X-Git-Url: http://git.onelab.eu/?a=blobdiff_plain;f=lib%2Fflow.c;h=211203105218b279fb9e355e46203cdd05522414;hb=d4570fd8ba5f36f21b9b631628e812de0189fa20;hp=137ce696d8374fac5738a2958f274d5801f025b1;hpb=080e28d0f0025c705011ce69857fa30d667a5524;p=sliver-openvswitch.git diff --git a/lib/flow.c b/lib/flow.c index 137ce696d..211203105 100644 --- a/lib/flow.c +++ b/lib/flow.c @@ -1,5 +1,5 @@ /* - * Copyright (c) 2008, 2009, 2010, 2011, 2012, 2013 Nicira, Inc. + * Copyright (c) 2008, 2009, 2010, 2011, 2012, 2013, 2014 Nicira, Inc. * * Licensed under the Apache License, Version 2.0 (the "License"); * you may not use this file except in compliance with the License. @@ -35,286 +35,260 @@ #include "ofpbuf.h" #include "openflow/openflow.h" #include "packets.h" +#include "odp-util.h" #include "random.h" #include "unaligned.h" COVERAGE_DEFINE(flow_extract); COVERAGE_DEFINE(miniflow_malloc); -static struct arp_eth_header * -pull_arp(struct ofpbuf *packet) -{ - return ofpbuf_try_pull(packet, ARP_ETH_HEADER_LEN); -} - -static struct ip_header * -pull_ip(struct ofpbuf *packet) -{ - if (packet->size >= IP_HEADER_LEN) { - struct ip_header *ip = packet->data; - int ip_len = IP_IHL(ip->ip_ihl_ver) * 4; - if (ip_len >= IP_HEADER_LEN && packet->size >= ip_len) { - return ofpbuf_pull(packet, ip_len); - } +/* 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 +}; + +/* miniflow_extract() assumes the following to be true to optimize the + * extraction process. */ +BUILD_ASSERT_DECL(offsetof(struct flow, dl_type) + 2 + == offsetof(struct flow, vlan_tci) && + offsetof(struct flow, dl_type) / 4 + == offsetof(struct flow, vlan_tci) / 4 ); + +BUILD_ASSERT_DECL(offsetof(struct flow, nw_frag) + 3 + == offsetof(struct flow, nw_proto) && + offsetof(struct flow, nw_tos) + 2 + == offsetof(struct flow, nw_proto) && + offsetof(struct flow, nw_ttl) + 1 + == offsetof(struct flow, nw_proto) && + offsetof(struct flow, nw_frag) / 4 + == offsetof(struct flow, nw_tos) / 4 && + offsetof(struct flow, nw_ttl) / 4 + == offsetof(struct flow, nw_tos) / 4 && + offsetof(struct flow, nw_proto) / 4 + == offsetof(struct flow, nw_tos) / 4); + +/* TCP flags in the first half of a BE32, zeroes in the other half. */ +BUILD_ASSERT_DECL(offsetof(struct flow, tcp_flags) + 2 + == offsetof(struct flow, pad) && + offsetof(struct flow, tcp_flags) / 4 + == offsetof(struct flow, pad) / 4); +#if WORDS_BIGENDIAN +#define TCP_FLAGS_BE32(tcp_ctl) ((OVS_FORCE ovs_be32)TCP_FLAGS_BE16(tcp_ctl) \ + << 16) +#else +#define TCP_FLAGS_BE32(tcp_ctl) ((OVS_FORCE ovs_be32)TCP_FLAGS_BE16(tcp_ctl)) +#endif + +BUILD_ASSERT_DECL(offsetof(struct flow, tp_src) + 2 + == offsetof(struct flow, tp_dst) && + offsetof(struct flow, tp_src) / 4 + == offsetof(struct flow, tp_dst) / 4); + +/* Removes 'size' bytes from the head end of '*datap', of size '*sizep', which + * must contain at least 'size' bytes of data. Returns the first byte of data + * removed. */ +static inline const void * +data_pull(void **datap, size_t *sizep, size_t size) +{ + char *data = (char *)*datap; + *datap = data + size; + *sizep -= size; + return data; +} + +/* If '*datap' has at least 'size' bytes of data, removes that many bytes from + * the head end of '*datap' and returns the first byte removed. Otherwise, + * returns a null pointer without modifying '*datap'. */ +static inline const void * +data_try_pull(void **datap, size_t *sizep, size_t size) +{ + return OVS_LIKELY(*sizep >= size) ? data_pull(datap, sizep, size) : NULL; +} + +/* Context for pushing data to a miniflow. */ +struct mf_ctx { + uint64_t map; + uint32_t *data; + uint32_t * const end; +}; + +/* miniflow_push_* macros allow filling in a miniflow data values in order. + * Assertions are needed only when the layout of the struct flow is modified. + * 'ofs' is a compile-time constant, which allows most of the code be optimized + * away. Some GCC versions gave warnigns on ALWAYS_INLINE, so these are + * defined as macros. */ + +#if (FLOW_WC_SEQ != 26) +#define MINIFLOW_ASSERT(X) ovs_assert(X) +#else +#define MINIFLOW_ASSERT(X) +#endif + +#define miniflow_push_uint32_(MF, OFS, VALUE) \ +{ \ + MINIFLOW_ASSERT(MF.data < MF.end && (OFS) % 4 == 0 \ + && !(MF.map & (UINT64_MAX << (OFS) / 4))); \ + *MF.data++ = VALUE; \ + MF.map |= UINT64_C(1) << (OFS) / 4; \ +} + +#define miniflow_push_be32_(MF, OFS, VALUE) \ + miniflow_push_uint32_(MF, OFS, (OVS_FORCE uint32_t)(VALUE)) + +#define miniflow_push_uint16_(MF, OFS, VALUE) \ +{ \ + MINIFLOW_ASSERT(MF.data < MF.end && \ + (((OFS) % 4 == 0 && !(MF.map & (UINT64_MAX << (OFS) / 4))) \ + || ((OFS) % 4 == 2 && MF.map & (UINT64_C(1) << (OFS) / 4) \ + && !(MF.map & (UINT64_MAX << ((OFS) / 4 + 1)))))); \ + \ + if ((OFS) % 4 == 0) { \ + *(uint16_t *)MF.data = VALUE; \ + MF.map |= UINT64_C(1) << (OFS) / 4; \ + } else if ((OFS) % 4 == 2) { \ + *((uint16_t *)MF.data + 1) = VALUE; \ + MF.data++; \ + } \ +} + +#define miniflow_push_be16_(MF, OFS, VALUE) \ + miniflow_push_uint16_(MF, OFS, (OVS_FORCE uint16_t)VALUE); + +/* Data at 'valuep' may be unaligned. */ +#define miniflow_push_words_(MF, OFS, VALUEP, N_WORDS) \ +{ \ + int ofs32 = (OFS) / 4; \ + \ + MINIFLOW_ASSERT(MF.data + (N_WORDS) <= MF.end && (OFS) % 4 == 0 \ + && !