X-Git-Url: http://git.onelab.eu/?a=blobdiff_plain;f=datapath%2Fflow.c;h=f264866a7eef793a107f24f3cd919212528672cd;hb=a328a943f173391cd9a9a54e257c8dabcd463402;hp=8a6ee23d3e833ce0a5761a321f291c670df116be;hpb=cdee00fd635d1e0f1eeb5d9c009daeb59abd4777;p=sliver-openvswitch.git diff --git a/datapath/flow.c b/datapath/flow.c index 8a6ee23d3..f264866a7 100644 --- a/datapath/flow.c +++ b/datapath/flow.c @@ -1,6 +1,6 @@ /* * Distributed under the terms of the GNU GPL version 2. - * Copyright (c) 2007, 2008, 2009, 2010 Nicira Networks. + * Copyright (c) 2007, 2008, 2009, 2010, 2011 Nicira Networks. * * Significant portions of this file may be copied from parts of the Linux * kernel, by Linus Torvalds and others. @@ -8,6 +8,7 @@ #include "flow.h" #include "datapath.h" +#include #include #include #include @@ -23,13 +24,17 @@ #include #include #include +#include #include #include #include +#include #include #include +#include +#include -#include "compat.h" +#include "vlan.h" static struct kmem_cache *flow_cache; static unsigned int hash_seed __read_mostly; @@ -83,6 +88,63 @@ static inline bool icmphdr_ok(struct sk_buff *skb) return skb->len >= skb_transport_offset(skb) + sizeof(struct icmphdr); } +u64 flow_used_time(unsigned long flow_jiffies) +{ + struct timespec cur_ts; + u64 cur_ms, idle_ms; + + ktime_get_ts(&cur_ts); + idle_ms = jiffies_to_msecs(jiffies - flow_jiffies); + cur_ms = (u64)cur_ts.tv_sec * MSEC_PER_SEC + + cur_ts.tv_nsec / NSEC_PER_MSEC; + + return cur_ms - idle_ms; +} + +static int parse_ipv6hdr(struct sk_buff *skb, struct sw_flow_key *key) +{ + unsigned int nh_ofs = skb_network_offset(skb); + unsigned int nh_len; + int payload_ofs; + struct ipv6hdr *nh; + uint8_t nexthdr; + + if (unlikely(skb->len < nh_ofs + sizeof(*nh))) + return -EINVAL; + + nh = ipv6_hdr(skb); + nexthdr = nh->nexthdr; + payload_ofs = (u8 *)(nh + 1) - skb->data; + + ipv6_addr_copy(&key->ipv6_src, &nh->saddr); + ipv6_addr_copy(&key->ipv6_dst, &nh->daddr); + key->nw_tos = ipv6_get_dsfield(nh) & ~INET_ECN_MASK; + key->nw_proto = NEXTHDR_NONE; + + payload_ofs = ipv6_skip_exthdr(skb, payload_ofs, &nexthdr); + if (unlikely(payload_ofs < 0)) + return -EINVAL; + + nh_len = payload_ofs - nh_ofs; + + /* Pull enough header bytes to account for the IP header plus the + * longest transport header that we parse, currently 20 bytes for TCP. + * To dig deeper than the transport header, transport parsers may need + * to pull more header bytes. + */ + if (unlikely(!pskb_may_pull(skb, min(nh_ofs + nh_len + 20, skb->len)))) + return -ENOMEM; + + skb_set_transport_header(skb, nh_ofs + nh_len); + key->nw_proto = nexthdr; + return nh_len; +} + +static bool icmp6hdr_ok(struct sk_buff *skb) +{ + return skb->len >= skb_transport_offset(skb) + sizeof(struct icmp6hdr); +} + #define TCP_FLAGS_OFFSET 13 #define TCP_FLAG_MASK 0x3f @@ -104,24 +166,23 @@ void flow_used(struct sw_flow *flow, struct sk_buff *skb) spin_unlock_bh(&flow->lock); } -struct sw_flow_actions *flow_actions_alloc(u32 actions_len) +struct sw_flow_actions *flow_actions_alloc(const struct nlattr *actions) { + int actions_len = nla_len(actions); struct sw_flow_actions *sfa; - if (actions_len % NLA_ALIGNTO) - return ERR_PTR(-EINVAL); - /* At least DP_MAX_PORTS actions are required to be able to flood a * packet to every port. Factor of 2 allows for setting VLAN tags, * etc. */ if (actions_len > 2 * DP_MAX_PORTS * nla_total_size(4)) return ERR_PTR(-EINVAL); - sfa = kmalloc(sizeof *sfa + actions_len, GFP_KERNEL); + sfa = kmalloc(sizeof(*sfa) + actions_len, GFP_KERNEL); if (!sfa) return ERR_PTR(-ENOMEM); sfa->actions_len = actions_len; + memcpy(sfa->actions, nla_data(actions), actions_len); return sfa; } @@ -175,7 +236,7 @@ void flow_put(struct sw_flow *flow) return; if (atomic_dec_and_test(&flow->refcnt)) { - kfree(flow->sf_acts); + kfree((struct sf_flow_acts __force *)flow->sf_acts); kmem_cache_free(flow_cache, flow); } } @@ -195,7 +256,7 @@ void flow_deferred_free_acts(struct sw_flow_actions *sf_acts) call_rcu(&sf_acts->rcu, rcu_free_acts_callback); } -static void parse_vlan(struct sk_buff *skb, struct odp_flow_key *key) +static void parse_vlan(struct sk_buff *skb, struct sw_flow_key *key) { struct qtag_prefix { __be16 eth_type; /* ETH_P_8021Q */ @@ -207,7 +268,7 @@ static void parse_vlan(struct sk_buff *skb, struct odp_flow_key *key) return; qp = (struct qtag_prefix *) skb->data; - key->dl_tci = qp->tci | htons(ODP_TCI_PRESENT); + key->dl_tci = qp->tci | htons(VLAN_TAG_PRESENT); __skb_pull(skb, sizeof(struct qtag_prefix)); } @@ -218,7 +279,7 @@ static __be16 parse_ethertype(struct sk_buff *skb) u8 ssap; /* Always 0xAA */ u8 ctrl; u8 oui[3]; - u16 ethertype; + __be16 ethertype; }; struct llc_snap_hdr *llc; __be16 proto; @@ -226,22 +287,93 @@ static __be16 parse_ethertype(struct sk_buff *skb) proto = *(__be16 *) skb->data; __skb_pull(skb, sizeof(__be16)); - if (ntohs(proto) >= ODP_DL_TYPE_ETH2_CUTOFF) + if (ntohs(proto) >= 1536) return proto; if (unlikely(skb->len < sizeof(struct llc_snap_hdr))) - return htons(ODP_DL_TYPE_NOT_ETH_TYPE); + return htons(ETH_P_802_2); llc = (struct llc_snap_hdr *) skb->data; if (llc->dsap != LLC_SAP_SNAP || llc->ssap != LLC_SAP_SNAP || (llc->oui[0] | llc->oui[1] | llc->oui[2]) != 0) - return htons(ODP_DL_TYPE_NOT_ETH_TYPE); + return htons(ETH_P_802_2); __skb_pull(skb, sizeof(struct llc_snap_hdr)); return llc->ethertype; } +static int parse_icmpv6(struct sk_buff *skb, struct sw_flow_key *key, + int nh_len) +{ + struct icmp6hdr *icmp = icmp6_hdr(skb); + + /* 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. + */ + key->tp_src = htons(icmp->icmp6_type); + key->tp_dst = htons(icmp->icmp6_code); + + if (icmp->icmp6_code == 0 && + (icmp->icmp6_type == NDISC_NEIGHBOUR_SOLICITATION || + icmp->icmp6_type == NDISC_NEIGHBOUR_ADVERTISEMENT)) { + int icmp_len = skb->len - skb_transport_offset(skb); + struct nd_msg *nd; + int offset; + + /* In order to process neighbor discovery options, we need the + * entire packet. + */ + if (unlikely(icmp_len < sizeof(*nd))) + return 0; + if (unlikely(skb_linearize(skb))) + return -ENOMEM; + + nd = (struct nd_msg *)skb_transport_header(skb); + ipv6_addr_copy(&key->nd_target, &nd->target); + + icmp_len -= sizeof(*nd); + offset = 0; + while (icmp_len >= 8) { + struct nd_opt_hdr *nd_opt = (struct nd_opt_hdr *)(nd->opt + offset); + int opt_len = nd_opt->nd_opt_len * 8; + + if (unlikely(!opt_len || opt_len > icmp_len)) + goto invalid; + + /* Store the link layer address if the appropriate + * option is provided. It is considered an error if + * the same link layer option is specified twice. + */ + if (nd_opt->nd_opt_type == ND_OPT_SOURCE_LL_ADDR + && opt_len == 8) { + if (unlikely(!is_zero_ether_addr(key->arp_sha))) + goto invalid; + memcpy(key->arp_sha, + &nd->opt[offset+sizeof(*nd_opt)], ETH_ALEN); + } else if (nd_opt->nd_opt_type == ND_OPT_TARGET_LL_ADDR + && opt_len == 8) { + if (unlikely(!is_zero_ether_addr(key->arp_tha))) + goto invalid; + memcpy(key->arp_tha, + &nd->opt[offset+sizeof(*nd_opt)], ETH_ALEN); + } + + icmp_len -= opt_len; + offset += opt_len; + } + } + + return 0; + +invalid: + memset(&key->nd_target, 0, sizeof(key->nd_target)); + memset(key->arp_sha, 0, sizeof(key->arp_sha)); + memset(key->arp_tha, 0, sizeof(key->arp_tha)); + + return 0; +} + /** * flow_extract - extracts a flow key from an Ethernet frame. * @skb: sk_buff that contains the frame, with skb->data pointing to the @@ -262,24 +394,25 @@ static __be16 parse_ethertype(struct sk_buff *skb) * - skb->network_header: just past the Ethernet header, or just past the * VLAN header, to the first byte of the Ethernet payload. * - * - skb->transport_header: If key->dl_type is ETH_P_IP on output, then just - * past the IPv4 header, if one is present and of a correct length, - * otherwise the same as skb->network_header. For other key->dl_type - * values it is left untouched. + * - skb->transport_header: If key->dl_type is ETH_P_IP or ETH_P_IPV6 + * on output, then just past the IP header, if one is present and + * of a correct length, otherwise the same as skb->network_header. + * For other key->dl_type values it is left untouched. */ -int flow_extract(struct sk_buff *skb, u16 in_port, struct odp_flow_key *key, +int flow_extract(struct sk_buff *skb, u16 in_port, struct sw_flow_key *key, bool *is_frag) { struct ethhdr *eth; - memset(key, 0, sizeof *key); + memset(key, 0, sizeof(*key)); key->tun_id = OVS_CB(skb)->tun_id; key->in_port = in_port; *is_frag = false; /* * We would really like to pull as many bytes as we could possibly - * want to parse into the linear data area. Currently that is: + * want to parse into the linear data area. Currently, for IPv4, + * that is: * * 14 Ethernet header * 4 VLAN header @@ -306,8 +439,12 @@ int flow_extract(struct sk_buff *skb, u16 in_port, struct odp_flow_key *key, /* dl_type, dl_vlan, dl_vlan_pcp. */ __skb_pull(skb, 2 * ETH_ALEN); - if (eth->h_proto == htons(ETH_P_8021Q)) + + if (vlan_tx_tag_present(skb)) + key->dl_tci = htons(vlan_get_tci(skb)); + else if (eth->h_proto == htons(ETH_P_8021Q)) parse_vlan(skb, key); + key->dl_type = parse_ethertype(skb); skb_reset_network_header(skb); __skb_push(skb, skb->data - (unsigned char *)eth); @@ -327,13 +464,14 @@ int flow_extract(struct sk_buff *skb, u16 in_port, struct odp_flow_key *key, } nh = ip_hdr(skb); - key->nw_src = nh->saddr; - key->nw_dst = nh->daddr; + key->ipv4_src = nh->saddr; + key->ipv4_dst = nh->daddr; key->nw_tos = nh->tos & ~INET_ECN_MASK; key->nw_proto = nh->protocol; /* Transport layer. */ - if (!