X-Git-Url: http://git.onelab.eu/?a=blobdiff_plain;f=lib%2Fflow.c;h=51851cfdb12950af524b918e3872999ea984ea49;hb=6e5fe8065cb056b3bc7f39051763480703e05496;hp=a42fea1e3d0f3113b0959b6f80cae9e25312642a;hpb=7431e17196fdb1c3189d67e3aeed4adeab4cf479;p=sliver-openvswitch.git diff --git a/lib/flow.c b/lib/flow.c index a42fea1e3..51851cfdb 100644 --- a/lib/flow.c +++ b/lib/flow.c @@ -35,10 +35,8 @@ #include "ofpbuf.h" #include "openflow/openflow.h" #include "packets.h" +#include "random.h" #include "unaligned.h" -#include "vlog.h" - -VLOG_DEFINE_THIS_MODULE(flow); COVERAGE_DEFINE(flow_extract); COVERAGE_DEFINE(miniflow_malloc); @@ -81,6 +79,12 @@ pull_udp(struct ofpbuf *packet) return ofpbuf_try_pull(packet, UDP_HEADER_LEN); } +static struct sctp_header * +pull_sctp(struct ofpbuf *packet) +{ + return ofpbuf_try_pull(packet, SCTP_HEADER_LEN); +} + static struct icmp_header * pull_icmp(struct ofpbuf *packet) { @@ -97,9 +101,11 @@ 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 (flow->mpls_depth++ == 0) { + if (top) { + top = false; flow->mpls_lse = mh->mpls_lse; } if (mh->mpls_lse & htonl(MPLS_BOS_MASK)) { @@ -158,7 +164,7 @@ parse_ethertype(struct ofpbuf *b) static int parse_ipv6(struct ofpbuf *packet, struct flow *flow) { - const struct ip6_hdr *nh; + const struct ovs_16aligned_ip6_hdr *nh; ovs_be32 tc_flow; int nexthdr; @@ -169,10 +175,10 @@ parse_ipv6(struct ofpbuf *packet, struct flow *flow) nexthdr = nh->ip6_nxt; - flow->ipv6_src = nh->ip6_src; - flow->ipv6_dst = nh->ip6_dst; + memcpy(&flow->ipv6_src, &nh->ip6_src, sizeof flow->ipv6_src); + memcpy(&flow->ipv6_dst, &nh->ip6_dst, sizeof flow->ipv6_dst); - tc_flow = get_unaligned_be32(&nh->ip6_flow); + tc_flow = get_16aligned_be32(&nh->ip6_flow); flow->nw_tos = ntohl(tc_flow) >> 20; flow->ipv6_label = tc_flow & htonl(IPV6_LABEL_MASK); flow->nw_ttl = nh->ip6_hlim; @@ -220,7 +226,7 @@ parse_ipv6(struct ofpbuf *packet, struct flow *flow) return EINVAL; } } else if (nexthdr == IPPROTO_FRAGMENT) { - const struct ip6_frag *frag_hdr = packet->data; + const struct ovs_16aligned_ip6_frag *frag_hdr = packet->data; nexthdr = frag_hdr->ip6f_nxt; if (!ofpbuf_try_pull(packet, sizeof *frag_hdr)) { @@ -265,6 +271,17 @@ parse_udp(struct ofpbuf *packet, struct ofpbuf *b, struct flow *flow) } } +static void +parse_sctp(struct ofpbuf *packet, struct ofpbuf *b, struct flow *flow) +{ + const struct sctp_header *sctp = pull_sctp(b); + if (sctp) { + flow->tp_src = sctp->sctp_src; + flow->tp_dst = sctp->sctp_dst; + packet->l7 = b->data; + } +} + static bool parse_icmpv6(struct ofpbuf *b, struct flow *flow) { @@ -352,11 +369,11 @@ invalid: * - packet->l4 to just past the IPv4 header, if one is present and has a * correct length, and otherwise NULL. * - * - packet->l7 to just past the TCP or UDP or ICMP header, if one is + * - packet->l7 to just past the TCP/UDP/SCTP/ICMP header, if one is * present and has a correct length, and otherwise NULL. */ void -flow_extract(struct ofpbuf *packet, uint32_t skb_priority, uint32_t skb_mark, +flow_extract(struct ofpbuf *packet, uint32_t skb_priority, uint32_t pkt_mark, const struct flow_tnl *tnl, const union flow_in_port *in_port, struct flow *flow) { @@ -375,7 +392,7 @@ flow_extract(struct ofpbuf *packet, uint32_t skb_priority, uint32_t skb_mark, flow->in_port = *in_port; } flow->skb_priority = skb_priority; - flow->skb_mark = skb_mark; + flow->pkt_mark = pkt_mark; packet->l2 = b.data; packet->l2_5 = NULL; @@ -412,8 +429,8 @@ flow_extract(struct ofpbuf *packet, uint32_t skb_priority, uint32_t skb_mark, if (nh) { packet->l4 = b.data; - flow->nw_src = get_unaligned_be32(&nh->ip_src); - flow->nw_dst = get_unaligned_be32(&nh->ip_dst); + flow->nw_src = get_16aligned_be32(&nh->ip_src); + flow->nw_dst = get_16aligned_be32(&nh->ip_dst); flow->nw_proto = nh->ip_proto; flow->nw_tos = nh->ip_tos; @@ -430,6 +447,8 @@ flow_extract(struct ofpbuf *packet, uint32_t skb_priority, uint32_t skb_mark, 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) { @@ -450,6 +469,8 @@ flow_extract(struct ofpbuf *packet, uint32_t skb_priority, uint32_t skb_mark, 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; @@ -467,8 +488,8 @@ flow_extract(struct ofpbuf *packet, uint32_t skb_priority, uint32_t skb_mark, flow->nw_proto = ntohs(arp->ar_op); } - flow->nw_src = arp->ar_spa; - flow->nw_dst = arp->ar_tpa; + flow->nw_src = get_16aligned_be32(&arp->ar_spa); + flow->nw_dst = get_16aligned_be32(&arp->ar_tpa); memcpy(flow->arp_sha, arp->ar_sha, ETH_ADDR_LEN); memcpy(flow->arp_tha, arp->ar_tha, ETH_ADDR_LEN); } @@ -493,13 +514,14 @@ flow_zero_wildcards(struct flow *flow, const struct flow_wildcards *wildcards) void flow_get_metadata(const struct flow *flow, struct flow_metadata *fmd) { - BUILD_ASSERT_DECL(FLOW_WC_SEQ == 20); + BUILD_ASSERT_DECL(FLOW_WC_SEQ == 21); fmd->tun_id = flow->tunnel.tun_id; fmd->tun_src = flow->tunnel.ip_src; fmd->tun_dst = flow->tunnel.ip_dst; fmd->metadata = flow->metadata; memcpy(fmd->regs, flow->regs, sizeof fmd->regs); + fmd->pkt_mark = flow->pkt_mark; fmd->in_port = flow->in_port.ofp_port; } @@ -581,15 +603,6 @@ flow_wildcards_init_catchall(struct flow_wildcards *wc) memset(&wc->masks, 0, sizeof wc->masks); } -/* Initializes 'wc' as an exact-match set of wildcards; that is, 'wc' does not - * wildcard any bits or fields. */ -void -flow_wildcards_init_exact(struct flow_wildcards *wc) -{ - memset(&wc->masks, 0xff, sizeof wc->masks); - memset(wc->masks.zeros, 0, sizeof wc->masks.zeros); -} - /* Returns true if 'wc' matches every packet, false if 'wc' fixes any bits or * fields. */ bool @@ -760,7 +773,7 @@ flow_hash_symmetric_l4(const struct flow *flow, uint32_t basis) if (fields.eth_type == htons(ETH_TYPE_IP)) { fields.ipv4_addr = flow->nw_src ^ flow->nw_dst; fields.ip_proto = flow->nw_proto; - if (fields.ip_proto == IPPROTO_TCP) { + if (fields.ip_proto == IPPROTO_TCP || fields.ip_proto == IPPROTO_SCTP) { fields.tp_port = flow->tp_src ^ flow->tp_dst; } } else if (fields.eth_type == htons(ETH_TYPE_IPV6)) { @@ -772,16 +785,57 @@ flow_hash_symmetric_l4(const struct flow *flow, uint32_t basis) ipv6_addr[i] = a[i] ^ b[i]; } fields.ip_proto = flow->nw_proto; - if (fields.ip_proto == IPPROTO_TCP) { + if (fields.ip_proto == IPPROTO_TCP || fields.ip_proto == IPPROTO_SCTP) { fields.tp_port = flow->tp_src ^ flow->tp_dst; } } return jhash_bytes(&fields, sizeof fields, basis); } +/* Initialize a flow