(MF.map & (UINT64_MAX << ofs32))); \ + \ + memcpy(MF.data, (VALUEP), (N_WORDS) * sizeof *MF.data); \ + MF.data += (N_WORDS); \ + MF.map |= ((UINT64_MAX >> (64 - (N_WORDS))) << ofs32); \ +} + +#define miniflow_push_uint32(MF, FIELD, VALUE) \ + miniflow_push_uint32_(MF, offsetof(struct flow, FIELD), VALUE) + +#define miniflow_push_be32(MF, FIELD, VALUE) \ + miniflow_push_be32_(MF, offsetof(struct flow, FIELD), VALUE) + +#define miniflow_push_uint32_check(MF, FIELD, VALUE) \ + { if (OVS_LIKELY(VALUE)) { \ + miniflow_push_uint32_(MF, offsetof(struct flow, FIELD), VALUE); \ + } \ } - return NULL; -} -static struct tcp_header * -pull_tcp(struct ofpbuf *packet) -{ - if (packet->size >= TCP_HEADER_LEN) { - struct tcp_header *tcp = packet->data; - int tcp_len = TCP_OFFSET(tcp->tcp_ctl) * 4; - if (tcp_len >= TCP_HEADER_LEN && packet->size >= tcp_len) { - return ofpbuf_pull(packet, tcp_len); - } +#define miniflow_push_be32_check(MF, FIELD, VALUE) \ + { if (OVS_LIKELY(VALUE)) { \ + miniflow_push_be32_(MF, offsetof(struct flow, FIELD), VALUE); \ + } \ } - return NULL; -} -static struct udp_header * -pull_udp(struct ofpbuf *packet) -{ - return ofpbuf_try_pull(packet, UDP_HEADER_LEN); -} +#define miniflow_push_uint16(MF, FIELD, VALUE) \ + miniflow_push_uint16_(MF, offsetof(struct flow, FIELD), VALUE) -static struct sctp_header * -pull_sctp(struct ofpbuf *packet) -{ - return ofpbuf_try_pull(packet, SCTP_HEADER_LEN); -} +#define miniflow_push_be16(MF, FIELD, VALUE) \ + miniflow_push_be16_(MF, offsetof(struct flow, FIELD), VALUE) -static struct icmp_header * -pull_icmp(struct ofpbuf *packet) -{ - return ofpbuf_try_pull(packet, ICMP_HEADER_LEN); -} +#define miniflow_push_words(MF, FIELD, VALUEP, N_WORDS) \ + miniflow_push_words_(MF, offsetof(struct flow, FIELD), VALUEP, N_WORDS) -static struct icmp6_hdr * -pull_icmpv6(struct ofpbuf *packet) +/* Pulls the MPLS headers at '*datap' and returns the count of them. */ +static inline int +parse_mpls(void **datap, size_t *sizep) { - return ofpbuf_try_pull(packet, sizeof(struct icmp6_hdr)); -} + const struct mpls_hdr *mh; + int count = 0; -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)) { + while ((mh = data_try_pull(datap, sizep, sizeof *mh))) { + count++; + if (mh->mpls_lse.lo & htons(1 << MPLS_BOS_SHIFT)) { break; } } + return MAX(count, FLOW_MAX_MPLS_LABELS); } -static void -parse_vlan(struct ofpbuf *b, struct flow *flow) +static inline ovs_be16 +parse_vlan(void **datap, size_t *sizep) { + const struct eth_header *eth = *datap; + struct qtag_prefix { ovs_be16 eth_type; /* ETH_TYPE_VLAN */ ovs_be16 tci; }; - if (b->size >= sizeof(struct qtag_prefix) + sizeof(ovs_be16)) { - struct qtag_prefix *qp = ofpbuf_pull(b, sizeof *qp); - flow->vlan_tci = qp->tci | htons(VLAN_CFI); + data_pull(datap, sizep, ETH_ADDR_LEN * 2); + + if (eth->eth_type == htons(ETH_TYPE_VLAN)) { + if (OVS_LIKELY(*sizep + >= sizeof(struct qtag_prefix) + sizeof(ovs_be16))) { + const struct qtag_prefix *qp = data_pull(datap, sizep, sizeof *qp); + return qp->tci | htons(VLAN_CFI); + } } + return 0; } -static ovs_be16 -parse_ethertype(struct ofpbuf *b) +static inline ovs_be16 +parse_ethertype(void **datap, size_t *sizep) { - struct llc_snap_header *llc; + const struct llc_snap_header *llc; ovs_be16 proto; - proto = *(ovs_be16 *) ofpbuf_pull(b, sizeof proto); - if (ntohs(proto) >= ETH_TYPE_MIN) { + proto = *(ovs_be16 *) data_pull(datap, sizep, sizeof proto); + if (OVS_LIKELY(ntohs(proto) >= ETH_TYPE_MIN)) { return proto; } - if (b->size < sizeof *llc) { + if (OVS_UNLIKELY(*sizep < sizeof *llc)) { return htons(FLOW_DL_TYPE_NONE); } - llc = b->data; - if (llc->llc.llc_dsap != LLC_DSAP_SNAP - || llc->llc.llc_ssap != LLC_SSAP_SNAP - || llc->llc.llc_cntl != LLC_CNTL_SNAP - || memcmp(llc->snap.snap_org, SNAP_ORG_ETHERNET, - sizeof llc->snap.snap_org)) { + llc = *datap; + if (OVS_UNLIKELY(llc->llc.llc_dsap != LLC_DSAP_SNAP + || llc->llc.llc_ssap != LLC_SSAP_SNAP + || llc->llc.llc_cntl != LLC_CNTL_SNAP + || memcmp(llc->snap.snap_org, SNAP_ORG_ETHERNET, + sizeof llc->snap.snap_org))) { return htons(FLOW_DL_TYPE_NONE); } - ofpbuf_pull(b, sizeof *llc); + data_pull(datap, sizep, sizeof *llc); - if (ntohs(llc->snap.snap_type) >= ETH_TYPE_MIN) { + if (OVS_LIKELY(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 ovs_16aligned_ip6_hdr *nh; - ovs_be32 tc_flow; - int nexthdr; - - nh = ofpbuf_try_pull(packet, sizeof *nh); - if (!nh) { - return EINVAL; - } - - nexthdr = nh->ip6_nxt; - - 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_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; - 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 = 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 = 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 ovs_16aligned_ip6_frag *frag_hdr = 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)) { - 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; - } - } - } - } - - 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; - flow->tcp_flags = tcp->tcp_ctl & htons(0x0fff); - packet->l7 = b->data; - } -} - -static void -parse_udp(struct ofpbuf *packet, struct ofpbuf *b, struct flow *flow) +static inline bool +parse_icmpv6(void **datap, size_t *sizep, const struct icmp6_hdr *icmp, + const struct in6_addr **nd_target, + uint8_t arp_buf[2][ETH_ADDR_LEN]) { - 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 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) -{ - 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) { + *nd_target = data_try_pull(datap, sizep, sizeof *nd_target); + if (OVS_UNLIKELY(!