(nh->frag_off & htons(IP_MF | IP_OFFSET))) { + if (!(nh->frag_off & htons(IP_MF | IP_OFFSET)) && + !(skb_shinfo(skb)->gso_type & SKB_GSO_UDP)) { if (key->nw_proto == IPPROTO_TCP) { if (tcphdr_ok(skb)) { struct tcphdr *tcp = tcp_hdr(skb); @@ -375,25 +513,434 @@ int flow_extract(struct sk_buff *skb, u16 in_port, struct odp_flow_key *key, if (key->nw_proto == ARPOP_REQUEST || key->nw_proto == ARPOP_REPLY) { - memcpy(&key->nw_src, arp->ar_sip, sizeof(key->nw_src)); - memcpy(&key->nw_dst, arp->ar_tip, sizeof(key->nw_dst)); + memcpy(&key->ipv4_src, arp->ar_sip, sizeof(key->ipv4_src)); + memcpy(&key->ipv4_dst, arp->ar_tip, sizeof(key->ipv4_dst)); + memcpy(key->arp_sha, arp->ar_sha, ETH_ALEN); + memcpy(key->arp_tha, arp->ar_tha, ETH_ALEN); + } + } + } else if (key->dl_type == htons(ETH_P_IPV6)) { + int nh_len; /* IPv6 Header + Extensions */ + + nh_len = parse_ipv6hdr(skb, key); + if (unlikely(nh_len < 0)) { + if (nh_len == -EINVAL) { + skb->transport_header = skb->network_header; + return 0; + } + return nh_len; + } + + /* Transport layer. */ + if (key->nw_proto == NEXTHDR_TCP) { + if (tcphdr_ok(skb)) { + struct tcphdr *tcp = tcp_hdr(skb); + key->tp_src = tcp->source; + key->tp_dst = tcp->dest; + } + } else if (key->nw_proto == NEXTHDR_UDP) { + if (udphdr_ok(skb)) { + struct udphdr *udp = udp_hdr(skb); + key->tp_src = udp->source; + key->tp_dst = udp->dest; + } + } else if (key->nw_proto == NEXTHDR_ICMP) { + if (icmp6hdr_ok(skb)) { + int error = parse_icmpv6(skb, key, nh_len); + if (error < 0) + return error; } } } return 0; } -u32 flow_hash(const struct odp_flow_key *key) +u32 flow_hash(const struct sw_flow_key *key) { - return jhash2((u32*)key, sizeof *key / sizeof(u32), hash_seed); + return jhash2((u32*)key, sizeof(*key) / sizeof(u32), hash_seed); } int flow_cmp(const struct tbl_node *node, void *key2_) { - const struct odp_flow_key *key1 = &flow_cast(node)->key; - const struct odp_flow_key *key2 = key2_; + const struct sw_flow_key *key1 = &flow_cast(node)->key; + const struct sw_flow_key *key2 = key2_; + + return !memcmp(key1, key2, sizeof(struct sw_flow_key)); +} + +/** + * flow_from_nlattrs - parses Netlink attributes into a flow key. + * @swkey: receives the extracted flow key. + * @key: Netlink attribute holding nested %ODP_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 \ + * [IPv4 [TCP|UDP|ICMP] | IPv6 [TCP|UDP|ICMPv6 [ND]] | ARP]] + */ +int flow_from_nlattrs(struct sw_flow_key *swkey, const struct nlattr *attr) +{ + const struct nlattr *nla; + u16 prev_type; + int rem; + + memset(swkey, 0, sizeof(*swkey)); + swkey->dl_type = htons(ETH_P_802_2); + + prev_type = ODP_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; + + int type = nla_type(nla); + + if (type > ODP_KEY_ATTR_MAX || nla_len(nla) != key_lens[type]) + return -EINVAL; + +#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): + swkey->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): + if (nla_get_u32(nla) >= DP_MAX_PORTS) + return -EINVAL; + swkey->in_port = nla_get_u32(nla); + break; + + case TRANSITION(ODP_KEY_ATTR_IN_PORT, ODP_KEY_ATTR_ETHERNET): + eth_key = nla_data(nla); + memcpy(swkey->dl_src, eth_key->eth_src, ETH_ALEN); + memcpy(swkey->dl_dst, eth_key->eth_dst, ETH_ALEN); + break; + + case TRANSITION(ODP_KEY_ATTR_ETHERNET, ODP_KEY_ATTR_8021Q): + q_key = nla_data(nla); + /* Only standard 0x8100 VLANs currently supported. */ + if (q_key->q_tpid != htons(ETH_P_8021Q)) + return -EINVAL; + if (q_key->q_tci & htons(VLAN_TAG_PRESENT)) + return -EINVAL; + swkey->dl_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): + swkey->dl_type = nla_get_be16(nla); + if (ntohs(swkey->dl_type) < 1536) + return -EINVAL; + break; + + case TRANSITION(ODP_KEY_ATTR_ETHERTYPE, ODP_KEY_ATTR_IPV4): + if (swkey->dl_type != htons(ETH_P_IP)) + return -EINVAL; + ipv4_key = nla_data(nla); + swkey->ipv4_src = ipv4_key->ipv4_src; + swkey->ipv4_dst = ipv4_key->ipv4_dst; + swkey->nw_proto = ipv4_key->ipv4_proto; + swkey->nw_tos = ipv4_key->ipv4_tos; + if (swkey->nw_tos & INET_ECN_MASK) + return -EINVAL; + break; + + case TRANSITION(ODP_KEY_ATTR_ETHERTYPE, ODP_KEY_ATTR_IPV6): + if (swkey->dl_type != htons(ETH_P_IPV6)) + return -EINVAL; + ipv6_key = nla_data(nla); + memcpy(&swkey->ipv6_src, ipv6_key->ipv6_src, + sizeof(swkey->ipv6_src)); + memcpy(&swkey->ipv6_dst, ipv6_key->ipv6_dst, + sizeof(swkey->ipv6_dst)); + swkey->nw_proto = ipv6_key->ipv6_proto; + swkey->nw_tos = ipv6_key->ipv6_tos; + if (swkey->nw_tos & INET_ECN_MASK) + return -EINVAL; + break; + + case TRANSITION(ODP_KEY_ATTR_IPV4, ODP_KEY_ATTR_TCP): + case TRANSITION(ODP_KEY_ATTR_IPV6, ODP_KEY_ATTR_TCP): + if (swkey->nw_proto != IPPROTO_TCP) + return -EINVAL; + tcp_key = nla_data(nla); + swkey->tp_src = tcp_key->tcp_src; + swkey->tp_dst = tcp_key->tcp_dst; + break; + + case TRANSITION(ODP_KEY_ATTR_IPV4, ODP_KEY_ATTR_UDP): + case TRANSITION(ODP_KEY_ATTR_IPV6, ODP_KEY_ATTR_UDP): + if (swkey->nw_proto != IPPROTO_UDP) + return -EINVAL; + udp_key = nla_data(nla); + swkey->tp_src = udp_key->udp_src; + swkey->tp_dst = udp_key->udp_dst; + break; + + case TRANSITION(ODP_KEY_ATTR_IPV4, ODP_KEY_ATTR_ICMP): + if (swkey->nw_proto != IPPROTO_ICMP) + return -EINVAL; + icmp_key = nla_data(nla); + swkey->tp_src = htons(icmp_key->icmp_type); + swkey->tp_dst = htons(icmp_key->icmp_code); + break; + + case TRANSITION(ODP_KEY_ATTR_IPV6, ODP_KEY_ATTR_ICMPV6): + if (swkey->nw_proto != IPPROTO_ICMPV6) + return -EINVAL; + icmpv6_key = nla_data(nla); + swkey->tp_src = htons(icmpv6_key->icmpv6_type); + swkey->tp_dst = htons(icmpv6_key->icmpv6_code); + break; + + case TRANSITION(ODP_KEY_ATTR_ETHERTYPE, ODP_KEY_ATTR_ARP): + if (swkey->dl_type != htons(ETH_P_ARP)) + return -EINVAL; + arp_key = nla_data(nla); + swkey->ipv4_src = arp_key->arp_sip; + swkey->ipv4_dst = arp_key->arp_tip; + if (arp_key->arp_op & htons(0xff00)) + return -EINVAL; + swkey->nw_proto = ntohs(arp_key->arp_op); + memcpy(swkey->arp_sha, arp_key->arp_sha, ETH_ALEN); + memcpy(swkey->arp_tha, arp_key->arp_tha, ETH_ALEN); + break; + + case TRANSITION(ODP_KEY_ATTR_ICMPV6, ODP_KEY_ATTR_ND): + if (swkey->tp_src != htons(NDISC_NEIGHBOUR_SOLICITATION) + && swkey->tp_src != htons(NDISC_NEIGHBOUR_ADVERTISEMENT)) + return -EINVAL; + nd_key = nla_data(nla); + memcpy(&swkey->nd_target, nd_key->nd_target, + sizeof(swkey->nd_target)); + memcpy(swkey->arp_sha, nd_key->nd_sll, ETH_ALEN); + memcpy(swkey->arp_tha, nd_key->nd_tll, ETH_ALEN); + break; + + default: + return -EINVAL; + } + + prev_type = type; + } + if (rem) + return -EINVAL; + + switch (prev_type) { + case ODP_KEY_ATTR_UNSPEC: + return -EINVAL; + + case ODP_KEY_ATTR_TUN_ID: + case ODP_KEY_ATTR_IN_PORT: + return -EINVAL; + + case ODP_KEY_ATTR_ETHERNET: + case ODP_KEY_ATTR_8021Q: + return 0; + + case ODP_KEY_ATTR_ETHERTYPE: + if (swkey->dl_type == htons(ETH_P_IP) || + swkey->dl_type == htons(ETH_P_ARP)) + return -EINVAL; + return 0; + + case ODP_KEY_ATTR_IPV4: + if (swkey->nw_proto == IPPROTO_TCP || + swkey->nw_proto == IPPROTO_UDP || + swkey->nw_proto == IPPROTO_ICMP) + return -EINVAL; + return 0; + + case ODP_KEY_ATTR_IPV6: + if (swkey->nw_proto == IPPROTO_TCP || + swkey->nw_proto == IPPROTO_UDP || + swkey->nw_proto == IPPROTO_ICMPV6) + return -EINVAL; + return 0; + + case ODP_KEY_ATTR_ICMPV6: + if (swkey->tp_src == htons(NDISC_NEIGHBOUR_SOLICITATION) || + swkey->tp_src == htons(NDISC_NEIGHBOUR_ADVERTISEMENT)) + return -EINVAL; + return 0; + + 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: + return 0; + } + + WARN_ON_ONCE(1); + return -EINVAL; +} + +int flow_to_nlattrs(const struct sw_flow_key *swkey, struct sk_buff *skb) +{ + struct odp_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); + + if (swkey->tun_id != cpu_to_be64(0)) + NLA_PUT_BE64(skb, ODP_KEY_ATTR_TUN_ID, swkey->tun_id); + + NLA_PUT_U32(skb, ODP_KEY_ATTR_IN_PORT, swkey->in_port); + + nla = nla_reserve(skb, ODP_KEY_ATTR_ETHERNET, sizeof(*eth_key)); + if (!nla) + goto nla_put_failure; + eth_key = nla_data(nla); + memcpy(eth_key->eth_src, swkey->dl_src, ETH_ALEN); + memcpy(eth_key->eth_dst, swkey->dl_dst, ETH_ALEN); + + if (swkey->dl_tci != htons(0)) { + struct odp_key_8021q q_key; + + q_key.q_tpid = htons(ETH_P_8021Q); + q_key.q_tci = swkey->dl_tci & ~htons(VLAN_TAG_PRESENT); + NLA_PUT(skb, ODP_KEY_ATTR_8021Q, sizeof(q_key), &q_key); + } + + if (swkey->dl_type == htons(ETH_P_802_2)) + return 0; + + NLA_PUT_BE16(skb, ODP_KEY_ATTR_ETHERTYPE, swkey->dl_type); + + if (swkey->dl_type == htons(ETH_P_IP)) { + struct odp_key_ipv4 *ipv4_key; + + nla = nla_reserve(skb, ODP_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)); + ipv4_key->ipv4_src = swkey->ipv4_src; + ipv4_key->ipv4_dst = swkey->ipv4_dst; + ipv4_key->ipv4_proto = swkey->nw_proto; + ipv4_key->ipv4_tos = swkey->nw_tos; + } else if (swkey->dl_type == htons(ETH_P_IPV6)) { + struct odp_key_ipv6 *ipv6_key; + + nla = nla_reserve(skb, ODP_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)); + memcpy(ipv6_key->ipv6_src, &swkey->ipv6_src, + sizeof(ipv6_key->ipv6_src)); + memcpy(ipv6_key->ipv6_dst, &swkey->ipv6_dst, + sizeof(ipv6_key->ipv6_dst)); + ipv6_key->ipv6_proto = swkey->nw_proto; + ipv6_key->ipv6_tos = swkey->nw_tos; + } else if (swkey->dl_type == htons(ETH_P_ARP)) { + struct odp_key_arp *arp_key; + + nla = nla_reserve(skb, ODP_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)); + arp_key->arp_sip = swkey->ipv4_src; + arp_key->arp_tip = swkey->ipv4_dst; + arp_key->arp_op = htons(swkey->nw_proto); + memcpy(arp_key->arp_sha, swkey->arp_sha, ETH_ALEN); + memcpy(arp_key->arp_tha, swkey->arp_tha, ETH_ALEN); + } + + if (swkey->dl_type == htons(ETH_P_IP) || + swkey->dl_type == htons(ETH_P_IPV6)) { + + if (swkey->nw_proto == IPPROTO_TCP) { + struct odp_key_tcp *tcp_key; + + nla = nla_reserve(skb, ODP_KEY_ATTR_TCP, sizeof(*tcp_key)); + if (!nla) + goto nla_put_failure; + tcp_key = nla_data(nla); + tcp_key->tcp_src = swkey->tp_src; + tcp_key->tcp_dst = swkey->tp_dst; + } else if (swkey->nw_proto == IPPROTO_UDP) { + struct odp_key_udp *udp_key; + + nla = nla_reserve(skb, ODP_KEY_ATTR_UDP, sizeof(*udp_key)); + if (!nla) + goto nla_put_failure; + udp_key = nla_data(nla); + udp_key->udp_src = swkey->tp_src; + udp_key->udp_dst = swkey->tp_dst; + } else if (swkey->dl_type == htons(ETH_P_IP) && + swkey->nw_proto == IPPROTO_ICMP) { + struct odp_key_icmp *icmp_key; + + nla = nla_reserve(skb, ODP_KEY_ATTR_ICMP, sizeof(*icmp_key)); + if (!nla) + goto nla_put_failure; + icmp_key = nla_data(nla); + icmp_key->icmp_type = ntohs(swkey->tp_src); + icmp_key->icmp_code = ntohs(swkey->tp_dst); + } else if (swkey->dl_type == htons(ETH_P_IPV6) && + swkey->nw_proto == IPPROTO_ICMPV6) { + struct odp_key_icmpv6 *icmpv6_key; + + nla = nla_reserve(skb, ODP_KEY_ATTR_ICMPV6, + sizeof(*icmpv6_key)); + if (!nla) + goto nla_put_failure; + icmpv6_key = nla_data(nla); + icmpv6_key->icmpv6_type = ntohs(swkey->tp_src); + icmpv6_key->icmpv6_code = ntohs(swkey->tp_dst); + + if (icmpv6_key->icmpv6_type == NDISC_NEIGHBOUR_SOLICITATION || + icmpv6_key->icmpv6_type == NDISC_NEIGHBOUR_ADVERTISEMENT) { + struct odp_key_nd *nd_key; + + nla = nla_reserve(skb, ODP_KEY_ATTR_ND, sizeof(*nd_key)); + if (!nla) + goto nla_put_failure; + nd_key = nla_data(nla); + memcpy(nd_key->nd_target, &swkey->nd_target, + sizeof(nd_key->nd_target)); + memcpy(nd_key->nd_sll, swkey->arp_sha, ETH_ALEN); + memcpy(nd_key->nd_tll, swkey->arp_tha, ETH_ALEN); + } + } + } + + return 0; - return !memcmp(key1, key2, sizeof(struct odp_flow_key)); +nla_put_failure: + return -EMSGSIZE; } /* Initializes the flow module. @@ -405,7 +952,7 @@ int flow_init(void) if (flow_cache == NULL) return -ENOMEM; - get_random_bytes(&hash_seed, sizeof hash_seed); + get_random_bytes(&hash_seed, sizeof(hash_seed)); return 0; }