"PRIu8" ipv6", - flow->nw_proto, flow->nw_tos); - print_ipv6_addr(ds, &flow->ipv6_src); - ds_put_cstr(ds, "->"); - print_ipv6_addr(ds, &flow->ipv6_dst); - - } else { - ds_put_format(ds, " proto%"PRIu8 - " tos%"PRIu8 - " ip"IP_FMT"->"IP_FMT, - flow->nw_proto, - flow->nw_tos, - IP_ARGS(&flow->nw_src), - IP_ARGS(&flow->nw_dst)); + uint32_t bad = 0; + + if (!flags) { + return; } - if (flow->tp_src || flow->tp_dst) { - ds_put_format(ds, " port%"PRIu16"->%"PRIu16, - ntohs(flow->tp_src), ntohs(flow->tp_dst)); + while (flags) { + uint32_t bit = rightmost_1bit(flags); + const char *s; + + s = bit_to_string(bit); + if (s) { + ds_put_format(ds, "%s%c", s, del); + } else { + bad |= bit; + } + + flags &= ~bit; } - 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)); + + 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) +{ + struct match match; + + match_wc_init(&match, flow); + match_format(&match, ds, OFP_DEFAULT_PRIORITY); } void @@ -511,15 +570,7 @@ flow_print(FILE *stream, const struct flow *flow) void flow_wildcards_init_catchall(struct flow_wildcards *wc) { - 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; - memset(wc->reg_masks, 0, sizeof wc->reg_masks); - wc->vlan_tci_mask = htons(0); - wc->zero = 0; + memset(&wc->masks, 0, sizeof wc->masks); } /* Initializes 'wc' as an exact-match set of wildcards; that is, 'wc' does not @@ -527,40 +578,23 @@ flow_wildcards_init_catchall(struct flow_wildcards *wc) void flow_wildcards_init_exact(struct flow_wildcards *wc) { - 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; - memset(wc->reg_masks, 0xff, sizeof wc->reg_masks); - wc->vlan_tci_mask = htons(UINT16_MAX); - wc->zero = 0; + memset(&wc->masks, 0xff, sizeof wc->masks); + memset(wc->masks.zeros, 0, sizeof wc->masks.zeros); } -/* Returns true if 'wc' is exact-match, false if 'wc' wildcards any bits or +/* Returns true if 'wc' matches every packet, false if 'wc' fixes any bits or * fields. */ bool -flow_wildcards_is_exact(const struct flow_wildcards *wc) +flow_wildcards_is_catchall(const struct flow_wildcards *wc) { - int i; + const uint32_t *wc_u32 = (const uint32_t *) &wc->masks; + size_t i; - 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) - || !ipv6_mask_is_exact(&wc->ipv6_src_mask) - || !ipv6_mask_is_exact(&wc->ipv6_dst_mask)) { - return false; - } - - for (i = 0; i < FLOW_N_REGS; i++) { - if (wc->reg_masks[i] != UINT32_MAX) { + for (i = 0; i < FLOW_U32S; i++) { + if (wc_u32[i]) { return false; } } - return true; } @@ -572,31 +606,21 @@ flow_wildcards_combine(struct flow_wildcards *dst, const struct flow_wildcards *src1, const struct flow_wildcards *src2) { - int i; + 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; - 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); - for (i = 0; i < FLOW_N_REGS; i++) { - dst->reg_masks[i] = src1->reg_masks[i] & src2->reg_masks[i]; + for (i = 0; i < FLOW_U32S; i++) { + dst_u32[i] = src1_u32[i] & src2_u32[i]; } - dst->vlan_tci_mask = src1->vlan_tci_mask & src2->vlan_tci_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 == 56 + FLOW_N_REGS * 4); - return hash_bytes(wc, sizeof *wc, 0); + return flow_hash(&wc->masks, basis); } /* Returns true if 'a' and 'b' represent the same wildcards, false if they are @@ -605,25 +629,7 @@ bool flow_wildcards_equal(const struct flow_wildcards *a, const struct flow_wildcards *b) { - int i; - - 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 - || !ipv6_addr_equals(&a->ipv6_src_mask, &b->ipv6_src_mask) - || !ipv6_addr_equals(&a->ipv6_dst_mask, &b->ipv6_dst_mask)) { - return false; - } - - for (i = 0; i < FLOW_N_REGS; i++) { - if (a->reg_masks[i] != b->reg_masks[i]) { - return false; - } - } - - return true; + return flow_equal(&a->masks, &b->masks); } /* Returns true if at least one bit or field is wildcarded in 'a' but not in @@ -632,90 +638,35 @@ bool flow_wildcards_has_extra(const struct flow_wildcards *a, const struct flow_wildcards *b) { - int i; - struct in6_addr ipv6_masked; + const uint32_t *a_u32 = (const uint32_t *) &a->masks; + const uint32_t *b_u32 = (const uint32_t *) &b->masks; + size_t i; - for (i = 0; i < FLOW_N_REGS; i++) { - if ((a->reg_masks[i] & b->reg_masks[i]) != b->reg_masks[i]) { + for (i = 0; i < FLOW_U32S; i++) { + if ((a_u32[i] & b_u32[i]) != b_u32[i]) { return true; } } - - 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; - } - - 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); -} - -static bool -set_nw_mask(ovs_be32 *maskp, ovs_be32 mask) -{ - if (ip_is_cidr(mask)) { - *maskp = mask; - return true; - } else { - return false; - } -} - -/* 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) -{ - return set_nw_mask(&wc->nw_src_mask, mask); + return false; } -/* 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 'a' and 'b' are equal, except that 0-bits (wildcarded bits) + * in 'wc' do not need to be equal in 'a' and 'b'. */ bool -flow_wildcards_set_nw_dst_mask(struct flow_wildcards *wc, ovs_be32 mask) +flow_equal_except(const struct flow *a, const struct flow *b, + const struct flow_wildcards *wc) { - return set_nw_mask(&wc->nw_dst_mask, mask); -} + const uint32_t *a_u32 = (const uint32_t *) a; + const uint32_t *b_u32 = (const uint32_t *) b; + const uint32_t *wc_u32 = (const uint32_t *) &wc->masks; + size_t i; -static bool -set_ipv6_mask(struct in6_addr *maskp, const struct in6_addr *mask) -{ - if (ipv6_is_cidr(mask)) { - *maskp = *mask; - return true; - } else { - return false; + for (i = 0; i < FLOW_U32S; i++) { + if ((a_u32[i] ^ b_u32[i]) & wc_u32[i]) { + return false; + } } -} - -/* Sets the IPv6 source wildcard mask to CIDR 'mask' (consisting of N - * high-order 1-bit and 128-N low-order 0-bits). Returns true if successful, - * false if 'mask' is not a CIDR mask. */ -bool -flow_wildcards_set_ipv6_src_mask(struct flow_wildcards *wc, - const struct in6_addr *mask) -{ - return set_ipv6_mask(&wc->ipv6_src_mask, mask); -} - -/* Sets the IPv6 destination wildcard mask to CIDR 'mask' (consisting of - * N high-order 1-bit and 128-N low-order 0-bits). Returns true if - * successful, false if 'mask' is not a CIDR mask. */ -bool -flow_wildcards_set_ipv6_dst_mask(struct flow_wildcards *wc, - const struct in6_addr *mask) -{ - return set_ipv6_mask(&wc->ipv6_dst_mask, mask); + return true; } /* Sets the wildcard mask for register 'idx' in 'wc' to 'mask'. @@ -723,7 +674,7 @@ flow_wildcards_set_ipv6_dst_mask(struct flow_wildcards *wc, void flow_wildcards_set_reg_mask(struct flow_wildcards *wc, int idx, uint32_t mask) { - wc->reg_masks[idx] = mask; + wc->masks.regs[idx] = mask; } /* Hashes 'flow' based on its L2 through L4 protocol information. */ @@ -737,7 +688,7 @@ flow_hash_symmetric_l4(const struct flow *flow, uint32_t basis) }; 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; @@ -757,7 +708,7 @@ flow_hash_symmetric_l4(const struct flow *flow, uint32_t basis) fields.ipv4_addr = flow->nw_src ^ flow->nw_dst; fields.ip_proto = flow->nw_proto; if (fields.ip_proto == IPPROTO_TCP) { - fields.tp_addr = flow->tp_src ^ flow->tp_dst; + 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]; @@ -769,8 +720,595 @@ flow_hash_symmetric_l4(const struct flow *flow, uint32_t basis) } fields.ip_proto = flow->nw_proto; if (fields.ip_proto == IPPROTO_TCP) { - fields.tp_addr = flow->tp_src ^ flow->tp_dst; + fields.tp_port = flow->tp_src ^ flow->tp_dst; } } 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_dl_vlan(struct flow *flow, ovs_be16 vid) +{ + if (vid == htons(OFP10_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 VID that 'flow' matches to 'vid', which is interpreted as an + * OpenFlow 1.2 "vlan_vid" value, that is, the low 13 bits of 'vlan_tci' (VID + * plus CFI). */ +void +flow_set_vlan_vid(struct flow *flow, ovs_be16 vid) +{ + ovs_be16 mask = htons(VLAN_VID_MASK | VLAN_CFI); + flow->vlan_tci &= ~mask; + flow->vlan_tci |= vid & mask; +} + +/* 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 some 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_ttl = flow->nw_ttl; + 