*nd_target)) { return false; } - flow->nd_target = *nd_target; - while (b->size >= 8) { + while (*sizep >= 8) { /* The minimum size of an option is 8 bytes, which also is * the size of Ethernet link-layer options. */ - const struct nd_opt_hdr *nd_opt = b->data; + const struct nd_opt_hdr *nd_opt = *datap; int opt_len = nd_opt->nd_opt_len * 8; - if (!opt_len || opt_len > b->size) { + if (!opt_len || opt_len > *sizep) { goto invalid; } @@ -323,21 +297,21 @@ parse_icmpv6(struct ofpbuf *b, struct flow *flow) * 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); + if (OVS_LIKELY(eth_addr_is_zero(arp_buf[0]))) { + memcpy(arp_buf[0], 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); + if (OVS_LIKELY(eth_addr_is_zero(arp_buf[1]))) { + memcpy(arp_buf[1], nd_opt + 1, ETH_ADDR_LEN); } else { goto invalid; } } - if (!ofpbuf_try_pull(b, opt_len)) { + if (OVS_UNLIKELY(!data_try_pull(datap, sizep, opt_len))) { goto invalid; } } @@ -346,155 +320,307 @@ parse_icmpv6(struct ofpbuf *b, struct flow *flow) 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', 'tnl', and - * 'in_port'. - * - * Initializes 'packet' header pointers as follows: +/* Initializes 'flow' members from 'packet' and 'md' * - * - packet->l2 to the start of the Ethernet header. + * Initializes 'packet' header l2 pointer to the start of the Ethernet + * header, and the layer offsets as follows: * - * - packet->l2_5 to the start of the MPLS shim header. + * - packet->l2_5_ofs to the start of the MPLS shim header, or UINT16_MAX + * when there is no MPLS shim header. * - * - packet->l3 to just past the Ethernet header, or just past the + * - packet->l3_ofs to just past the Ethernet header, or just past the * vlan_header if one is present, to the first byte of the payload of the - * Ethernet frame. + * Ethernet frame. UINT16_MAX if the frame is too short to contain an + * Ethernet header. * - * - packet->l4 to just past the IPv4 header, if one is present and has a - * correct length, and otherwise NULL. - * - * - packet->l7 to just past the TCP/UDP/SCTP/ICMP header, if one is - * present and has a correct length, and otherwise NULL. + * - packet->l4_ofs to just past the IPv4 header, if one is present and + * has at least the content used for the fields of interest for the flow, + * otherwise UINT16_MAX. */ void -flow_extract(struct ofpbuf *packet, uint32_t skb_priority, uint32_t pkt_mark, - const struct flow_tnl *tnl, const union flow_in_port *in_port, +flow_extract(struct ofpbuf *packet, const struct pkt_metadata *md, struct flow *flow) { - struct ofpbuf b = *packet; - struct eth_header *eth; + uint32_t buf[FLOW_U32S]; + struct miniflow mf; COVERAGE_INC(flow_extract); - memset(flow, 0, sizeof *flow); + miniflow_initialize(&mf, buf); + miniflow_extract(packet, md, &mf); + miniflow_expand(&mf, flow); +} - if (tnl) { - ovs_assert(tnl != &flow->tunnel); - flow->tunnel = *tnl; - } - if (in_port) { - flow->in_port = *in_port; +/* Caller is responsible for initializing 'dst->values' with enough storage + * for FLOW_U32S * 4 bytes. */ +void +miniflow_extract(struct ofpbuf *packet, const struct pkt_metadata *md, + struct miniflow *dst) +{ + void *data = ofpbuf_data(packet); + size_t size = ofpbuf_size(packet); + char *l2; + struct mf_ctx mf = { 0, dst->values, dst->values + FLOW_U32S }; + ovs_be16 dl_type; + uint8_t nw_frag, nw_tos, nw_ttl, nw_proto; + + /* Metadata. */ + if (md) { + if (md->tunnel.ip_dst) { + miniflow_push_words(mf, tunnel, &md->tunnel, + sizeof md->tunnel / 4); + } + miniflow_push_uint32_check(mf, skb_priority, md->skb_priority); + miniflow_push_uint32_check(mf, pkt_mark, md->pkt_mark); + miniflow_push_uint32_check(mf, recirc_id, md->recirc_id); + miniflow_push_uint32(mf, in_port, odp_to_u32(md->in_port.odp_port)); } - flow->skb_priority = skb_priority; - flow->pkt_mark = pkt_mark; - packet->l2 = b.data; - packet->l2_5 = NULL; - packet->l3 = NULL; - packet->l4 = NULL; - packet->l7 = NULL; - - if (b.size < sizeof *eth) { - return; - } + /* Initialize packet's layer pointer and offsets. */ + l2 = data; + ofpbuf_set_frame(packet, data); - /* Link layer. */ - eth = b.data; - memcpy(flow->dl_src, eth->eth_src, ETH_ADDR_LEN); - memcpy(flow->dl_dst, eth->eth_dst, ETH_ADDR_LEN); + /* Must have full Ethernet header to proceed. */ + if (OVS_UNLIKELY(size < sizeof(struct eth_header))) { + goto out; + } else { + ovs_be16 vlan_tci; - /* dl_type, vlan_tci. */ - ofpbuf_pull(&b, ETH_ADDR_LEN * 2); - if (eth->eth_type == htons(ETH_TYPE_VLAN)) { - parse_vlan(&b, flow); + /* Link layer. */ + BUILD_ASSERT(offsetof(struct flow, dl_dst) + 6 + == offsetof(struct flow, dl_src)); + miniflow_push_words(mf, dl_dst, data, ETH_ADDR_LEN * 2 / 4); + /* dl_type, vlan_tci. */ + vlan_tci = parse_vlan(&data, &size); + dl_type = parse_ethertype(&data, &size); + miniflow_push_be16(mf, dl_type, dl_type); + miniflow_push_be16(mf, vlan_tci, vlan_tci); } - 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); + /* Parse mpls. */ + if (OVS_UNLIKELY(eth_type_mpls(dl_type))) { + int count; + const void *mpls = data; + + packet->l2_5_ofs = (char *)data - l2; + count = parse_mpls(&data, &size); + miniflow_push_words(mf, mpls_lse, mpls, count); } /* Network layer. */ - packet->l3 = b.data; - 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_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; - 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; - } + packet->l3_ofs = (char *)data - l2; + + nw_frag = 0; + if (OVS_LIKELY(dl_type == htons(ETH_TYPE_IP))) { + const struct ip_header *nh = data; + int ip_len; + + if (OVS_UNLIKELY(size < IP_HEADER_LEN)) { + goto out; + } + ip_len = IP_IHL(nh->ip_ihl_ver) * 4; + + if (OVS_UNLIKELY(ip_len < IP_HEADER_LEN)) { + goto out; + } + + /* Push both source and destination address at once. */ + miniflow_push_words(mf, nw_src, &nh->ip_src, 2); + + nw_tos = nh->ip_tos; + nw_ttl = nh->ip_ttl; + nw_proto = nh->ip_proto; + if (OVS_UNLIKELY(IP_IS_FRAGMENT(nh->ip_frag_off))) { + nw_frag = FLOW_NW_FRAG_ANY; + if (nh->ip_frag_off & htons(IP_FRAG_OFF_MASK)) { + 