with random fields that matter for nx_hash_fields. */ +void +flow_random_hash_fields(struct flow *flow) +{ + uint16_t rnd = random_uint16(); + + /* Initialize to all zeros. */ + memset(flow, 0, sizeof *flow); + + eth_addr_random(flow->dl_src); + eth_addr_random(flow->dl_dst); + + flow->vlan_tci = (OVS_FORCE ovs_be16) (random_uint16() & VLAN_VID_MASK); + + /* Make most of the random flows IPv4, some IPv6, and rest random. */ + flow->dl_type = rnd < 0x8000 ? htons(ETH_TYPE_IP) : + rnd < 0xc000 ? htons(ETH_TYPE_IPV6) : (OVS_FORCE ovs_be16)rnd; + + if (dl_type_is_ip_any(flow->dl_type)) { + if (flow->dl_type == htons(ETH_TYPE_IP)) { + flow->nw_src = (OVS_FORCE ovs_be32)random_uint32(); + flow->nw_dst = (OVS_FORCE ovs_be32)random_uint32(); + } else { + random_bytes(&flow->ipv6_src, sizeof flow->ipv6_src); + random_bytes(&flow->ipv6_dst, sizeof flow->ipv6_dst); + } + /* Make most of IP flows TCP, some UDP or SCTP, and rest random. */ + rnd = random_uint16(); + flow->nw_proto = rnd < 0x8000 ? IPPROTO_TCP : + rnd < 0xc000 ? IPPROTO_UDP : + rnd < 0xd000 ? IPPROTO_SCTP : (uint8_t)rnd; + if (flow->nw_proto == IPPROTO_TCP || + flow->nw_proto == IPPROTO_UDP || + flow->nw_proto == IPPROTO_SCTP) { + flow->tp_src = (OVS_FORCE ovs_be16)random_uint16(); + flow->tp_dst = (OVS_FORCE ovs_be16)random_uint16(); + } + } +} + /* Masks the fields in 'wc' that are used by the flow hash 'fields'. */ void -flow_mask_hash_fields(struct flow_wildcards *wc, enum nx_hash_fields fields) +flow_mask_hash_fields(const struct flow *flow, struct flow_wildcards *wc, + enum nx_hash_fields fields) { switch (fields) { case NX_HASH_FIELDS_ETH_SRC: @@ -791,11 +845,18 @@ flow_mask_hash_fields(struct flow_wildcards *wc, enum nx_hash_fields fields) case NX_HASH_FIELDS_SYMMETRIC_L4: memset(&wc->masks.dl_src, 0xff, sizeof wc->masks.dl_src); memset(&wc->masks.dl_dst, 0xff, sizeof wc->masks.dl_dst); - memset(&wc->masks.nw_proto, 0xff, sizeof wc->masks.nw_proto); - memset(&wc->masks.nw_src, 0xff, sizeof wc->masks.nw_src); - memset(&wc->masks.nw_dst, 0xff, sizeof wc->masks.nw_dst); - memset(&wc->masks.tp_src, 0xff, sizeof wc->masks.tp_src); - memset(&wc->masks.tp_dst, 0xff, sizeof wc->masks.tp_dst); + if (flow->dl_type == htons(ETH_TYPE_IP)) { + memset(&wc->masks.nw_src, 0xff, sizeof wc->masks.nw_src); + memset(&wc->masks.nw_dst, 0xff, sizeof wc->masks.nw_dst); + } else if (flow->dl_type == htons(ETH_TYPE_IPV6)) { + memset(&wc->masks.ipv6_src, 0xff, sizeof wc->masks.ipv6_src); + memset(&wc->masks.ipv6_dst, 0xff, sizeof wc->masks.ipv6_dst); + } + if (is_ip_any(flow)) { + memset(&wc->masks.nw_proto, 0xff, sizeof wc->masks.nw_proto); + memset(&wc->masks.tp_src, 0xff, sizeof wc->masks.tp_src); + memset(&wc->masks.tp_dst, 0xff, sizeof wc->masks.tp_dst); + } wc->masks.vlan_tci |= htons(VLAN_VID_MASK | VLAN_CFI); break; @@ -966,8 +1027,8 @@ flow_compose(struct ofpbuf *b, const struct flow *flow) ip->ip_tos = flow->nw_tos; ip->ip_ttl = flow->nw_ttl; ip->ip_proto = flow->nw_proto; - ip->ip_src = flow->nw_src; - ip->ip_dst = flow->nw_dst; + put_16aligned_be32(&ip->ip_src, flow->nw_src); + put_16aligned_be32(&ip->ip_dst, flow->nw_dst); if (flow->nw_frag & FLOW_NW_FRAG_ANY) { ip->ip_frag_off |= htons(IP_MORE_FRAGMENTS); @@ -990,6 +1051,12 @@ flow_compose(struct ofpbuf *b, const struct flow *flow) 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; @@ -1019,8 +1086,8 @@ flow_compose(struct ofpbuf *b, const struct flow *flow) if (flow->nw_proto == ARP_OP_REQUEST || flow->nw_proto == ARP_OP_REPLY) { - arp->ar_spa = flow->nw_src; - arp->ar_tpa = flow->nw_dst; + put_16aligned_be32(&arp->ar_spa, flow->nw_src); + put_16aligned_be32(&arp->ar_tpa, flow->nw_dst); memcpy(arp->ar_sha, flow->arp_sha, ETH_ADDR_LEN); memcpy(arp->ar_tha, flow->arp_tha, ETH_ADDR_LEN); } @@ -1057,13 +1124,38 @@ miniflow_alloc_values(struct miniflow *flow, int n) } } +/* Completes an initialization of 'dst' as a miniflow copy of 'src' begun by + * the caller. The caller must have already initialized 'dst->map' properly + * to indicate the nonzero uint32_t elements of 'src'. 'n' must be the number + * of 1-bits in 'dst->map'. + * + * This function initializes 'dst->values' (either inline if possible or with + * malloc() otherwise) and copies the nonzero uint32_t elements of 'src' into + * it. */ +static void +miniflow_init__(struct miniflow *dst, const struct flow *src, int n) +{ + const uint32_t *src_u32 = (const uint32_t *) src; + unsigned int ofs; + int i; + + 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_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; @@ -1077,16 +1169,17 @@ miniflow_init(struct miniflow *dst, const struct flow *src) } } - /* Initialize dst->values. */ - dst->values = miniflow_alloc_values(dst, n); - ofs = 0; - for (i = 0; i < MINI_N_MAPS; i++) { - uint32_t map; + miniflow_init__(dst, src, n); +} - for (map = dst->map[i]; map; map = zero_rightmost_1bit(map)) { - dst->values[ofs++] = src_u32[raw_ctz(map) + i * 32]; - } - } +/* Initializes 'dst' as a copy of 'src', using 'mask->map' as 'dst''s map. The + * caller must eventually free 'dst' with miniflow_destroy(). */ +void +miniflow_init_with_minimask(struct miniflow *dst, const struct flow *src, + const struct minimask *mask) +{ + memcpy(dst->map, mask->masks.map, sizeof dst->map); + miniflow_init__(dst, src, miniflow_n_values(dst)); } /* Initializes 'dst' as a copy of 'src'. The caller must eventually free 'dst' @@ -1100,6 +1193,21 @@ miniflow_clone(struct miniflow *dst, const struct miniflow *src) memcpy(dst->values, src->values, n * sizeof *dst->values); } +/* Initializes 'dst' with the data in 'src', destroying 'src'. + * The caller must eventually free 'dst' with miniflow_destroy(). */ +void +miniflow_move(struct miniflow *dst, struct miniflow *src) +{ + if (src->values == src->inline_values) { + dst->values = dst->inline_values; + memcpy(dst->values, src->values, + miniflow_n_values(src) * sizeof *dst->values); + } else { + dst->values = src->values; + } + memcpy(dst->map, src->map, sizeof dst->map); +} + /* Frees any memory owned by 'flow'. Does not free the storage in which 'flow' * itself resides; the caller is responsible for that. */ void @@ -1169,16 +1277,35 @@ miniflow_get_vid(const struct miniflow *flow) bool miniflow_equal(const struct miniflow *a, const struct miniflow *b) { + const uint32_t *ap = a->values; + const uint32_t *bp = b->values; int