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); + 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); + 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) || + flow->dl_type == htons(ETH_TYPE_RARP)) { + 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); + } + } +} + +/* Compressed 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; +} + +static uint32_t * +miniflow_alloc_values(struct miniflow *flow, int n) +{ + if (n <= MINI_N_INLINE) { + return flow->inline_values; + } else { + COVERAGE_INC(miniflow_malloc); + return xmalloc(n * sizeof *flow->values); + } +} + +/* 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); + for (i = 0; i < FLOW_U32S; i++) { + if (src_u32[i]) { + dst->map[i / 32] |= 1u << (i % 32); + n++; + } + } + + /* Initialize dst->values. */ + dst->values = miniflow_alloc_values(dst, n); + ofs = 0; + for (i = 0; i < MINI_N_MAPS; i++) { + uint32_t map; + + for (map = dst->map[i]; map; map = zero_rightmost_1bit(map)) { + dst->values[ofs++] = src_u32[raw_ctz(map) + i * 32]; + } + } +} + +/* Initializes 'dst' as a copy of 'src'. The caller must eventually free 'dst' + * with miniflow_destroy(). */ +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->values = miniflow_alloc_values(dst, n); + memcpy(dst->values, src->values, n * sizeof *dst->values); +} + +/* Frees any memory owned by 'flow'. Does not free the storage in which 'flow' + * itself resides; the caller is responsible for that. */ +void +miniflow_destroy(struct miniflow *flow) +{ + if (flow->values != flow->inline_values) { + free(flow->values); + } +} + +/* Initializes 'dst' as a copy of 'src'. */ +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++]; + } + } +} + +static const uint32_t * +miniflow_get__(const struct miniflow *flow, unsigned int u32_ofs) +{ + if (!(flow->map[u32_ofs / 32] & (1u << (u32_ofs % 32)))) { + 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; + } +} + +/* Returns the uint32_t that would be at byte offset '4 * u32_ofs' if 'flow' + * were expanded into a "struct flow". */ +uint32_t +miniflow_get(const struct miniflow *flow, unsigned int u32_ofs) +{ + return *miniflow_get__(flow, u32_ofs); +} + +/* Returns the ovs_be16 that would be at byte offset 'u8_ofs' if 'flow' were + * expanded into a "struct flow". */ +static ovs_be16 +miniflow_get_be16(const struct miniflow *flow, unsigned int u8_ofs) +{ + const uint32_t *u32p = miniflow_get__(flow, u8_ofs / 4); + const ovs_be16 *be16p = (const ovs_be16 *) u32p; + return be16p[u8_ofs % 4 != 0]; +} + +/* Returns the VID within the vlan_tci member of the "struct flow" represented + * by 'flow'. */ +uint16_t +miniflow_get_vid(const struct miniflow *flow) +{ + ovs_be16 tci = miniflow_get_be16(flow, offsetof(struct flow, vlan_tci)); + return vlan_tci_to_vid(tci); +} + +/* Returns true if 'a' and 'b' are the same flow, false otherwise. */ +bool +miniflow_equal(const struct miniflow *a, const struct miniflow *b) +{ + int i; + + for (i = 0; i < MINI_N_MAPS; i++) { + if (a->map[i] != b->map[i]) { + return false; + } + } + + return !memcmp(a->values, b->values, + miniflow_n_values(a) * sizeof *a->values); +} + +/* Returns true if 'a' and 'b' are equal at the places where there are 1-bits + * in 'mask', false if they differ. */ +bool +miniflow_equal_in_minimask(const struct miniflow *a, const struct miniflow *b, + const struct minimask *mask) +{ + const uint32_t *p; + int i; + + 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; + + if ((miniflow_get(a, ofs) ^ miniflow_get(b, ofs)) & *p) { + return false; + } + p++; + } + } + + return true; +} + +/* Returns true if 'a' and 'b' are equal at the places where there are 1-bits + * in 'mask', false if they differ. */ +bool +miniflow_equal_flow_in_minimask(const struct miniflow *a, const struct flow *b, + const struct minimask *mask) +{ + const uint32_t *b_u32 = (const uint32_t *) b; + const uint32_t *p; + int i; + + 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; + + if ((miniflow_get(a, ofs) ^ b_u32[ofs]) & *p) { + return