datapath: Remove tunnel path MTU discovery support.
[sliver-openvswitch.git] / datapath / tunnel.c
index d39f40b..b6864bf 100644 (file)
@@ -358,343 +358,6 @@ void ovs_tnl_rcv(struct vport *vport, struct sk_buff *skb)
        ovs_vport_receive(vport, skb);
 }
 
-static bool check_ipv4_address(__be32 addr)
-{
-       if (ipv4_is_multicast(addr) || ipv4_is_lbcast(addr)
-           || ipv4_is_loopback(addr) || ipv4_is_zeronet(addr))
-               return false;
-
-       return true;
-}
-
-static bool ipv4_should_icmp(struct sk_buff *skb)
-{
-       struct iphdr *old_iph = ip_hdr(skb);
-
-       /* Don't respond to L2 broadcast. */
-       if (is_multicast_ether_addr(eth_hdr(skb)->h_dest))
-               return false;
-
-       /* Don't respond to L3 broadcast or invalid addresses. */
-       if (!check_ipv4_address(old_iph->daddr) ||
-           !check_ipv4_address(old_iph->saddr))
-               return false;
-
-       /* Only respond to the first fragment. */
-       if (old_iph->frag_off & htons(IP_OFFSET))
-               return false;
-
-       /* Don't respond to ICMP error messages. */
-       if (old_iph->protocol == IPPROTO_ICMP) {
-               u8 icmp_type, *icmp_typep;
-
-               icmp_typep = skb_header_pointer(skb, (u8 *)old_iph +
-                                               (old_iph->ihl << 2) +
-                                               offsetof(struct icmphdr, type) -
-                                               skb->data, sizeof(icmp_type),
-                                               &icmp_type);
-
-               if (!icmp_typep)
-                       return false;
-
-               if (*icmp_typep > NR_ICMP_TYPES
-                       || (*icmp_typep <= ICMP_PARAMETERPROB
-                               && *icmp_typep != ICMP_ECHOREPLY
-                               && *icmp_typep != ICMP_ECHO))
-                       return false;
-       }
-
-       return true;
-}
-
-static void ipv4_build_icmp(struct sk_buff *skb, struct sk_buff *nskb,
-                           unsigned int mtu, unsigned int payload_length)
-{
-       struct iphdr *iph, *old_iph = ip_hdr(skb);
-       struct icmphdr *icmph;
-       u8 *payload;
-
-       iph = (struct iphdr *)skb_put(nskb, sizeof(struct iphdr));
-       icmph = (struct icmphdr *)skb_put(nskb, sizeof(struct icmphdr));
-       payload = skb_put(nskb, payload_length);
-
-       /* IP */
-       iph->version            =       4;
-       iph->ihl                =       sizeof(struct iphdr) >> 2;
-       iph->tos                =       (old_iph->tos & IPTOS_TOS_MASK) |
-                                       IPTOS_PREC_INTERNETCONTROL;
-       iph->tot_len            =       htons(sizeof(struct iphdr)
-                                             + sizeof(struct icmphdr)
-                                             + payload_length);
-       get_random_bytes(&iph->id, sizeof(iph->id));
-       iph->frag_off           =       0;
-       iph->ttl                =       IPDEFTTL;
-       iph->protocol           =       IPPROTO_ICMP;
-       iph->daddr              =       old_iph->saddr;
-       iph->saddr              =       old_iph->daddr;
-
-       ip_send_check(iph);
-
-       /* ICMP */
-       icmph->type             =       ICMP_DEST_UNREACH;
-       icmph->code             =       ICMP_FRAG_NEEDED;
-       icmph->un.gateway       =       htonl(mtu);
-       icmph->checksum         =       0;
-
-       nskb->csum = csum_partial((u8 *)icmph, sizeof(struct icmphdr), 0);
-       nskb->csum = skb_copy_and_csum_bits(skb, (u8 *)old_iph - skb->data,
-                                           payload, payload_length,
-                                           nskb->csum);
-       icmph->checksum = csum_fold(nskb->csum);
-}
-
-#if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
-static bool ipv6_should_icmp(struct sk_buff *skb)
-{
-       struct ipv6hdr *old_ipv6h = ipv6_hdr(skb);
-       int addr_type;