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_SCTP) { - parse_sctp(packet, &b, flow); - } else if (flow->nw_proto == IPPROTO_ICMP) { - const struct icmp_header *icmp = pull_icmp(&b); - if (icmp) { - flow->tp_src = htons(icmp->icmp_type); - flow->tp_dst = htons(icmp->icmp_code); - packet->l7 = b.data; + } + if (OVS_UNLIKELY(size < ip_len)) { + goto out; + } + data_pull(&data, &size, ip_len); + + } else if (dl_type == htons(ETH_TYPE_IPV6)) { + const struct ovs_16aligned_ip6_hdr *nh; + ovs_be32 tc_flow; + + if (OVS_UNLIKELY(size < sizeof *nh)) { + goto out; + } + nh = data_pull(&data, &size, sizeof *nh); + + miniflow_push_words(mf, ipv6_src, &nh->ip6_src, + sizeof nh->ip6_src / 4); + miniflow_push_words(mf, ipv6_dst, &nh->ip6_dst, + sizeof nh->ip6_dst / 4); + + tc_flow = get_16aligned_be32(&nh->ip6_flow); + { + ovs_be32 label = tc_flow & htonl(IPV6_LABEL_MASK); + miniflow_push_be32_check(mf, ipv6_label, label); + } + + nw_tos = ntohl(tc_flow) >> 20; + nw_ttl = nh->ip6_hlim; + nw_proto = nh->ip6_nxt; + + while (1) { + if (OVS_LIKELY((nw_proto != IPPROTO_HOPOPTS) + && (nw_proto != IPPROTO_ROUTING) + && (nw_proto != IPPROTO_DSTOPTS) + && (nw_proto != IPPROTO_AH) + && (nw_proto != 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 (OVS_UNLIKELY(size < 8)) { + goto out; + } + + if ((nw_proto == IPPROTO_HOPOPTS) + || (nw_proto == IPPROTO_ROUTING) + || (nw_proto == 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 = data; + nw_proto = ext_hdr->ip6e_nxt; + if (OVS_UNLIKELY(!data_try_pull(&data, &size, + (ext_hdr->ip6e_len + 1) * 8))) { + goto out; + } + } else if (nw_proto == 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 = data; + nw_proto = ext_hdr->ip6e_nxt; + if (OVS_UNLIKELY(!data_try_pull(&data, &size, + (ext_hdr->ip6e_len + 2) * 4))) { + goto out; + } + } else if (nw_proto == IPPROTO_FRAGMENT) { + const struct ovs_16aligned_ip6_frag *frag_hdr = data; + + nw_proto = frag_hdr->ip6f_nxt; + if (!data_try_pull(&data, &size, sizeof *frag_hdr)) { + goto out; + } + + /* We only process the first fragment. */ + if (frag_hdr->ip6f_offlg != htons(0)) { + nw_frag = FLOW_NW_FRAG_ANY; + if ((frag_hdr->ip6f_offlg & IP6F_OFF_MASK) != htons(0)) { + nw_frag |= FLOW_NW_FRAG_LATER; + nw_proto = IPPROTO_FRAGMENT; + break; } } } } - } else if (flow->dl_type == htons(ETH_TYPE_IPV6)) { - if (parse_ipv6(&b, flow)) { - return; - } + } else { + if (dl_type == htons(ETH_TYPE_ARP) || + dl_type == htons(ETH_TYPE_RARP)) { + uint8_t arp_buf[2][ETH_ADDR_LEN]; + const struct arp_eth_header *arp = (const struct arp_eth_header *) + data_try_pull(&data, &size, ARP_ETH_HEADER_LEN); + + if (OVS_LIKELY(arp) && OVS_LIKELY(arp->ar_hrd == htons(1)) + && OVS_LIKELY(arp->ar_pro == htons(ETH_TYPE_IP)) + && OVS_LIKELY(arp->ar_hln == ETH_ADDR_LEN) + && OVS_LIKELY(arp->ar_pln == 4)) { + miniflow_push_words(mf, nw_src, &arp->ar_spa, 1); + miniflow_push_words(mf, nw_dst, &arp->ar_tpa, 1); + + /* We only match on the lower 8 bits of the opcode. */ + if (OVS_LIKELY(ntohs(arp->ar_op) <= 0xff)) { + miniflow_push_be32(mf, nw_frag, htonl(ntohs(arp->ar_op))); + } - 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_SCTP) { - parse_sctp(packet, &b, flow); - } else if (flow->nw_proto == IPPROTO_ICMPV6) { - if (parse_icmpv6(&b, flow)) { - packet->l7 = b.data; + /* Must be adjacent. */ + BUILD_ASSERT(offsetof(struct flow, arp_sha) + 6 + == offsetof(struct flow, arp_tha)); + + memcpy(arp_buf[0], arp->ar_sha, ETH_ADDR_LEN); + memcpy(arp_buf[1], arp->ar_tha, ETH_ADDR_LEN); + miniflow_push_words(mf, arp_sha, arp_buf, + ETH_ADDR_LEN * 2 / 4); } } - } 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) - && arp->ar_hln == ETH_ADDR_LEN - && arp->ar_pln == 4) { - /* We only match on the lower 8 bits of the opcode. */ - if (ntohs(arp->ar_op) <= 0xff) { - flow->nw_proto = ntohs(arp->ar_op); + goto out; + } + + packet->l4_ofs = (char *)data - l2; + miniflow_push_be32(mf, nw_frag, + BYTES_TO_BE32(nw_frag, nw_tos, nw_ttl, nw_proto)); + + if (OVS_LIKELY(!(nw_frag & FLOW_NW_FRAG_LATER))) { + if (OVS_LIKELY(nw_proto == IPPROTO_TCP)) { + if (OVS_LIKELY(size >= TCP_HEADER_LEN)) { + const struct tcp_header *tcp = data; + + miniflow_push_be32(mf, tcp_flags, + TCP_FLAGS_BE32(tcp->tcp_ctl)); + miniflow_push_words(mf, tp_src, &tcp->tcp_src, 1); + } + } else if (OVS_LIKELY(nw_proto == IPPROTO_UDP)) { + if (OVS_LIKELY(size >= UDP_HEADER_LEN)) { + const struct udp_header *udp = data; + + miniflow_push_words(mf, tp_src, &udp->udp_src, 1); } + } else if (OVS_LIKELY(nw_proto == IPPROTO_SCTP)) { + if (OVS_LIKELY(size >= SCTP_HEADER_LEN)) { + const struct sctp_header *sctp = data; - 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); + miniflow_push_words(mf, tp_src, &sctp->sctp_src, 1); + } + } else if (OVS_LIKELY(nw_proto == IPPROTO_ICMP)) { + if (OVS_LIKELY(size >= ICMP_HEADER_LEN)) { + const struct icmp_header *icmp = data; + + miniflow_push_be16(mf, tp_src, htons(icmp->icmp_type)); + miniflow_push_be16(mf, tp_dst, htons(icmp->icmp_code)); + } + } else if (OVS_LIKELY(nw_proto == IPPROTO_ICMPV6)) { + if (OVS_LIKELY(size >= sizeof(struct icmp6_hdr))) { + const struct in6_addr *nd_target = NULL; + uint8_t arp_buf[2][ETH_ADDR_LEN]; + const struct icmp6_hdr *icmp = data_pull(&data, &size, + sizeof *icmp); + memset(arp_buf, 0, sizeof arp_buf); + if (OVS_LIKELY(parse_icmpv6(&data, &size, icmp, &nd_target, + arp_buf))) { + if (nd_target) { + miniflow_push_words(mf, nd_target, nd_target, + sizeof *nd_target / 4); + } + miniflow_push_words(mf, arp_sha, arp_buf, + ETH_ADDR_LEN * 2 / 4); + miniflow_push_be16(mf, tp_src, htons(icmp->icmp6_type)); + miniflow_push_be16(mf, tp_dst, htons(icmp->icmp6_code)); + } + } } } + if (md) { + miniflow_push_uint32_check(mf, dp_hash, md->dp_hash); + } + out: + dst->map = mf.map; } /* For every bit of a field that is wildcarded in 'wildcards', sets the @@ -511,12 +637,26 @@ flow_zero_wildcards(struct flow *flow, const struct flow_wildcards *wildcards) } } +void +flow_unwildcard_tp_ports(const struct flow *flow, struct flow_wildcards *wc) +{ + if (flow->nw_proto != IPPROTO_ICMP) { + memset(&wc->masks.tp_src, 0xff, sizeof wc->masks.tp_src); + memset(&wc->masks.tp_dst, 0xff, sizeof wc->masks.tp_dst); + } else { + wc->masks.tp_src = htons(0xff); + wc->masks.tp_dst = htons(0xff); + } +} + /* Initializes 'fmd' with the metadata found in 'flow'. */ void flow_get_metadata(const struct flow *flow, struct flow_metadata *fmd) { - BUILD_ASSERT_DECL(FLOW_WC_SEQ == 22); + BUILD_ASSERT_DECL(FLOW_WC_SEQ == 26); + fmd->dp_hash = flow->dp_hash; + fmd->recirc_id = flow->recirc_id; fmd->tun_id = flow->tunnel.tun_id; fmd->tun_src = flow->tunnel.ip_src; fmd->tun_dst = flow->tunnel.ip_dst; @@ -578,6 +718,24 @@ format_flags(struct ds *ds, const char *(*bit_to_string)(uint32_t), ds_chomp(ds, del); } +void +format_flags_masked(struct ds *ds, const char *name, + const char *(*bit_to_string)(uint32_t), uint32_t flags, + uint32_t mask) +{ + if (name) { + ds_put_format(ds, "%s=", name); + } + while (mask) { + uint32_t bit = rightmost_1bit(mask); + const char *s = bit_to_string(bit); + + ds_put_format(ds, "%s%s", (flags & bit) ? "+" : "-", + s ? s : "[Unknown]"); + mask &= ~bit; + } +} + void flow_format(struct ds *ds, const struct flow *flow) { @@ -604,6 +762,15 @@ flow_wildcards_init_catchall(struct flow_wildcards *wc) memset(&wc->masks, 0, sizeof wc->masks); } +/* Clear the metadata and register wildcard masks. They are not packet + * header fields. */ +void +flow_wildcards_clear_non_packet_fields(struct flow_wildcards *wc) +{ + memset(&wc->masks.metadata, 0, sizeof wc->masks.metadata); + memset(&wc->masks.regs, 0, sizeof wc->masks.regs); +} + /* Returns true if 'wc' matches every packet, false if 'wc' fixes any bits or * fields. */ bool @@ -656,28 +823,6 @@ flow_wildcards_or(struct flow_wildcards *dst, } } -/* 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; - int ofs = 0; - uint64_t map; - - for (map = src->map; map; map = zero_rightmost_1bit(map)) { - dst_u32[raw_ctz64(map)] |= src->values[ofs++]; - } -} - -/* 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); -} - /* Returns a hash of the wildcards in 'wc'. */ uint32_t flow_wildcards_hash(const struct flow_wildcards *wc, uint32_t basis) @@ -739,6 +884,78 @@ flow_wildcards_set_reg_mask(struct flow_wildcards *wc, int idx, uint32_t mask) wc->masks.regs[idx] = mask; } +/* Calculates the 5-tuple hash from the given miniflow. + * This returns the same value as flow_hash_5tuple for the corresponding + * flow. */ +uint32_t +miniflow_hash_5tuple(const struct miniflow *flow, uint32_t basis) +{ + uint32_t hash = basis; + + if (flow) { + ovs_be16 dl_type = MINIFLOW_GET_BE16(flow, dl_type); + + hash = mhash_add(hash, MINIFLOW_GET_U8(flow, nw_proto)); + + /* Separate loops for better optimization. */ + if (dl_type == htons(ETH_TYPE_IPV6)) { + uint64_t map = MINIFLOW_MAP(ipv6_src) | MINIFLOW_MAP(ipv6_dst) + | MINIFLOW_MAP(tp_src); /* Covers both ports */ + uint32_t value; + + MINIFLOW_FOR_EACH_IN_MAP(value, flow, map) { + hash = mhash_add(hash, value); + } + } else { + uint64_t map = MINIFLOW_MAP(nw_src) | MINIFLOW_MAP(nw_dst) + | MINIFLOW_MAP(tp_src); /* Covers both ports */ + uint32_t value; + + MINIFLOW_FOR_EACH_IN_MAP(value, flow, map) { + hash = mhash_add(hash, value); + } + } + hash = mhash_finish(hash, 42); /* Arbitrary number. */ + } + return hash; +} + +BUILD_ASSERT_DECL(offsetof(struct flow, tp_src) + 2 + == offsetof(struct flow, tp_dst) && + offsetof(struct flow, tp_src) / 4 + == offsetof(struct flow, tp_dst) / 4); +BUILD_ASSERT_DECL(offsetof(struct flow, ipv6_src) + 16 + == offsetof(struct flow, ipv6_dst)); + +/* Calculates the 5-tuple hash from the given flow. */ +uint32_t +flow_hash_5tuple(const struct flow *flow, uint32_t basis) +{ + uint32_t hash = basis; + + if (flow) { + const uint32_t *flow_u32 = (const uint32_t *)flow; + + hash = mhash_add(hash, flow->nw_proto); + + if (flow->dl_type == htons(ETH_TYPE_IPV6)) { + int ofs = offsetof(struct flow, ipv6_src) / 4; + int end = ofs + 2 * sizeof flow->ipv6_src / 4; + + while (ofs < end) { + hash = mhash_add(hash, flow_u32[ofs++]); + } + } else { + hash = mhash_add(hash, (OVS_FORCE uint32_t) flow->nw_src); + hash = mhash_add(hash, (OVS_FORCE uint32_t) flow->nw_dst); + } + hash = mhash_add(hash, flow_u32[offsetof(struct flow, tp_src) / 4]); + + hash = mhash_finish(hash, 42); /* Arbitrary number. */ + } + return hash; +} + /* Hashes 'flow' based on its L2 through L4 protocol information. */ uint32_t flow_hash_symmetric_l4(const struct flow *flow, uint32_t basis) @@ -850,14 +1067,13 @@ flow_mask_hash_fields(const struct flow *flow, struct flow_wildcards *wc, } if (is_ip_any(flow)) { memset(&wc->masks.nw_proto, 0xff, sizeof wc->masks.nw_proto); - memset(&wc->masks.tp_src, 0xff, sizeof wc->masks.tp_src); - memset(&wc->masks.tp_dst, 0xff, sizeof wc->masks.tp_dst); + flow_unwildcard_tp_ports(flow, wc); } wc->masks.vlan_tci |= htons(VLAN_VID_MASK | VLAN_CFI); break; default: - NOT_REACHED(); + OVS_NOT_REACHED(); } } @@ -875,7 +1091,7 @@ flow_hash_fields(const struct flow *flow, enum nx_hash_fields fields, return flow_hash_symmetric_l4(flow, basis); } - NOT_REACHED(); + OVS_NOT_REACHED(); } /* Returns a string representation of 'fields'. */ @@ -964,35 +1180,288 @@ flow_set_vlan_pcp(struct flow *flow, uint8_t pcp) flow->vlan_tci |= htons((pcp << VLAN_PCP_SHIFT) | VLAN_CFI); } +/* Returns the number of MPLS LSEs present in 'flow' + * + * Returns 0 if the 'dl_type' of 'flow' is not an MPLS ethernet type. + * Otherwise traverses 'flow''s MPLS label stack stopping at the + * first entry that has the BoS bit set. If no such entry exists then + * the maximum number of LSEs that can be stored in 'flow' is returned. + */ +int +flow_count_mpls_labels(const struct flow *flow, struct flow_wildcards *wc) +{ + if (wc) { + wc->masks.dl_type = OVS_BE16_MAX; + } + if (eth_type_mpls(flow->dl_type)) { + int i; + int len = FLOW_MAX_MPLS_LABELS; + + for (i = 0; i < len; i++) { + if (wc) { + wc->masks.mpls_lse[i] |= htonl(MPLS_BOS_MASK); + } + if (flow->mpls_lse[i] & htonl(MPLS_BOS_MASK)) { + return i + 1; + } + } + + return len; + } else { + return 0; + } +} + +/* Returns the number consecutive of MPLS LSEs, starting at the + * innermost LSE, that are common in 'a' and 'b'. + * + * 'an' must be flow_count_mpls_labels(a). + * 'bn' must be flow_count_mpls_labels(b). + */ +int +flow_count_common_mpls_labels(const struct flow *a, int an, + const struct flow *b, int bn, + struct flow_wildcards *wc) +{ + int min_n = MIN(an, bn); + if (min_n == 0) { + return 0; + } else { + int common_n = 0; + int a_last = an - 1; + int b_last = bn - 1; + int i; + + for (i = 0; i < min_n; i++) { + if (wc) { + wc->masks.mpls_lse[a_last - i] = OVS_BE32_MAX; + wc->masks.mpls_lse[b_last - i] = OVS_BE32_MAX; + } + if (a->mpls_lse[a_last - i] != b->mpls_lse[b_last - i]) { + break; + } else { + common_n++; + } + } + + return common_n; + } +} + +/* Adds a new outermost MPLS label to 'flow' and changes 'flow''s Ethernet type + * to 'mpls_eth_type', which must be an MPLS Ethertype. + * + * If the new label is the first MPLS label in 'flow', it is generated as; + * + * - label: 2, if 'flow' is IPv6, otherwise 0. + * + * - TTL: IPv4 or IPv6 TTL, if present and nonzero, otherwise 64. + * + * - TC: IPv4 or IPv6 TOS, if present, otherwise 0. + * + * - BoS: 1. + * + * If the new label is the second or label MPLS label in 'flow', it is + * generated as; + * + * - label: Copied from outer label. + * + * - TTL: Copied from outer label. + * + * - TC: Copied from outer label. + * + * - BoS: 0. + * + * 'n' must be flow_count_mpls_labels(flow). 'n' must be less than + * FLOW_MAX_MPLS_LABELS (because otherwise flow->mpls_lse[] would overflow). + */ +void +flow_push_mpls(struct flow *flow, int n, ovs_be16 mpls_eth_type, + struct flow_wildcards *wc) +{ + ovs_assert(eth_type_mpls(mpls_eth_type)); + ovs_assert(n < FLOW_MAX_MPLS_LABELS); + + memset(wc->masks.mpls_lse, 0xff, sizeof wc->masks.mpls_lse); + if (n) { + int i; + + for (i = n; i >= 1; i--) { + flow->mpls_lse[i] = flow->mpls_lse[i - 1]; + } + flow->mpls_lse[0] = (flow->mpls_lse[1] + & htonl(~MPLS_BOS_MASK)); + } else { + int label = 0; /* IPv4 Explicit Null. */ + int tc = 0; + int ttl = 64; + + if (flow->dl_type == htons(ETH_TYPE_IPV6)) { + label = 2; + } + + if (is_ip_any(flow)) { + tc = (flow->nw_tos & IP_DSCP_MASK) >> 2; + wc->masks.nw_tos |= IP_DSCP_MASK; + + if (flow->nw_ttl) { + ttl = flow->nw_ttl; + } + wc->masks.nw_ttl = 0xff; + } + + flow->mpls_lse[0] = set_mpls_lse_values(ttl, tc, 1, htonl(label)); + + /* Clear all L3 and L4 fields. */ + BUILD_ASSERT(FLOW_WC_SEQ == 26); + memset((char *) flow + FLOW_SEGMENT_2_ENDS_AT, 0, + sizeof(struct flow) - FLOW_SEGMENT_2_ENDS_AT); + } + flow->dl_type = mpls_eth_type; +} + +/* Tries to remove the outermost MPLS label from 'flow'. Returns true if + * successful, false otherwise. On success, sets 'flow''s Ethernet type to + * 'eth_type'. + * + * 'n' must be flow_count_mpls_labels(flow). */ +bool +flow_pop_mpls(struct flow *flow, int n, ovs_be16 eth_type, + struct flow_wildcards *wc) +{ + int i; + + if (n == 0) { + /* Nothing to pop. */ + return false; + } else if (n == FLOW_MAX_MPLS_LABELS + && !(flow->mpls_lse[n - 1] & htonl(MPLS_BOS_MASK))) { + /* Can't pop because we don't know what to fill in mpls_lse[n - 1]. */ + return false; + } + + memset(wc->masks.mpls_lse, 0xff, sizeof wc->masks.mpls_lse); + for (i = 1; i < n; i++) { + flow->mpls_lse[i - 1] = flow->mpls_lse[i]; + } + flow->mpls_lse[n - 1] = 0; + flow->dl_type = eth_type; + return true; +} + /* Sets the MPLS Label that 'flow' matches to 'label', which is interpreted * as an OpenFlow 1.1 "mpls_label" value. */ void -flow_set_mpls_label(struct flow *flow, ovs_be32 label) +flow_set_mpls_label(struct flow *flow, int idx, ovs_be32 label) { - set_mpls_lse_label(&flow->mpls_lse, label); + set_mpls_lse_label(&flow->mpls_lse[idx], label); } /* Sets the MPLS TTL that 'flow' matches to 'ttl', which should be in the * range 0...255. */ void -flow_set_mpls_ttl(struct flow *flow, uint8_t ttl) +flow_set_mpls_ttl(struct flow *flow, int idx, uint8_t ttl) { - set_mpls_lse_ttl(&flow->mpls_lse, ttl); + set_mpls_lse_ttl(&flow->mpls_lse[idx], ttl); } /* Sets the MPLS TC that 'flow' matches to 'tc', which should be in the * range 0...7. */ void -flow_set_mpls_tc(struct flow *flow, uint8_t tc) +flow_set_mpls_tc(struct flow *flow, int idx, uint8_t tc) { - set_mpls_lse_tc(&flow->mpls_lse, tc); + set_mpls_lse_tc(&flow->mpls_lse[idx], tc); } /* Sets the MPLS BOS bit that 'flow' matches to which should be 0 or 1. */ void -flow_set_mpls_bos(struct flow *flow, uint8_t bos) +flow_set_mpls_bos(struct flow *flow, int idx, uint8_t bos) +{ + set_mpls_lse_bos(&flow->mpls_lse[idx], bos); +} + +/* Sets the entire MPLS LSE. */ +void +flow_set_mpls_lse(struct flow *flow, int idx, ovs_be32 lse) +{ + flow->mpls_lse[idx] = lse; +} + +static size_t +flow_compose_l4(struct ofpbuf *b, const struct flow *flow) { - set_mpls_lse_bos(&flow->mpls_lse, bos); + size_t l4_len = 0; + + if (!