X-Git-Url: https://pintos-os.org/cgi-bin/gitweb.cgi?a=blobdiff_plain;f=datapath%2Fflow.c;h=2d08c9d71a2e76256e1be5bce6a4eff22e28911f;hb=03a8a29eb87ef21d0e56c200ebe43878ef773d63;hp=2b80c6d820b8b78a11ae2f5ab17866ef7b5ec713;hpb=76abe283baa043bae84163873b0c7c498bfb260a;p=openvswitch diff --git a/datapath/flow.c b/datapath/flow.c index 2b80c6d8..2d08c9d7 100644 --- a/datapath/flow.c +++ b/datapath/flow.c @@ -29,6 +29,7 @@ #include #include #include +#include #include #include #include @@ -138,8 +139,8 @@ static int parse_ipv6hdr(struct sk_buff *skb, struct sw_flow_key *key, nexthdr = nh->nexthdr; payload_ofs = (u8 *)(nh + 1) - skb->data; - key->ip.nw_proto = NEXTHDR_NONE; - key->ip.nw_tos = ipv6_get_dsfield(nh) & ~INET_ECN_MASK; + key->ip.proto = NEXTHDR_NONE; + key->ip.tos = ipv6_get_dsfield(nh) & ~INET_ECN_MASK; ipv6_addr_copy(&key->ipv6.addr.src, &nh->saddr); ipv6_addr_copy(&key->ipv6.addr.dst, &nh->daddr); @@ -149,7 +150,7 @@ static int parse_ipv6hdr(struct sk_buff *skb, struct sw_flow_key *key, nh_len = payload_ofs - nh_ofs; skb_set_transport_header(skb, nh_ofs + nh_len); - key->ip.nw_proto = nexthdr; + key->ip.proto = nexthdr; return nh_len; } @@ -167,7 +168,7 @@ void flow_used(struct sw_flow *flow, struct sk_buff *skb) u8 tcp_flags = 0; if (flow->key.eth.type == htons(ETH_P_IP) && - flow->key.ip.nw_proto == IPPROTO_TCP) { + flow->key.ip.proto == IPPROTO_TCP) { u8 *tcp = (u8 *)tcp_hdr(skb); tcp_flags = *(tcp + TCP_FLAGS_OFFSET) & TCP_FLAG_MASK; } @@ -216,14 +217,153 @@ struct sw_flow *flow_alloc(void) return flow; } -void flow_free_tbl(struct tbl_node *node) +static struct hlist_head __rcu *find_bucket(struct flow_table * table, u32 hash) { - struct sw_flow *flow = flow_cast(node); + return flex_array_get(table->buckets, + (hash & (table->n_buckets - 1))); +} + +static struct flex_array __rcu *alloc_buckets(unsigned int n_buckets) +{ + struct flex_array __rcu * buckets; + int i, err; + + buckets = flex_array_alloc(sizeof(struct hlist_head *), + n_buckets, GFP_KERNEL); + if (!buckets) + return NULL; + + err = flex_array_prealloc(buckets, 0, n_buckets, GFP_KERNEL); + if (err) { + flex_array_free(buckets); + return NULL; + } + + for (i = 0; i < n_buckets; i++) + INIT_HLIST_HEAD((struct hlist_head *) + flex_array_get(buckets, i)); + + return buckets; +} + +static void free_buckets(struct flex_array * buckets) +{ + flex_array_free(buckets); +} + +struct flow_table *flow_tbl_alloc(int new_size) +{ + struct flow_table *table = kmalloc(sizeof(*table), GFP_KERNEL); + + if (!table) + return NULL; + + table->buckets = alloc_buckets(new_size); + if (!table->buckets) { + kfree(table); + return NULL; + } + table->n_buckets = new_size; + table->count = 0; + + return table; +} + +static void flow_free(struct sw_flow *flow) +{ flow->dead = true; flow_put(flow); } +void flow_tbl_destroy(struct flow_table *table) +{ + int i; + + if (!table) + return; + + for (i = 0; i < table->n_buckets; i++) { + struct sw_flow *flow; + struct hlist_head *head = flex_array_get(table->buckets, i); + struct hlist_node *node, *n; + + hlist_for_each_entry_safe(flow, node, n, head, hash_node) { + hlist_del_init_rcu(&flow->hash_node); + flow_free(flow); + } + } + + free_buckets(table->buckets); + kfree(table); +} + +static void flow_tbl_destroy_rcu_cb(struct rcu_head *rcu) +{ + struct flow_table *table = container_of(rcu, struct flow_table, rcu); + + flow_tbl_destroy(table); +} + +void flow_tbl_deferred_destroy(struct flow_table *table) +{ + if (!table) + return; + + call_rcu(&table->rcu, flow_tbl_destroy_rcu_cb); +} + +struct sw_flow *flow_tbl_next(struct flow_table *table, u32 *bucket, u32 *last) +{ + struct sw_flow *flow; + struct hlist_head *head; + struct hlist_node *n; + int i; + + while (*bucket < table->n_buckets) { + i = 0; + head = flex_array_get(table->buckets, *bucket); + hlist_for_each_entry_rcu(flow, n, head, hash_node) { + if (i < *last) { + i++; + continue; + } + *last = i + 1; + return flow; + } + (*bucket)++; + *last = 0; + } + + return NULL; +} + +struct flow_table *flow_tbl_expand(struct flow_table *table) +{ + struct flow_table *new_table; + int n_buckets = table->n_buckets * 2; + int i; + + new_table = flow_tbl_alloc(n_buckets); + if (!new_table) + return ERR_PTR(-ENOMEM); + + for (i = 0; i < table->n_buckets; i++) { + struct sw_flow *flow; + struct hlist_head *head; + struct hlist_node *n, *pos; + + head = flex_array_get(table->buckets, i); + + hlist_for_each_entry_safe(flow, n, pos, head, hash_node) { + hlist_del_init_rcu(&flow->hash_node); + flow_tbl_insert(new_table, flow); + } + } + + return new_table; +} + /* RCU callback used by flow_deferred_free. */ static void rcu_free_flow_callback(struct rcu_head *rcu) { @@ -485,40 +625,40 @@ int flow_extract(struct sk_buff *skb, u16 in_port, struct sw_flow_key *key, nh = ip_hdr(skb); key->ipv4.addr.src = nh->saddr; key->ipv4.addr.dst = nh->daddr; - key->ip.nw_tos = nh->tos & ~INET_ECN_MASK; - key->ip.nw_proto = nh->protocol; + key->ip.tos = nh->tos & ~INET_ECN_MASK; + key->ip.proto = nh->protocol; /* Transport layer. */ - if (!(nh->frag_off & htons(IP_MF | IP_OFFSET)) && - !(skb_shinfo(skb)->gso_type & SKB_GSO_UDP)) { - if (key->ip.nw_proto == IPPROTO_TCP) { - key_len = SW_FLOW_KEY_OFFSET(ipv4.tp); - if (tcphdr_ok(skb)) { - struct tcphdr *tcp = tcp_hdr(skb); - key->ipv4.tp.src = tcp->source; - key->ipv4.tp.dst = tcp->dest; - } - } else if (key->ip.nw_proto == IPPROTO_UDP) { - key_len = SW_FLOW_KEY_OFFSET(ipv4.tp); - if (udphdr_ok(skb)) { - struct udphdr *udp = udp_hdr(skb); - key->ipv4.tp.src = udp->source; - key->ipv4.tp.dst = udp->dest; - } - } else if (key->ip.nw_proto == IPPROTO_ICMP) { - key_len = SW_FLOW_KEY_OFFSET(ipv4.tp); - if (icmphdr_ok(skb)) { - struct icmphdr *icmp = icmp_hdr(skb); - /* The ICMP type and code fields use the 16-bit - * transport port fields, so we need to store them - * in 16-bit network byte order. */ - key->ipv4.tp.src = htons(icmp->type); - key->ipv4.tp.dst = htons(icmp->code); - } - } - } else + if ((nh->frag_off & htons(IP_MF | IP_OFFSET)) || + (skb_shinfo(skb)->gso_type & SKB_GSO_UDP)) *is_frag = true; + if (key->ip.proto == IPPROTO_TCP) { + key_len = SW_FLOW_KEY_OFFSET(ipv4.tp); + if (!*is_frag && tcphdr_ok(skb)) { + struct tcphdr *tcp = tcp_hdr(skb); + key->ipv4.tp.src = tcp->source; + key->ipv4.tp.dst = tcp->dest; + } + } else if (key->ip.proto == IPPROTO_UDP) { + key_len = SW_FLOW_KEY_OFFSET(ipv4.tp); + if (!*is_frag && udphdr_ok(skb)) { + struct udphdr *udp = udp_hdr(skb); + key->ipv4.tp.src = udp->source; + key->ipv4.tp.dst = udp->dest; + } + } else if (key->ip.proto == IPPROTO_ICMP) { + key_len = SW_FLOW_KEY_OFFSET(ipv4.tp); + if (!