i; for (i = 0; i < MINI_N_MAPS; i++) { - if (a->map[i] != b->map[i]) { - return false; + const uint32_t a_map = a->map[i]; + const uint32_t b_map = b->map[i]; + uint32_t map; + + if (a_map == b_map) { + for (map = a_map; map; map = zero_rightmost_1bit(map)) { + if (*ap++ != *bp++) { + return false; + } + } + } else { + for (map = a_map | b_map; map; map = zero_rightmost_1bit(map)) { + uint32_t bit = rightmost_1bit(map); + uint32_t a_value = a_map & bit ? *ap++ : 0; + uint32_t b_value = b_map & bit ? *bp++ : 0; + + if (a_value != b_value) { + return false; + } + } } } - return !memcmp(a->values, b->values, - miniflow_n_values(a) * sizeof *a->values); + return true; } /* Returns true if 'a' and 'b' are equal at the places where there are 1-bits @@ -1238,10 +1365,24 @@ miniflow_equal_flow_in_minimask(const struct miniflow *a, const struct flow *b, uint32_t miniflow_hash(const struct miniflow *flow, uint32_t basis) { - BUILD_ASSERT_DECL(MINI_N_MAPS == 2); - return hash_3words(flow->map[0], flow->map[1], - hash_words(flow->values, miniflow_n_values(flow), - basis)); + const uint32_t *p = flow->values; + uint32_t hash = basis; + int i; + + for (i = 0; i < MINI_N_MAPS; i++) { + uint32_t hash_map = 0; + uint32_t map; + + for (map = flow->map[i]; 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); + } + return mhash_finish(hash, p - flow->values); } /* Returns a hash value for the bits of 'flow' where there are 1-bits in @@ -1262,9 +1403,10 @@ miniflow_hash_in_minimask(const struct miniflow *flow, 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); + if (*p) { + int ofs = raw_ctz(map) + i * 32; + hash = mhash_add(hash, miniflow_get(flow, ofs) & *p); + } p++; } } @@ -1281,21 +1423,23 @@ uint32_t flow_hash_in_minimask(const struct flow *flow, const struct minimask *mask, uint32_t basis) { - const uint32_t *flow_u32 = (const uint32_t *) flow; + const uint32_t *flow_u32; const uint32_t *p = mask->masks.values; uint32_t hash; int i; hash = basis; + flow_u32 = (const uint32_t *) flow; 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); + if (*p) { + hash = mhash_add(hash, flow_u32[raw_ctz(map)] & *p); + } p++; } + flow_u32 += 32; } return mhash_finish(hash, (p - mask->masks.values) * 4); @@ -1317,6 +1461,14 @@ minimask_clone(struct minimask *dst, const struct minimask *src) miniflow_clone(&dst->masks, &src->masks); } +/* Initializes 'dst' with the data in 'src', destroying 'src'. + * The caller must eventually free 'dst' with minimask_destroy(). */ +void +minimask_move(struct minimask *dst, struct minimask *src) +{ + miniflow_move(&dst->masks, &src->masks); +} + /* Initializes 'dst_' as the bit-wise "and" of 'a_' and 'b_'. * * The caller must provide room for FLOW_U32S "uint32_t"s in 'storage', for use @@ -1428,7 +1580,17 @@ bool minimask_is_catchall(const struct minimask *mask_) { const struct miniflow *mask = &mask_->masks; + const uint32_t *p = mask->values; + int i; - BUILD_ASSERT(MINI_N_MAPS == 2); - return !(mask->map[0] | mask->map[1]); + for (i = 0; i < MINI_N_MAPS; i++) { + uint32_t map; + + for (map = mask->map[i]; map; map = zero_rightmost_1bit(map)) { + if (*p++) { + return false; + } + } + } + return true; }