-       int payload_off = (u8 *)(old_ipv6h + 1) - skb->data;
-       u8 nexthdr = ipv6_hdr(skb)->nexthdr;
-       __be16 frag_off;
-
-       /* Check source address is valid. */
-       addr_type = ipv6_addr_type(&old_ipv6h->saddr);
-       if (addr_type & IPV6_ADDR_MULTICAST || addr_type == IPV6_ADDR_ANY)
-               return false;
-
-       /* Don't reply to unspecified addresses. */
-       if (ipv6_addr_type(&old_ipv6h->daddr) == IPV6_ADDR_ANY)
-               return false;
-
-       /* Don't respond to ICMP error messages. */
-       payload_off = ipv6_skip_exthdr(skb, payload_off, &nexthdr, &frag_off);
-       if (payload_off < 0)
-               return false;
-
-       if (nexthdr == NEXTHDR_ICMP) {
-               u8 icmp_type, *icmp_typep;
-
-               icmp_typep = skb_header_pointer(skb, payload_off +
-                                               offsetof(struct icmp6hdr,
-                                                       icmp6_type),
-                                               sizeof(icmp_type), &icmp_type);
-
-               if (!icmp_typep || !(*icmp_typep & ICMPV6_INFOMSG_MASK))
-                       return false;
-       }
-
-       return true;
-}
-
-static void ipv6_build_icmp(struct sk_buff *skb, struct sk_buff *nskb,
-                           unsigned int mtu, unsigned int payload_length)
-{
-       struct ipv6hdr *ipv6h, *old_ipv6h = ipv6_hdr(skb);
-       struct icmp6hdr *icmp6h;
-       u8 *payload;
-
-       ipv6h = (struct ipv6hdr *)skb_put(nskb, sizeof(struct ipv6hdr));
-       icmp6h = (struct icmp6hdr *)skb_put(nskb, sizeof(struct icmp6hdr));
-       payload = skb_put(nskb, payload_length);
-
-       /* IPv6 */
-       ipv6h->version          =       6;
-       ipv6h->priority         =       0;
-       memset(&ipv6h->flow_lbl, 0, sizeof(ipv6h->flow_lbl));
-       ipv6h->payload_len      =       htons(sizeof(struct icmp6hdr)
-                                             + payload_length);
-       ipv6h->nexthdr          =       NEXTHDR_ICMP;
-       ipv6h->hop_limit        =       IPV6_DEFAULT_HOPLIMIT;
-       ipv6h->daddr            =       old_ipv6h->saddr;
-       ipv6h->saddr            =       old_ipv6h->daddr;
-
-       /* ICMPv6 */
-       icmp6h->icmp6_type      =       ICMPV6_PKT_TOOBIG;
-       icmp6h->icmp6_code      =       0;
-       icmp6h->icmp6_cksum     =       0;
-       icmp6h->icmp6_mtu       =       htonl(mtu);
-
-       nskb->csum = csum_partial((u8 *)icmp6h, sizeof(struct icmp6hdr), 0);
-       nskb->csum = skb_copy_and_csum_bits(skb, (u8 *)old_ipv6h - skb->data,
-                                           payload, payload_length,
-                                           nskb->csum);
-       icmp6h->icmp6_cksum = csum_ipv6_magic(&ipv6h->saddr, &ipv6h->daddr,
-                                               sizeof(struct icmp6hdr)
-                                               + payload_length,
-                                               ipv6h->nexthdr, nskb->csum);
-}
-#endif /* IPv6 */
-
-bool ovs_tnl_frag_needed(struct vport *vport,
-                        const struct tnl_mutable_config *mutable,
-                        struct sk_buff *skb, unsigned int mtu)
-{
-       unsigned int eth_hdr_len = ETH_HLEN;
-       unsigned int total_length = 0, header_length = 0, payload_length;
-       struct ethhdr *eh, *old_eh = eth_hdr(skb);
-       struct sk_buff *nskb;
-
-       /* Sanity check */
-       if (skb->protocol == htons(ETH_P_IP)) {
-               if (mtu < IP_MIN_MTU)
-                       return false;
-
-               if (!ipv4_should_icmp(skb))
-                       return true;
-       }
-#if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
-       else if (skb->protocol == htons(ETH_P_IPV6)) {
-               if (mtu < IPV6_MIN_MTU)
-                       return false;
-
-               /*
-                * In theory we should do PMTUD on IPv6 multicast messages but
-                * we don't have an address to send from so just fragment.