*is_frag && icmphdr_ok(skb)) { + struct icmphdr *icmp = icmp_hdr(skb); + /* The ICMP type and code fields use the 16-bit + * transport port fields, so we need to store them + * in 16-bit network byte order. */ + key->ipv4.tp.src = htons(icmp->type); + key->ipv4.tp.dst = htons(icmp->code); + } + } + } else if (key->eth.type == htons(ETH_P_ARP) && arphdr_ok(skb)) { struct arp_eth_header *arp; @@ -531,10 +671,10 @@ int flow_extract(struct sk_buff *skb, u16 in_port, struct sw_flow_key *key, /* We only match on the lower 8 bits of the opcode. */ if (ntohs(arp->ar_op) <= 0xff) - key->ip.nw_proto = ntohs(arp->ar_op); + key->ip.proto = ntohs(arp->ar_op); - if (key->ip.nw_proto == ARPOP_REQUEST - || key->ip.nw_proto == ARPOP_REPLY) { + if (key->ip.proto == ARPOP_REQUEST + || key->ip.proto == ARPOP_REPLY) { memcpy(&key->ipv4.addr.src, arp->ar_sip, sizeof(key->ipv4.addr.src)); memcpy(&key->ipv4.addr.dst, arp->ar_tip, sizeof(key->ipv4.addr.dst)); memcpy(key->ipv4.arp.sha, arp->ar_sha, ETH_ALEN); @@ -555,21 +695,21 @@ int flow_extract(struct sk_buff *skb, u16 in_port, struct sw_flow_key *key, } /* Transport layer. */ - if (key->ip.nw_proto == NEXTHDR_TCP) { + if (key->ip.proto == NEXTHDR_TCP) { key_len = SW_FLOW_KEY_OFFSET(ipv6.tp); if (tcphdr_ok(skb)) { struct tcphdr *tcp = tcp_hdr(skb); key->ipv6.tp.src = tcp->source; key->ipv6.tp.dst = tcp->dest; } - } else if (key->ip.nw_proto == NEXTHDR_UDP) { + } else if (key->ip.proto == NEXTHDR_UDP) { key_len = SW_FLOW_KEY_OFFSET(ipv6.tp); if (udphdr_ok(skb)) { struct udphdr *udp = udp_hdr(skb); key->ipv6.tp.src = udp->source; key->ipv6.tp.dst = udp->dest; } - } else if (key->ip.nw_proto == NEXTHDR_ICMP) { + } else if (key->ip.proto == NEXTHDR_ICMP) { key_len = SW_FLOW_KEY_OFFSET(ipv6.tp); if (icmp6hdr_ok(skb)) { error = parse_icmpv6(skb, key, &key_len, nh_len); @@ -589,25 +729,73 @@ u32 flow_hash(const struct sw_flow_key *key, int key_len) return jhash2((u32*)key, DIV_ROUND_UP(key_len, sizeof(u32)), hash_seed); } -int flow_cmp(const struct tbl_node *node, void *key2_, int len) +struct sw_flow * flow_tbl_lookup(struct flow_table *table, + struct sw_flow_key *key, int key_len) +{ + struct sw_flow *flow; + struct hlist_node *n; + struct hlist_head *head; + u32 hash; + + hash = flow_hash(key, key_len); + + head = find_bucket(table, hash); + hlist_for_each_entry_rcu(flow, n, head, hash_node) { + + if (flow->hash == hash && + !memcmp(&flow->key, key, key_len)) { + return flow; + } + } + return NULL; +} + +void flow_tbl_insert(struct flow_table *table, struct sw_flow *flow) { - const struct sw_flow_key *key1 = &flow_cast(node)->key; - const struct sw_flow_key *key2 = key2_; + struct hlist_head *head; + + head = find_bucket(table, flow->hash); + hlist_add_head_rcu(&flow->hash_node, head); + table->count++; +} - return !memcmp(key1, key2, len); +void flow_tbl_remove(struct flow_table *table, struct sw_flow *flow) +{ + if (!hlist_unhashed(&flow->hash_node)) { + hlist_del_init_rcu(&flow->hash_node); + table->count--; + BUG_ON(table->count < 0); + } } +/* The size of the argument for each %OVS_KEY_ATTR_* Netlink attribute. */ +static const u32 key_lens[OVS_KEY_ATTR_MAX + 1] = { + [OVS_KEY_ATTR_TUN_ID] = 8, + [OVS_KEY_ATTR_IN_PORT] = 4, + [OVS_KEY_ATTR_ETHERNET] = sizeof(struct ovs_key_ethernet), + [OVS_KEY_ATTR_8021Q] = sizeof(struct ovs_key_8021q), + [OVS_KEY_ATTR_ETHERTYPE] = 2, + [OVS_KEY_ATTR_IPV4] = sizeof(struct ovs_key_ipv4), + [OVS_KEY_ATTR_IPV6] = sizeof(struct ovs_key_ipv6), + [OVS_KEY_ATTR_TCP] = sizeof(struct ovs_key_tcp), + [OVS_KEY_ATTR_UDP] = sizeof(struct ovs_key_udp), + [OVS_KEY_ATTR_ICMP] = sizeof(struct ovs_key_icmp), + [OVS_KEY_ATTR_ICMPV6] = sizeof(struct ovs_key_icmpv6), + [OVS_KEY_ATTR_ARP] = sizeof(struct ovs_key_arp), + [OVS_KEY_ATTR_ND] = sizeof(struct ovs_key_nd), +}; + /** * flow_from_nlattrs - parses Netlink attributes into a flow key. * @swkey: receives the extracted flow key. * @key_lenp: number of bytes used in @swkey. - * @attr: Netlink attribute holding nested %ODP_KEY_ATTR_* Netlink attribute + * @attr: Netlink