Expand tunnel IDs from 32 to 64 bits.

[sliver-openvswitch.git] / datapath / tunnel.c

diff --git a/datapath/tunnel.c b/datapath/tunnel.c
index 37ec4d9..3534909 100644 (file)
--- a/datapath/tunnel.c
+++ b/datapath/tunnel.c
@@ -29,6 +29,7 @@
 #include <net/xfrm.h>
 
 #include "actions.h"
+#include "checksum.h"
 #include "datapath.h"
 #include "table.h"
 #include "tunnel.h"
@@ -71,7 +72,7 @@
 static struct tbl *port_table __read_mostly;
 
 static void cache_cleaner(struct work_struct *work);
-DECLARE_DELAYED_WORK(cache_cleaner_wq, cache_cleaner);
+static DECLARE_DELAYED_WORK(cache_cleaner_wq, cache_cleaner);
 
 /*
  * These are just used as an optimization: they don't require any kind of
@@ -165,11 +166,11 @@ static unsigned int *find_port_pool(const struct tnl_mutable_config *mutable)
 }
 
 struct port_lookup_key {
+       const struct tnl_mutable_config *mutable;
+       __be64 key;
        u32 tunnel_type;
        __be32 saddr;
        __be32 daddr;
-       __be32 key;
-       const struct tnl_mutable_config *mutable;
 };
 
 /*
@@ -191,7 +192,8 @@ static int port_cmp(const struct tbl_node *node, void *target)
 
 static u32 port_hash(struct port_lookup_key *k)
 {
-       return jhash_3words(k->key, k->saddr, k->daddr, k->tunnel_type);
+       u32 x = jhash_3words(k->saddr, k->daddr, k->tunnel_type, 0);
+       return jhash_2words(k->key >> 32, k->key, x);
 }
 
 static u32 mutable_hash(const struct tnl_mutable_config *mutable)
@@ -276,12 +278,15 @@ static int move_port(struct vport *vport, struct tnl_mutable_config *new_mutable
 
        err = tbl_insert(port_table, &tnl_vport->tbl_node, hash);
        if (err) {
+               (*find_port_pool(tnl_vport->mutable))--;
                check_table_empty();
                return err;
        }
 
 table_updated:
+       (*find_port_pool(tnl_vport->mutable))--;
        assign_config_rcu(vport, new_mutable);
+       (*find_port_pool(tnl_vport->mutable))++;
 
        return 0;
 }
@@ -301,7 +306,7 @@ static int del_port(struct vport *vport)
        return 0;
 }
 
-struct vport *tnl_find_port(__be32 saddr, __be32 daddr, __be32 key,
+struct vport *tnl_find_port(__be32 saddr, __be32 daddr, __be64 key,
                            int tunnel_type,
                            const struct tnl_mutable_config **mutable)
 {
@@ -364,34 +369,37 @@ found:
 
 static inline void ecn_decapsulate(struct sk_buff *skb)
 {
-       u8 tos = ip_hdr(skb)->tos;
-
-       if (INET_ECN_is_ce(tos)) {
+       /* This is accessing the outer IP header of the tunnel, which we've
+        * already validated to be OK.  skb->data is currently set to the start
+        * of the inner Ethernet header, and we've validated ETH_HLEN.
+        */
+       if (unlikely(INET_ECN_is_ce(ip_hdr(skb)->tos))) {
                __be16 protocol = skb->protocol;
-               unsigned int nw_header = skb_network_offset(skb);
+
+               skb_set_network_header(skb, ETH_HLEN);
 
                if (skb->protocol == htons(ETH_P_8021Q)) {
                        if (unlikely(!pskb_may_pull(skb, VLAN_ETH_HLEN)))
                                return;
 
                        protocol = vlan_eth_hdr(skb)->h_vlan_encapsulated_proto;
-                       nw_header += VLAN_HLEN;
+                       skb_set_network_header(skb, VLAN_ETH_HLEN);
                }
 
                if (protocol == htons(ETH_P_IP)) {
-                       if (unlikely(!pskb_may_pull(skb, nw_header
+                       if (unlikely(!pskb_may_pull(skb, skb_network_offset(skb)
                            + sizeof(struct iphdr))))
                                return;
 
-                       IP_ECN_set_ce((struct iphdr *)(skb->data + nw_header));
+                       IP_ECN_set_ce(ip_hdr(skb));
                }
 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
                else if (protocol == htons(ETH_P_IPV6)) {
-                       if (unlikely(!pskb_may_pull(skb, nw_header
+                       if (unlikely(!pskb_may_pull(skb, skb_network_offset(skb)
                            + sizeof(struct ipv6hdr))))
                                return;
 
-                       IP6_ECN_set_ce((struct ipv6hdr *)(skb->data + nw_header));
+                       IP6_ECN_set_ce(ipv6_hdr(skb));
                }
 #endif
        }
@@ -400,17 +408,25 @@ static inline void ecn_decapsulate(struct sk_buff *skb)
 /* Called with rcu_read_lock. */
 void tnl_rcv(struct vport *vport, struct sk_buff *skb)
 {
-       skb->pkt_type = PACKET_HOST;
-       skb->protocol = eth_type_trans(skb, skb->dev);
+       /* Packets received by this function are in the following state:
+        * - skb->data points to the inner Ethernet header.
+        * - The inner Ethernet header is in the linear data area.
+        * - skb->csum does not include the inner Ethernet header.
+        * - The layer pointers point at the outer headers.
+        */
+
+       struct ethhdr *eh = (struct ethhdr *)skb->data;
+
+       if (likely(ntohs(eh->h_proto) >= 1536))
+               skb->protocol = eh->h_proto;
+       else
+               skb->protocol = htons(ETH_P_802_2);
 
        skb_dst_drop(skb);
        nf_reset(skb);
        secpath_reset(skb);
-       skb_reset_network_header(skb);
 
        ecn_decapsulate(skb);
-
-       skb_push(skb, ETH_HLEN);
        compute_ip_summed(skb, false);
 
        vport_receive(vport, skb);
@@ -583,7 +599,7 @@ static void ipv6_build_icmp(struct sk_buff *skb, struct sk_buff *nskb,
 #endif /* IPv6 */
 
 bool tnl_frag_needed(struct vport *vport, const struct tnl_mutable_config *mutable,
-                    struct sk_buff *skb, unsigned int mtu, __be32 flow_key)
+                    struct sk_buff *skb, unsigned int mtu, __be64 flow_key)
 {
        unsigned int eth_hdr_len = ETH_HLEN;
        unsigned int total_length = 0, header_length = 0, payload_length;
@@ -803,7 +819,7 @@ static void cache_cleaner(struct work_struct *work)
        schedule_cache_cleaner();
 
        rcu_read_lock();
-       tbl_foreach(port_table, cache_cleaner_cb, NULL);
+       tbl_foreach(rcu_dereference(port_table), cache_cleaner_cb, NULL);
        rcu_read_unlock();
 }