(flow->nw_frag & FLOW_NW_FRAG_ANY) + || !(flow->nw_frag & FLOW_NW_FRAG_LATER)) { + if (flow->nw_proto == IPPROTO_TCP) { + struct tcp_header *tcp; + + l4_len = sizeof *tcp; + tcp = ofpbuf_put_zeros(b, l4_len); + tcp->tcp_src = flow->tp_src; + tcp->tcp_dst = flow->tp_dst; + tcp->tcp_ctl = TCP_CTL(ntohs(flow->tcp_flags), 5); + } else if (flow->nw_proto == IPPROTO_UDP) { + struct udp_header *udp; + + l4_len = sizeof *udp; + udp = ofpbuf_put_zeros(b, l4_len); + udp->udp_src = flow->tp_src; + udp->udp_dst = flow->tp_dst; + } else if (flow->nw_proto == IPPROTO_SCTP) { + struct sctp_header *sctp; + + l4_len = sizeof *sctp; + sctp = ofpbuf_put_zeros(b, l4_len); + sctp->sctp_src = flow->tp_src; + sctp->sctp_dst = flow->tp_dst; + } else if (flow->nw_proto == IPPROTO_ICMP) { + struct icmp_header *icmp; + + l4_len = sizeof *icmp; + icmp = ofpbuf_put_zeros(b, l4_len); + icmp->icmp_type = ntohs(flow->tp_src); + icmp->icmp_code = ntohs(flow->tp_dst); + icmp->icmp_csum = csum(icmp, ICMP_HEADER_LEN); + } else if (flow->nw_proto == IPPROTO_ICMPV6) { + struct icmp6_hdr *icmp; + + l4_len = sizeof *icmp; + icmp = ofpbuf_put_zeros(b, l4_len); + icmp->icmp6_type = ntohs(flow->tp_src); + icmp->icmp6_code = ntohs(flow->tp_dst); + + if (icmp->icmp6_code == 0 && + (icmp->icmp6_type == ND_NEIGHBOR_SOLICIT || + icmp->icmp6_type == ND_NEIGHBOR_ADVERT)) { + struct in6_addr *nd_target; + struct nd_opt_hdr *nd_opt; + + l4_len += sizeof *nd_target; + nd_target = ofpbuf_put_zeros(b, sizeof *nd_target); + *nd_target = flow->nd_target; + + if (!eth_addr_is_zero(flow->arp_sha)) { + l4_len += 8; + nd_opt = ofpbuf_put_zeros(b, 8); + nd_opt->nd_opt_len = 1; + nd_opt->nd_opt_type = ND_OPT_SOURCE_LINKADDR; + memcpy(nd_opt + 1, flow->arp_sha, ETH_ADDR_LEN); + } + if (!eth_addr_is_zero(flow->arp_tha)) { + l4_len += 8; + nd_opt = ofpbuf_put_zeros(b, 8); + nd_opt->nd_opt_len = 1; + nd_opt->nd_opt_type = ND_OPT_TARGET_LINKADDR; + memcpy(nd_opt + 1, flow->arp_tha, ETH_ADDR_LEN); + } + } + icmp->icmp6_cksum = (OVS_FORCE uint16_t) + csum(icmp, (char *)ofpbuf_tail(b) - (char *)icmp); + } + } + return l4_len; } /* Puts into 'b' a packet that flow_extract() would parse as having the given @@ -1004,21 +1473,24 @@ flow_set_mpls_bos(struct flow *flow, uint8_t bos) void flow_compose(struct ofpbuf *b, const struct flow *flow) { + size_t l4_len; + + /* eth_compose() sets l3 pointer and makes sure it is 32-bit aligned. */ eth_compose(b, flow->dl_dst, flow->dl_src, ntohs(flow->dl_type), 0); if (flow->dl_type == htons(FLOW_DL_TYPE_NONE)) { - struct eth_header *eth = b->l2; - eth->eth_type = htons(b->size); + struct eth_header *eth = ofpbuf_l2(b); + eth->eth_type = htons(ofpbuf_size(b)); return; } if (flow->vlan_tci & htons(VLAN_CFI)) { - eth_push_vlan(b, flow->vlan_tci); + eth_push_vlan(b, htons(ETH_TYPE_VLAN), flow->vlan_tci); } if (flow->dl_type == htons(ETH_TYPE_IP)) { struct ip_header *ip; - b->l3 = ip = ofpbuf_put_zeros(b, sizeof *ip); + ip = ofpbuf_put_zeros(b, sizeof *ip); ip->ip_ihl_ver = IP_IHL_VER(5, 4); ip->ip_tos = flow->nw_tos; ip->ip_ttl = flow->nw_ttl; @@ -1032,49 +1504,36 @@ flow_compose(struct ofpbuf *b, const struct flow *flow) ip->ip_frag_off |= htons(100); } } - if (!(flow->nw_frag & FLOW_NW_FRAG_ANY) - || !(flow->nw_frag & FLOW_NW_FRAG_LATER)) { - if (flow->nw_proto == IPPROTO_TCP) { - struct tcp_header *tcp; - - b->l4 = tcp = ofpbuf_put_zeros(b, sizeof *tcp); - tcp->tcp_src = flow->tp_src; - tcp->tcp_dst = flow->tp_dst; - tcp->tcp_ctl = TCP_CTL(ntohs(flow->tcp_flags), 5); - } else if (flow->nw_proto == IPPROTO_UDP) { - struct udp_header *udp; - - b->l4 = udp = ofpbuf_put_zeros(b, sizeof *udp); - udp->udp_src = flow->tp_src; - udp->udp_dst = flow->tp_dst; - } else if (flow->nw_proto == IPPROTO_SCTP) { - struct sctp_header *sctp; - - b->l4 = sctp = ofpbuf_put_zeros(b, sizeof *sctp); - sctp->sctp_src = flow->tp_src; - sctp->sctp_dst = flow->tp_dst; - } else if (flow->nw_proto == IPPROTO_ICMP) { - struct icmp_header *icmp; - - b->l4 = icmp = ofpbuf_put_zeros(b, sizeof *icmp); - icmp->icmp_type = ntohs(flow->tp_src); - icmp->icmp_code = ntohs(flow->tp_dst); - icmp->icmp_csum = csum(icmp, ICMP_HEADER_LEN); - } - b->l7 = ofpbuf_tail(b); - } - ip = b->l3; - ip->ip_tot_len = htons((uint8_t *) b->data + b->size - - (uint8_t *) b->l3); + ofpbuf_set_l4(b, ofpbuf_tail(b)); + + l4_len = flow_compose_l4(b, flow); + + ip->ip_tot_len = htons(b->l4_ofs - b->l3_ofs + l4_len); ip->ip_csum = csum(ip, sizeof *ip); } else if (flow->dl_type == htons(ETH_TYPE_IPV6)) { - /* XXX */ + struct ovs_16aligned_ip6_hdr *nh; + + nh = ofpbuf_put_zeros(b, sizeof *nh); + put_16aligned_be32(&nh->ip6_flow, htonl(6 << 28) | + htonl(flow->nw_tos << 20) | flow->ipv6_label); + nh->ip6_hlim = flow->nw_ttl; + nh->ip6_nxt = flow->nw_proto; + + memcpy(&nh->ip6_src, &flow->ipv6_src, sizeof(nh->ip6_src)); + memcpy(&nh->ip6_dst, &flow->ipv6_dst, sizeof(nh->ip6_dst)); + + ofpbuf_set_l4(b, ofpbuf_tail(b)); + + l4_len = flow_compose_l4(b, flow); + + nh->ip6_plen = htons(l4_len); } else if (flow->dl_type == htons(ETH_TYPE_ARP) || flow->dl_type == htons(ETH_TYPE_RARP)) { struct arp_eth_header *arp; - b->l3 = arp = ofpbuf_put_zeros(b, sizeof *arp); + arp = ofpbuf_put_zeros(b, sizeof *arp); + ofpbuf_set_l3(b, arp); arp->ar_hrd = htons(1); arp->ar_pro = htons(ETH_TYPE_IP); arp->ar_hln = ETH_ADDR_LEN; @@ -1091,8 +1550,17 @@ flow_compose(struct ofpbuf *b, const struct flow *flow) } if (eth_type_mpls(flow->dl_type)) { - b->l2_5 = b->l3; - push_mpls(b, flow->dl_type, flow->mpls_lse); + int n; + + b->l2_5_ofs = b->l3_ofs; + for (n = 1; n < FLOW_MAX_MPLS_LABELS; n++) { + if (flow->mpls_lse[n - 1] & htonl(MPLS_BOS_MASK)) { + break; + } + } + while (n > 0) { + push_mpls(b, flow->dl_type, flow->mpls_lse[--n]); + } } } @@ -1101,7 +1569,7 @@ flow_compose(struct ofpbuf *b, const struct flow *flow) static int miniflow_n_values(const struct miniflow *flow) { - return popcount64(flow->map); + return count_1bits(flow->map); } static uint32_t * @@ -1117,12 +1585,16 @@ miniflow_alloc_values(struct miniflow *flow, int n) /* Completes an initialization of 'dst' as a miniflow copy of 'src' begun by * the caller. The caller must have already initialized 'dst->map' properly - * to indicate the nonzero uint32_t elements of 'src'. 