attribute holding nested %OVS_KEY_ATTR_* Netlink attribute * sequence. * * This state machine accepts the following forms, with [] for optional * elements and | for alternatives: * - * [tun_id] in_port ethernet [8021q] [ethertype \ + * [tun_id] [in_port] ethernet [8021q] [ethertype \ * [IPv4 [TCP|UDP|ICMP] | IPv6 [TCP|UDP|ICMPv6 [ND]] | ARP]] */ int flow_from_nlattrs(struct sw_flow_key *swkey, int *key_lenp, @@ -620,63 +808,50 @@ int flow_from_nlattrs(struct sw_flow_key *swkey, int *key_lenp, int key_len; memset(swkey, 0, sizeof(*swkey)); + swkey->eth.in_port = USHRT_MAX; swkey->eth.type = htons(ETH_P_802_2); key_len = SW_FLOW_KEY_OFFSET(eth); - prev_type = ODP_KEY_ATTR_UNSPEC; + prev_type = OVS_KEY_ATTR_UNSPEC; nla_for_each_nested(nla, attr, rem) { - static const u32 key_lens[ODP_KEY_ATTR_MAX + 1] = { - [ODP_KEY_ATTR_TUN_ID] = 8, - [ODP_KEY_ATTR_IN_PORT] = 4, - [ODP_KEY_ATTR_ETHERNET] = sizeof(struct odp_key_ethernet), - [ODP_KEY_ATTR_8021Q] = sizeof(struct odp_key_8021q), - [ODP_KEY_ATTR_ETHERTYPE] = 2, - [ODP_KEY_ATTR_IPV4] = sizeof(struct odp_key_ipv4), - [ODP_KEY_ATTR_IPV6] = sizeof(struct odp_key_ipv6), - [ODP_KEY_ATTR_TCP] = sizeof(struct odp_key_tcp), - [ODP_KEY_ATTR_UDP] = sizeof(struct odp_key_udp), - [ODP_KEY_ATTR_ICMP] = sizeof(struct odp_key_icmp), - [ODP_KEY_ATTR_ICMPV6] = sizeof(struct odp_key_icmpv6), - [ODP_KEY_ATTR_ARP] = sizeof(struct odp_key_arp), - [ODP_KEY_ATTR_ND] = sizeof(struct odp_key_nd), - }; - - const struct odp_key_ethernet *eth_key; - const struct odp_key_8021q *q_key; - const struct odp_key_ipv4 *ipv4_key; - const struct odp_key_ipv6 *ipv6_key; - const struct odp_key_tcp *tcp_key; - const struct odp_key_udp *udp_key; - const struct odp_key_icmp *icmp_key; - const struct odp_key_icmpv6 *icmpv6_key; - const struct odp_key_arp *arp_key; - const struct odp_key_nd *nd_key; + const struct ovs_key_ethernet *eth_key; + const struct ovs_key_8021q *q_key; + const struct ovs_key_ipv4 *ipv4_key; + const struct ovs_key_ipv6 *ipv6_key; + const struct ovs_key_tcp *tcp_key; + const struct ovs_key_udp *udp_key; + const struct ovs_key_icmp *icmp_key; + const struct ovs_key_icmpv6 *icmpv6_key; + const struct ovs_key_arp *arp_key; + const struct ovs_key_nd *nd_key; int type = nla_type(nla); - if (type > ODP_KEY_ATTR_MAX || nla_len(nla) != key_lens[type]) + if (type > OVS_KEY_ATTR_MAX || nla_len(nla) != key_lens[type]) goto invalid; #define TRANSITION(PREV_TYPE, TYPE) (((PREV_TYPE) << 16) | (TYPE)) switch (TRANSITION(prev_type, type)) { - case TRANSITION(ODP_KEY_ATTR_UNSPEC, ODP_KEY_ATTR_TUN_ID): + case TRANSITION(OVS_KEY_ATTR_UNSPEC, OVS_KEY_ATTR_TUN_ID): swkey->eth.tun_id = nla_get_be64(nla); break; - case TRANSITION(ODP_KEY_ATTR_UNSPEC, ODP_KEY_ATTR_IN_PORT): - case TRANSITION(ODP_KEY_ATTR_TUN_ID, ODP_KEY_ATTR_IN_PORT): + case TRANSITION(OVS_KEY_ATTR_UNSPEC, OVS_KEY_ATTR_IN_PORT): + case TRANSITION(OVS_KEY_ATTR_TUN_ID, OVS_KEY_ATTR_IN_PORT): if (nla_get_u32(nla) >= DP_MAX_PORTS) goto invalid; swkey->eth.in_port = nla_get_u32(nla); break; - case TRANSITION(ODP_KEY_ATTR_IN_PORT, ODP_KEY_ATTR_ETHERNET): + case TRANSITION(OVS_KEY_ATTR_UNSPEC, OVS_KEY_ATTR_ETHERNET): + case TRANSITION(OVS_KEY_ATTR_TUN_ID, OVS_KEY_ATTR_ETHERNET): + case TRANSITION(OVS_KEY_ATTR_IN_PORT, OVS_KEY_ATTR_ETHERNET): eth_key = nla_data(nla); memcpy(swkey->eth.src, eth_key->eth_src, ETH_ALEN); memcpy(swkey->eth.dst, eth_key->eth_dst, ETH_ALEN); break; - case TRANSITION(ODP_KEY_ATTR_ETHERNET, ODP_KEY_ATTR_8021Q): + case TRANSITION(OVS_KEY_ATTR_ETHERNET, OVS_KEY_ATTR_8021Q): q_key = nla_data(nla); /* Only standard 