false; + } + p++; + } + } + + return true; +} + +/* Returns a hash value for 'flow', given 'basis'. */ +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)); +} + +/* Returns a hash value for the bits of 'flow' where there are 1-bits in + * 'mask', given 'basis'. + * + * The hash values returned by this function are the same as those returned by + * flow_hash_in_minimask(), only the form of the arguments differ. */ +uint32_t +miniflow_hash_in_minimask(const struct miniflow *flow, + const struct minimask *mask, uint32_t basis) +{ + const uint32_t *p = mask->masks.values; + uint32_t hash; + int i; + + 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++; + } + } + + return mhash_finish(hash, p - mask->masks.values); +} + +/* Returns a hash value for the bits of 'flow' where there are 1-bits in + * 'mask', given 'basis'. + * + * The hash values returned by this function are the same as those returned by + * miniflow_hash_in_minimask(), only the form of the arguments differ. */ +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 *p = mask->masks.values; + uint32_t hash; + int i; + + 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, flow_u32[ofs] & *p); + p++; + } + } + + return mhash_finish(hash, p - mask->masks.values); +} + +/* Initializes 'dst' as a copy of 'src'. The caller must eventually free 'dst' + * with minimask_destroy(). */ +void +minimask_init(struct minimask *mask, const struct flow_wildcards *wc) +{ + miniflow_init(&mask->masks, &wc->masks); +} + +/* Initializes 'dst' as a copy of 'src'. The caller must eventually free 'dst' + * with minimask_destroy(). */ +void +minimask_clone(struct minimask *dst, const struct minimask *src) +{ + miniflow_clone(&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 + * by 'dst_'. The caller must *not* free 'dst_' with minimask_destroy(). */ +void +minimask_combine(struct minimask *dst_, + const struct minimask *a_, const struct minimask *b_, + uint32_t storage[FLOW_U32S]) +{ + struct miniflow *dst = &dst_->masks; + const struct miniflow *a = &a_->masks; + const struct miniflow *b = &b_->masks; + int i, n; + + 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; + } + } + } +} + +/* Frees any memory owned by 'mask'. Does not free the storage in which 'mask' + * itself resides; the caller is responsible for that. */ +void +minimask_destroy(struct minimask *mask) +{ + miniflow_destroy(&mask->masks); +} + +/* Initializes 'dst' as a copy of 'src'. */ +void +minimask_expand(const struct minimask *mask, struct flow_wildcards *wc) +{ + miniflow_expand(&mask->masks, &wc->masks); +} + +/* Returns the uint32_t that would be at byte offset '4 * u32_ofs' if 'mask' + * were expanded into a "struct flow_wildcards". */ +uint32_t +minimask_get(const struct minimask *mask, unsigned int u32_ofs) +{ + return miniflow_get(&mask->masks, u32_ofs); +} + +/* Returns the VID mask within the vlan_tci member of the "struct + * flow_wildcards" represented by 'mask'. */ +uint16_t +minimask_get_vid_mask(const struct minimask *mask) +{ + return miniflow_get_vid(&mask->masks); +} + +/* Returns true if 'a' and 'b' are the same flow mask, false otherwise. */ +bool +minimask_equal(const struct minimask *a, const struct minimask *b) +{ + return miniflow_equal(&a->masks, &b->masks); +} + +/* Returns a hash value for 'mask', given 'basis'. */ +uint32_t +minimask_hash(const struct minimask *mask, uint32_t basis) +{ + return miniflow_hash(&mask->masks, basis); +} + +/* Returns true if at least one bit is wildcarded in 'a_' but not in 'b_', + * false otherwise. */ +bool +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; + + 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; + } + } + } + + return false; +} + +/* Returns true if 'mask' matches every packet, false if 'mask' fixes any bits + * or fields. */ +bool +minimask_is_catchall(const struct minimask *mask_) +{ + const struct miniflow *mask = &mask_->masks; + + BUILD_ASSERT(MINI_N_MAPS == 2); + return !(mask->map[0] | mask->map[1]); +}