-                */
-               if (ipv6_addr_type(&ipv6_hdr(skb)->daddr) & IPV6_ADDR_MULTICAST)
-                       return false;
-
-               if (!ipv6_should_icmp(skb))
-                       return true;
-       }
-#endif
-       else
-               return false;
-
-       /* Allocate */
-       if (old_eh->h_proto == htons(ETH_P_8021Q))
-               eth_hdr_len = VLAN_ETH_HLEN;
-
-       payload_length = skb->len - eth_hdr_len;
-       if (skb->protocol == htons(ETH_P_IP)) {
-               header_length = sizeof(struct iphdr) + sizeof(struct icmphdr);
-               total_length = min_t(unsigned int, header_length +
-                                                  payload_length, 576);
-       }
-#if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
-       else {
-               header_length = sizeof(struct ipv6hdr) +
-                               sizeof(struct icmp6hdr);
-               total_length = min_t(unsigned int, header_length +
-                                                 payload_length, IPV6_MIN_MTU);
-       }
-#endif
-
-       payload_length = total_length - header_length;
-
-       nskb = dev_alloc_skb(NET_IP_ALIGN + eth_hdr_len + header_length +
-                            payload_length);
-       if (!nskb)
-               return false;
-
-       skb_reserve(nskb, NET_IP_ALIGN);
-
-       /* Ethernet / VLAN */
-       eh = (struct ethhdr *)skb_put(nskb, eth_hdr_len);
-       memcpy(eh->h_dest, old_eh->h_source, ETH_ALEN);
-       memcpy(eh->h_source, mutable->eth_addr, ETH_ALEN);
-       nskb->protocol = eh->h_proto = old_eh->h_proto;
-       if (old_eh->h_proto == htons(ETH_P_8021Q)) {
-               struct vlan_ethhdr *vh = (struct vlan_ethhdr *)eh;
-
-               vh->h_vlan_TCI = vlan_eth_hdr(skb)->h_vlan_TCI;
-               vh->h_vlan_encapsulated_proto = skb->protocol;
-       } else
-               vlan_set_tci(nskb, vlan_get_tci(skb));
-       skb_reset_mac_header(nskb);
-
-       /* Protocol */
-       if (skb->protocol == htons(ETH_P_IP))
-               ipv4_build_icmp(skb, nskb, mtu, payload_length);
-#if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
-       else
-               ipv6_build_icmp(skb, nskb, mtu, payload_length);
-#endif
-
-       if (unlikely(compute_ip_summed(nskb, false))) {
-               kfree_skb(nskb);
-               return false;
-       }
-
-       ovs_vport_receive(vport, nskb);
-
-       return true;
-}
-
-static bool check_mtu(struct sk_buff *skb,
-                     struct vport *vport,
-                     const struct tnl_mutable_config *mutable,
-                     const struct rtable *rt, __be16 *frag_offp,
-                     int tunnel_hlen)
-{
-       bool pmtud;
-       __be16 frag_off;
-       int mtu = 0;
-       unsigned int packet_length = skb->len - ETH_HLEN;
-
-       if (OVS_CB(skb)->tun_key->ipv4_dst) {
-               pmtud = false;
-               frag_off = OVS_CB(skb)->tun_key->tun_flags & OVS_TNL_F_DONT_FRAGMENT ?