0x8100 VLANs currently supported. */ if (q_key->q_tpid != htons(ETH_P_8021Q)) @@ -686,96 +861,96 @@ int flow_from_nlattrs(struct sw_flow_key *swkey, int *key_lenp, swkey->eth.tci = q_key->q_tci | htons(VLAN_TAG_PRESENT); break; - case TRANSITION(ODP_KEY_ATTR_8021Q, ODP_KEY_ATTR_ETHERTYPE): - case TRANSITION(ODP_KEY_ATTR_ETHERNET, ODP_KEY_ATTR_ETHERTYPE): + case TRANSITION(OVS_KEY_ATTR_8021Q, OVS_KEY_ATTR_ETHERTYPE): + case TRANSITION(OVS_KEY_ATTR_ETHERNET, OVS_KEY_ATTR_ETHERTYPE): swkey->eth.type = nla_get_be16(nla); if (ntohs(swkey->eth.type) < 1536) goto invalid; break; - case TRANSITION(ODP_KEY_ATTR_ETHERTYPE, ODP_KEY_ATTR_IPV4): + case TRANSITION(OVS_KEY_ATTR_ETHERTYPE, OVS_KEY_ATTR_IPV4): key_len = SW_FLOW_KEY_OFFSET(ipv4.addr); if (swkey->eth.type != htons(ETH_P_IP)) goto invalid; ipv4_key = nla_data(nla); - swkey->ip.nw_proto = ipv4_key->ipv4_proto; - swkey->ip.nw_tos = ipv4_key->ipv4_tos; + swkey->ip.proto = ipv4_key->ipv4_proto; + swkey->ip.tos = ipv4_key->ipv4_tos; swkey->ipv4.addr.src = ipv4_key->ipv4_src; swkey->ipv4.addr.dst = ipv4_key->ipv4_dst; - if (swkey->ip.nw_tos & INET_ECN_MASK) + if (swkey->ip.tos & INET_ECN_MASK) goto invalid; break; - case TRANSITION(ODP_KEY_ATTR_ETHERTYPE, ODP_KEY_ATTR_IPV6): + case TRANSITION(OVS_KEY_ATTR_ETHERTYPE, OVS_KEY_ATTR_IPV6): key_len = SW_FLOW_KEY_OFFSET(ipv6.addr); if (swkey->eth.type != htons(ETH_P_IPV6)) goto invalid; ipv6_key = nla_data(nla); - swkey->ip.nw_proto = ipv6_key->ipv6_proto; - swkey->ip.nw_tos = ipv6_key->ipv6_tos; + swkey->ip.proto = ipv6_key->ipv6_proto; + swkey->ip.tos = ipv6_key->ipv6_tos; memcpy(&swkey->ipv6.addr.src, ipv6_key->ipv6_src, sizeof(swkey->ipv6.addr.src)); memcpy(&swkey->ipv6.addr.dst, ipv6_key->ipv6_dst, sizeof(swkey->ipv6.addr.dst)); - if (swkey->ip.nw_tos & INET_ECN_MASK) + if (swkey->ip.tos & INET_ECN_MASK) goto invalid; break; - case TRANSITION(ODP_KEY_ATTR_IPV4, ODP_KEY_ATTR_TCP): + case TRANSITION(OVS_KEY_ATTR_IPV4, OVS_KEY_ATTR_TCP): key_len = SW_FLOW_KEY_OFFSET(ipv4.tp); - if (swkey->ip.nw_proto != IPPROTO_TCP) + if (swkey->ip.proto != IPPROTO_TCP) goto invalid; tcp_key = nla_data(nla); swkey->ipv4.tp.src = tcp_key->tcp_src; swkey->ipv4.tp.dst = tcp_key->tcp_dst; break; - case TRANSITION(ODP_KEY_ATTR_IPV6, ODP_KEY_ATTR_TCP): + case TRANSITION(OVS_KEY_ATTR_IPV6, OVS_KEY_ATTR_TCP): key_len = SW_FLOW_KEY_OFFSET(ipv6.tp); - if (swkey->ip.nw_proto != IPPROTO_TCP) + if (swkey->ip.proto != IPPROTO_TCP) goto invalid; tcp_key = nla_data(nla); swkey->ipv6.tp.src = tcp_key->tcp_src; swkey->ipv6.tp.dst = tcp_key->tcp_dst; break; - case TRANSITION(ODP_KEY_ATTR_IPV4, ODP_KEY_ATTR_UDP): + case TRANSITION(OVS_KEY_ATTR_IPV4, OVS_KEY_ATTR_UDP): key_len = SW_FLOW_KEY_OFFSET(ipv4.tp); - if (swkey->ip.nw_proto != IPPROTO_UDP) + if (swkey->ip.proto != IPPROTO_UDP) goto invalid; udp_key = nla_data(nla); swkey->ipv4.tp.src = udp_key->udp_src; swkey->ipv4.tp.dst = udp_key->udp_dst; break; - case TRANSITION(ODP_KEY_ATTR_IPV6, ODP_KEY_ATTR_UDP): + case TRANSITION(OVS_KEY_ATTR_IPV6, OVS_KEY_ATTR_UDP): key_len = SW_FLOW_KEY_OFFSET(ipv6.tp); - if (swkey->ip.nw_proto != IPPROTO_UDP) + if (swkey->ip.proto != IPPROTO_UDP) goto invalid; udp_key = nla_data(nla); swkey->ipv6.tp.src = udp_key->udp_src; swkey->ipv6.tp.dst = udp_key->udp_dst; break; - case TRANSITION(ODP_KEY_ATTR_IPV4, ODP_KEY_ATTR_ICMP): + case TRANSITION(OVS_KEY_ATTR_IPV4, OVS_KEY_ATTR_ICMP): key_len = SW_FLOW_KEY_OFFSET(ipv4.tp); - if (swkey->ip.nw_proto != IPPROTO_ICMP) + if (swkey->ip.proto != IPPROTO_ICMP) goto invalid; icmp_key = nla_data(nla); swkey->ipv4.tp.src = htons(icmp_key->icmp_type); swkey->ipv4.tp.dst = htons(icmp_key->icmp_code); break; - case TRANSITION(ODP_KEY_ATTR_IPV6, ODP_KEY_ATTR_ICMPV6): + case TRANSITION(OVS_KEY_ATTR_IPV6, OVS_KEY_ATTR_ICMPV6): key_len = SW_FLOW_KEY_OFFSET(ipv6.tp); - if (swkey->ip.nw_proto != IPPROTO_ICMPV6) + if (swkey->ip.proto != IPPROTO_ICMPV6) goto invalid; icmpv6_key = nla_data(nla); swkey->ipv6.tp.src = htons(icmpv6_key->icmpv6_type); swkey->ipv6.tp.dst = htons(icmpv6_key->icmpv6_code); break; - case TRANSITION(ODP_KEY_ATTR_ETHERTYPE, ODP_KEY_ATTR_ARP): + case TRANSITION(OVS_KEY_ATTR_ETHERTYPE, OVS_KEY_ATTR_ARP): key_len = SW_FLOW_KEY_OFFSET(ipv4.arp); if (swkey->eth.type != htons(ETH_P_ARP)) goto invalid; @@ -784,12 +959,12 @@ int flow_from_nlattrs(struct sw_flow_key *swkey, int *key_lenp, swkey->ipv4.addr.dst = arp_key->arp_tip; if (arp_key->arp_op & htons(0xff00)) goto invalid; - swkey->ip.nw_proto = ntohs(arp_key->arp_op); + swkey->ip.proto = ntohs(arp_key->arp_op); memcpy(swkey->ipv4.arp.sha, arp_key->arp_sha, ETH_ALEN); memcpy(swkey->ipv4.arp.tha, arp_key->arp_tha, ETH_ALEN); break; - case TRANSITION(ODP_KEY_ATTR_ICMPV6, ODP_KEY_ATTR_ND): + case TRANSITION(OVS_KEY_ATTR_ICMPV6, OVS_KEY_ATTR_ND): key_len = SW_FLOW_KEY_OFFSET(ipv6.nd); if (swkey->ipv6.tp.src != htons(NDISC_NEIGHBOUR_SOLICITATION) && swkey->ipv6.tp.src != htons(NDISC_NEIGHBOUR_ADVERTISEMENT)) @@ -811,48 +986,48 @@ int flow_from_nlattrs(struct sw_flow_key *swkey, int *key_lenp, goto invalid; switch (prev_type) { - case ODP_KEY_ATTR_UNSPEC: + case OVS_KEY_ATTR_UNSPEC: goto invalid; - case ODP_KEY_ATTR_TUN_ID: - case ODP_KEY_ATTR_IN_PORT: + case OVS_KEY_ATTR_TUN_ID: + case OVS_KEY_ATTR_IN_PORT: goto invalid; - case ODP_KEY_ATTR_ETHERNET: - case ODP_KEY_ATTR_8021Q: + case OVS_KEY_ATTR_ETHERNET: + case OVS_KEY_ATTR_8021Q: goto ok; - case ODP_KEY_ATTR_ETHERTYPE: + case OVS_KEY_ATTR_ETHERTYPE: if (swkey->eth.type == htons(ETH_P_IP) || swkey->eth.type == htons(ETH_P_ARP)) goto invalid; goto ok; - case ODP_KEY_ATTR_IPV4: - if (swkey->ip.nw_proto == IPPROTO_TCP || - swkey->ip.nw_proto == IPPROTO_UDP || - swkey->ip.nw_proto == IPPROTO_ICMP) + case OVS_KEY_ATTR_IPV4: + if (swkey->ip.proto == IPPROTO_TCP || + swkey->ip.proto == IPPROTO_UDP || + swkey->ip.proto == IPPROTO_ICMP) goto invalid; goto ok; - case ODP_KEY_ATTR_IPV6: - if (swkey->ip.nw_proto == IPPROTO_TCP || - swkey->ip.nw_proto == IPPROTO_UDP || - swkey->ip.nw_proto == IPPROTO_ICMPV6) + case OVS_KEY_ATTR_IPV6: + if (swkey->ip.proto == IPPROTO_TCP || + swkey->ip.proto == IPPROTO_UDP || + swkey->ip.proto == IPPROTO_ICMPV6) goto invalid; goto ok; - case ODP_KEY_ATTR_ICMPV6: + case OVS_KEY_ATTR_ICMPV6: if (swkey->ipv6.tp.src == htons(NDISC_NEIGHBOUR_SOLICITATION) || swkey->ipv6.tp.src == htons(NDISC_NEIGHBOUR_ADVERTISEMENT)) goto invalid; goto ok; - case ODP_KEY_ATTR_TCP: - case ODP_KEY_ATTR_UDP: - case ODP_KEY_ATTR_ICMP: - case ODP_KEY_ATTR_ARP: - case ODP_KEY_ATTR_ND: + case OVS_KEY_ATTR_TCP: + case OVS_KEY_ATTR_UDP: + case OVS_KEY_ATTR_ICMP: + case OVS_KEY_ATTR_ARP: + case OVS_KEY_ATTR_ND: goto ok; default: @@ -868,22 +1043,75 @@ ok: return error; } +/** + * flow_metadata_from_nlattrs - parses Netlink attributes into a flow key. + * @in_port: receives the extracted input port. + * @tun_id: receives the extracted tunnel ID. + * @key: Netlink attribute holding nested %OVS_KEY_ATTR_* Netlink attribute + * sequence. + * + * This parses a series of Netlink attributes that form a flow key, which must + * take the same form accepted by flow_from_nlattrs(), but only enough of it to + * get the