'n' must be the number - * of 1-bits in 'dst->map'. + * to indicate the significant uint32_t elements of 'src'. 'n' must be the + * number of 1-bits in 'dst->map'. + * + * Normally the significant elements are the ones that are non-zero. However, + * when a miniflow is initialized from a (mini)mask, the values can be zeroes, + * so that the flow and mask always have the same maps. * * This function initializes 'dst->values' (either inline if possible or with - * malloc() otherwise) and copies the nonzero uint32_t elements of 'src' into - * it. */ + * malloc() otherwise) and copies the uint32_t elements of 'src' indicated by + * 'dst->map' into it. */ static void miniflow_init__(struct miniflow *dst, const struct flow *src, int n) { @@ -1133,7 +1605,7 @@ miniflow_init__(struct miniflow *dst, const struct flow *src, int n) 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_ctz64(map)]; + dst->values[ofs++] = src_u32[raw_ctz(map)]; } } @@ -1214,42 +1686,15 @@ miniflow_expand(const struct miniflow *src, struct flow *dst) flow_union_with_miniflow(dst, src); } -static const uint32_t * -miniflow_get__(const struct miniflow *flow, unsigned int u32_ofs) -{ - if (!(flow->map & (UINT64_C(1) << u32_ofs))) { - static const uint32_t zero = 0; - return &zero; - } - return flow->values - + popcount64(flow->map & ((UINT64_C(1) << u32_ofs) - 1)); -} - /* Returns the uint32_t that would be at byte offset '4 * u32_ofs' if 'flow' * were expanded into a "struct flow". */ -uint32_t +static 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); + return (flow->map & UINT64_C(1) << u32_ofs) + ? *(flow->values + + count_1bits(flow->map & ((UINT64_C(1) << u32_ofs) - 1))) + : 0; } /* Returns true if 'a' and 'b' are the same flow, false otherwise. */ @@ -1295,7 +1740,7 @@ miniflow_equal_in_minimask(const struct miniflow *a, const struct miniflow *b, p = mask->masks.values; for (map = mask->masks.map; map; map = zero_rightmost_1bit(map)) { - int ofs = raw_ctz64(map); + int ofs = raw_ctz(map); if ((miniflow_get(a, ofs) ^ miniflow_get(b, ofs)) & *p) { return false; @@ -1319,7 +1764,7 @@ miniflow_equal_flow_in_minimask(const struct miniflow *a, const struct flow *b, p = mask->masks.values; for (map = mask->masks.map; map; map = zero_rightmost_1bit(map)) { - int ofs = raw_ctz64(map); + int ofs = raw_ctz(map); if ((miniflow_get(a, ofs) ^ b_u32[ofs]) & *p) { return false; @@ -1330,78 +1775,6 @@ miniflow_equal_flow_in_minimask(const struct miniflow *a, const struct flow *b, return true; } -/* Returns a hash value for 'flow', given 'basis'. */ -uint32_t -miniflow_hash(const struct miniflow *flow, uint32_t 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 - * '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; - uint64_t map; - - hash = basis; - - for (map = mask->masks.map; map; map = zero_rightmost_1bit(map)) { - if (*p) { - int ofs = raw_ctz64(map); - hash = mhash_add(hash, miniflow_get(flow, ofs) & *p); - } - p++; - } - - return mhash_finish(hash, (p - mask->masks.values) * 4); -} - -/* 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; - uint64_t map; - - hash = basis; - for (map = mask->masks.map; map; map = zero_rightmost_1bit(map)) { - if (*p) { - hash = mhash_add(hash, flow_u32[raw_ctz64(map)] & *p); - } - p++; - } - - 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(). */ @@ -1446,7 +1819,7 @@ minimask_combine(struct minimask *dst_, dst->map = 0; for (map = a->map & b->map; map; map = zero_rightmost_1bit(map)) { - int ofs = raw_ctz64(map); + int ofs = raw_ctz(map); uint32_t mask = miniflow_get(a, ofs) & miniflow_get(b, ofs); if (mask) { @@ -1479,14 +1852,6 @@ 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) @@ -1494,26 +1859,17 @@ 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_', +/* Returns true if at least one bit matched by 'b' is wildcarded by 'a', * false otherwise. */ bool -minimask_has_extra(const struct minimask *a_, const struct minimask *b_) +minimask_has_extra(const struct minimask *a, const struct minimask *b) { - const struct miniflow *a = &a_->masks; - const struct miniflow *b = &b_->masks; + const uint32_t *p = b->masks.values; uint64_t map; - for (map = a->map | b->map; map; map = zero_rightmost_1bit(map)) { - int ofs = raw_ctz64(map); - uint32_t a_u32 = miniflow_get(a, ofs); - uint32_t b_u32 = miniflow_get(b, ofs); + for (map = b->masks.map; map; map = zero_rightmost_1bit(map)) { + uint32_t a_u32 = minimask_get(a, raw_ctz(map)); + uint32_t b_u32 = *p++; if ((a_u32 & b_u32) != b_u32) { return true; @@ -1522,20 +1878,3 @@ minimask_has_extra(const struct minimask *a_, const struct minimask *b_) 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; - const uint32_t *p = mask->values; - uint64_t map; - - for (map = mask->map; map; map = zero_rightmost_1bit(map)) { - if (*p++) { - return false; - } - } - return true; -}