-                                 htons(IP_DF) : 0;
-       } else {
-               pmtud = mutable->flags & TNL_F_PMTUD;
-               frag_off = mutable->flags & TNL_F_DF_DEFAULT ? htons(IP_DF) : 0;
-       }
-
-       /* Allow for one level of tagging in the packet length. */
-       if (!vlan_tx_tag_present(skb) &&
-           eth_hdr(skb)->h_proto == htons(ETH_P_8021Q))
-               packet_length -= VLAN_HLEN;
-
-       if (pmtud) {
-               int vlan_header = 0;
-
-               /* The tag needs to go in packet regardless of where it
-                * currently is, so subtract it from the MTU.
-                */
-               if (vlan_tx_tag_present(skb) ||
-                   eth_hdr(skb)->h_proto == htons(ETH_P_8021Q))
-                       vlan_header = VLAN_HLEN;
-
-               mtu = dst_mtu(&rt_dst(rt))
-                       - ETH_HLEN
-                       - tunnel_hlen
-                       - vlan_header;
-       }
-
-       if (skb->protocol == htons(ETH_P_IP)) {
-               struct iphdr *iph = ip_hdr(skb);
-
-               if (pmtud && iph->frag_off & htons(IP_DF)) {
-                       mtu = max(mtu, IP_MIN_MTU);
-
-                       if (packet_length > mtu &&
-                           ovs_tnl_frag_needed(vport, mutable, skb, mtu))
-                               return false;
-               }
-       }
-#if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
-       else if (skb->protocol == htons(ETH_P_IPV6)) {
-               /* IPv6 requires end hosts to do fragmentation
-                * if the packet is above the minimum MTU.
-                */
-               if (packet_length > IPV6_MIN_MTU)
-                       frag_off = htons(IP_DF);
-
-               if (pmtud) {
-                       mtu = max(mtu, IPV6_MIN_MTU);
-
-                       if (packet_length > mtu &&
-                           ovs_tnl_frag_needed(vport, mutable, skb, mtu))
-                               return false;
-               }
-       }
-#endif
-
-       *frag_offp = frag_off;
-       return true;
-}
-
 static struct rtable *find_route(struct net *net,
                __be32 *saddr, __be32 daddr, u8 ipproto,
                u8 tos)
@@ -850,7 +513,7 @@ int ovs_tnl_send(struct vport *vport, struct sk_buff *skb)
        struct ovs_key_ipv4_tunnel tun_key;
        int sent_len = 0;
        int tunnel_hlen;
-       __be16 frag_off = 0;
+       __be16 frag_off;
        __be32 daddr;
        __be32 saddr;
        u8 ttl;
@@ -899,6 +562,8 @@ int ovs_tnl_send(struct vport *vport, struct sk_buff *skb)
                saddr = OVS_CB(skb)->tun_key->ipv4_src;
                tos = OVS_CB(skb)->tun_key->ipv4_tos;
                ttl = OVS_CB(skb)->tun_key->ipv4_ttl;
+                frag_off = OVS_CB(skb)->tun_key->tun_flags &
+                               OVS_TNL_F_DONT_FRAGMENT ?  htons(IP_DF) : 0;
        } else {
                u8 inner_tos;
                daddr = mutable->key.daddr;
@@ -939,6 +604,7 @@ int ovs_tnl_send(struct vport *vport, struct sk_buff *skb)
 #endif
                }
 
+               frag_off = mutable->flags & TNL_F_DF_DEFAULT ? htons(IP_DF) : 0;
        }
 
        /* Route lookup */
@@ -960,12 +626,6 @@ int ovs_tnl_send(struct vport *vport, struct sk_buff *skb)
                goto err_free_rt;
        }
 
-       /* MTU */
-       if (unlikely(!check_mtu(skb, vport, mutable, rt, &frag_off, tunnel_hlen))) {
-               err = VPORT_E_TX_DROPPED;
-               goto err_free_rt;
-       }
-
        /* TTL Fixup. */
        if (!OVS_CB(skb)->tun_key->ipv4_dst) {
                if (!(mutable->flags & TNL_F_TTL_INHERIT)) {