metadata, that is, the parts of the flow key that cannot be + * extracted from the packet itself. + */ +int flow_metadata_from_nlattrs(u16 *in_port, __be64 *tun_id, + const struct nlattr *attr) +{ + const struct nlattr *nla; + u16 prev_type; + int rem; + + *in_port = USHRT_MAX; + *tun_id = 0; + + prev_type = OVS_KEY_ATTR_UNSPEC; + nla_for_each_nested(nla, attr, rem) { + int type = nla_type(nla); + + if (type > OVS_KEY_ATTR_MAX || nla_len(nla) != key_lens[type]) + return -EINVAL; + + switch (TRANSITION(prev_type, type)) { + case TRANSITION(OVS_KEY_ATTR_UNSPEC, OVS_KEY_ATTR_TUN_ID): + *tun_id = nla_get_be64(nla); + break; + + case TRANSITION(OVS_KEY_ATTR_UNSPEC, OVS_KEY_ATTR_IN_PORT): + case TRANSITION(OVS_KEY_ATTR_TUN_ID, OVS_KEY_ATTR_IN_PORT): + if (nla_get_u32(nla) >= DP_MAX_PORTS) + return -EINVAL; + *in_port = nla_get_u32(nla); + break; + + default: + return 0; + } + + prev_type = type; + } + if (rem) + return -EINVAL; + return 0; +} + int flow_to_nlattrs(const struct sw_flow_key *swkey, struct sk_buff *skb) { - struct odp_key_ethernet *eth_key; + struct ovs_key_ethernet *eth_key; struct nlattr *nla; /* This is an imperfect sanity-check that FLOW_BUFSIZE doesn't need - * to be updated, but will at least raise awareness when new ODP key - * types are added. */ - BUILD_BUG_ON(__ODP_KEY_ATTR_MAX != 14); + * to be updated, but will at least raise awareness when new + * datapath key types are added. */ + BUILD_BUG_ON(__OVS_KEY_ATTR_MAX != 14); if (swkey->eth.tun_id != cpu_to_be64(0)) - NLA_PUT_BE64(skb, ODP_KEY_ATTR_TUN_ID, swkey->eth.tun_id); + NLA_PUT_BE64(skb, OVS_KEY_ATTR_TUN_ID, swkey->eth.tun_id); - NLA_PUT_U32(skb, ODP_KEY_ATTR_IN_PORT, swkey->eth.in_port); + if (swkey->eth.in_port != USHRT_MAX) + NLA_PUT_U32(skb, OVS_KEY_ATTR_IN_PORT, swkey->eth.in_port); - nla = nla_reserve(skb, ODP_KEY_ATTR_ETHERNET, sizeof(*eth_key)); + nla = nla_reserve(skb, OVS_KEY_ATTR_ETHERNET, sizeof(*eth_key)); if (!nla) goto nla_put_failure; eth_key = nla_data(nla); @@ -891,55 +1119,55 @@ int flow_to_nlattrs(const struct sw_flow_key *swkey, struct sk_buff *skb) memcpy(eth_key->eth_dst, swkey->eth.dst, ETH_ALEN); if (swkey->eth.tci != htons(0)) { - struct odp_key_8021q q_key; + struct ovs_key_8021q q_key; q_key.q_tpid = htons(ETH_P_8021Q); q_key.q_tci = swkey->eth.tci & ~htons(VLAN_TAG_PRESENT); - NLA_PUT(skb, ODP_KEY_ATTR_8021Q, sizeof(q_key), &q_key); + NLA_PUT(skb, OVS_KEY_ATTR_8021Q, sizeof(q_key), &q_key); } if (swkey->eth.type == htons(ETH_P_802_2)) return 0; - NLA_PUT_BE16(skb, ODP_KEY_ATTR_ETHERTYPE, swkey->eth.type); + NLA_PUT_BE16(skb, OVS_KEY_ATTR_ETHERTYPE, swkey->eth.type); if (swkey->eth.type == htons(ETH_P_IP)) { - struct odp_key_ipv4 *ipv4_key; + struct ovs_key_ipv4 *ipv4_key; - nla = nla_reserve(skb, ODP_KEY_ATTR_IPV4, sizeof(*ipv4_key)); + nla = nla_reserve(skb, OVS_KEY_ATTR_IPV4, sizeof(*ipv4_key)); if (!nla) goto nla_put_failure; ipv4_key = nla_data(nla); - memset(ipv4_key, 0, sizeof(struct odp_key_ipv4)); + memset(ipv4_key, 0, sizeof(struct ovs_key_ipv4)); ipv4_key->ipv4_src = swkey->ipv4.addr.src; ipv4_key->ipv4_dst = swkey->ipv4.addr.dst; - ipv4_key->ipv4_proto = swkey->ip.nw_proto; - ipv4_key->ipv4_tos = swkey->ip.nw_tos; + ipv4_key->ipv4_proto = swkey->ip.proto; + ipv4_key->ipv4_tos = swkey->ip.tos; } else if (swkey->eth.type == htons(ETH_P_IPV6)) { - struct odp_key_ipv6 *ipv6_key; + struct ovs_key_ipv6 *ipv6_key; - nla = nla_reserve(skb, ODP_KEY_ATTR_IPV6, sizeof(*ipv6_key)); + nla = nla_reserve(skb, OVS_KEY_ATTR_IPV6, sizeof(*ipv6_key)); if (!nla) goto nla_put_failure; ipv6_key = nla_data(nla); - memset(ipv6_key, 0, sizeof(struct odp_key_ipv6)); + memset(ipv6_key, 0, sizeof(struct ovs_key_ipv6)); memcpy(ipv6_key->ipv6_src, &swkey->ipv6.addr.src, sizeof(ipv6_key->ipv6_src)); memcpy(ipv6_key->ipv6_dst, &swkey->ipv6.addr.dst, sizeof(ipv6_key->ipv6_dst)); - ipv6_key->ipv6_proto = swkey->ip.nw_proto; - ipv6_key->ipv6_tos = swkey->ip.nw_tos; + ipv6_key->ipv6_proto = swkey->ip.proto; + ipv6_key->ipv6_tos = swkey->ip.tos; } else if (swkey->eth.type == htons(ETH_P_ARP)) { - struct odp_key_arp *arp_key; + struct ovs_key_arp *arp_key; - nla = nla_reserve(skb, ODP_KEY_ATTR_ARP, sizeof(*arp_key)); + nla = nla_reserve(skb, OVS_KEY_ATTR_ARP, sizeof(*arp_key)); if (!nla) goto nla_put_failure; arp_key = nla_data(nla); - memset(arp_key, 0, sizeof(struct odp_key_arp)); + memset(arp_key, 0, sizeof(struct ovs_key_arp)); arp_key->arp_sip = swkey->ipv4.addr.src; arp_key->arp_tip = swkey->ipv4.addr.dst; - arp_key->arp_op = htons(swkey->ip.nw_proto); + arp_key->arp_op = htons(swkey->ip.proto); memcpy(arp_key->arp_sha, swkey->ipv4.arp.sha, ETH_ALEN); memcpy(arp_key->arp_tha, swkey->ipv4.arp.tha, ETH_ALEN); } @@ -947,10 +1175,10 @@ int flow_to_nlattrs(const struct sw_flow_key *swkey, struct sk_buff *skb) if (swkey->eth.type == htons(ETH_P_IP) || swkey->eth.type == htons(ETH_P_IPV6)) { - if (swkey->ip.nw_proto == IPPROTO_TCP) { - struct odp_key_tcp *tcp_key; + if (swkey->ip.proto == IPPROTO_TCP) { + struct ovs_key_tcp *tcp_key; - nla = nla_reserve(skb, ODP_KEY_ATTR_TCP, sizeof(*tcp_key)); + nla = nla_reserve(skb, OVS_KEY_ATTR_TCP, sizeof(*tcp_key)); if (!nla) goto nla_put_failure; tcp_key = nla_data(nla); @@ -961,10 +1189,10 @@ int flow_to_nlattrs(const struct sw_flow_key *swkey, struct sk_buff *skb) tcp_key->tcp_src = swkey->ipv6.tp.src; tcp_key->tcp_dst = swkey->ipv6.tp.dst; } - } else if (swkey->ip.nw_proto == IPPROTO_UDP) { - struct odp_key_udp *udp_key; + } else if (swkey->ip.proto == IPPROTO_UDP) { + struct ovs_key_udp *udp_key; - nla = nla_reserve(skb, ODP_KEY_ATTR_UDP, sizeof(*udp_key)); + nla = nla_reserve(skb, OVS_KEY_ATTR_UDP, sizeof(*udp_key)); if (!nla) goto nla_put_failure; udp_key = nla_data(nla); @@ -976,20 +1204,20 @@ int flow_to_nlattrs(const struct sw_flow_key *swkey, struct sk_buff *skb) udp_key->udp_dst = swkey->ipv6.tp.dst; } } else if (swkey->eth.type == htons(ETH_P_IP) && - swkey->ip.nw_proto == IPPROTO_ICMP) { - struct odp_key_icmp *icmp_key; + swkey->ip.proto == IPPROTO_ICMP) { + struct ovs_key_icmp *icmp_key; - nla = nla_reserve(skb, ODP_KEY_ATTR_ICMP, sizeof(*icmp_key)); + nla = nla_reserve(skb, OVS_KEY_ATTR_ICMP, sizeof(*icmp_key)); if (!nla) goto nla_put_failure; icmp_key = nla_data(nla); icmp_key->icmp_type = ntohs(swkey->ipv4.tp.src); icmp_key->icmp_code = ntohs(swkey->ipv4.tp.dst); } else if (swkey->eth.type == htons(ETH_P_IPV6) && - swkey->ip.nw_proto == IPPROTO_ICMPV6) { - struct odp_key_icmpv6 *icmpv6_key; + swkey->ip.proto == IPPROTO_ICMPV6) { + struct ovs_key_icmpv6 *icmpv6_key; - nla = nla_reserve(skb, ODP_KEY_ATTR_ICMPV6, + nla = nla_reserve(skb, OVS_KEY_ATTR_ICMPV6, sizeof(*icmpv6_key)); if (!nla) goto nla_put_failure; @@ -999,9 +1227,9 @@ int flow_to_nlattrs(const struct sw_flow_key *swkey, struct sk_buff *skb) if (icmpv6_key->icmpv6_type == NDISC_NEIGHBOUR_SOLICITATION || icmpv6_key->icmpv6_type == NDISC_NEIGHBOUR_ADVERTISEMENT) { - struct odp_key_nd *nd_key; + struct ovs_key_nd *nd_key; - nla = nla_reserve(skb, ODP_KEY_ATTR_ND, sizeof(*nd_key)); + nla = nla_reserve(skb, OVS_KEY_ATTR_ND, sizeof(*nd_key)); if (!nla) goto nla_put_failure; nd_key = nla_data(nla);