]> err.no Git - linux-2.6/blobdiff - net/ipv4/ipip.c
Merge branch 'linus' into xen-64bit
[linux-2.6] / net / ipv4 / ipip.c
index da7147090a192103ecb4f88fbf10611e2a7b45b3..4c6d2caf92032c4fef47b2cb042e531eae0aa193 100644 (file)
@@ -1,8 +1,6 @@
 /*
  *     Linux NET3:     IP/IP protocol decoder.
  *
- *     Version: $Id: ipip.c,v 1.50 2001/10/02 02:22:36 davem Exp $
- *
  *     Authors:
  *             Sam Lantinga (slouken@cs.ucdavis.edu)  02/01/95
  *
 
 static int ipip_net_id;
 struct ipip_net {
+       struct ip_tunnel *tunnels_r_l[HASH_SIZE];
+       struct ip_tunnel *tunnels_r[HASH_SIZE];
+       struct ip_tunnel *tunnels_l[HASH_SIZE];
+       struct ip_tunnel *tunnels_wc[1];
+       struct ip_tunnel **tunnels[4];
+
        struct net_device *fb_tunnel_dev;
 };
 
@@ -130,12 +134,6 @@ static int ipip_fb_tunnel_init(struct net_device *dev);
 static int ipip_tunnel_init(struct net_device *dev);
 static void ipip_tunnel_setup(struct net_device *dev);
 
-static struct ip_tunnel *tunnels_r_l[HASH_SIZE];
-static struct ip_tunnel *tunnels_r[HASH_SIZE];
-static struct ip_tunnel *tunnels_l[HASH_SIZE];
-static struct ip_tunnel *tunnels_wc[1];
-static struct ip_tunnel **tunnels[4] = { tunnels_wc, tunnels_l, tunnels_r, tunnels_r_l };
-
 static DEFINE_RWLOCK(ipip_lock);
 
 static struct ip_tunnel * ipip_tunnel_lookup(struct net *net,
@@ -144,21 +142,22 @@ static struct ip_tunnel * ipip_tunnel_lookup(struct net *net,
        unsigned h0 = HASH(remote);
        unsigned h1 = HASH(local);
        struct ip_tunnel *t;
+       struct ipip_net *ipn = net_generic(net, ipip_net_id);
 
-       for (t = tunnels_r_l[h0^h1]; t; t = t->next) {
+       for (t = ipn->tunnels_r_l[h0^h1]; t; t = t->next) {
                if (local == t->parms.iph.saddr &&
                    remote == t->parms.iph.daddr && (t->dev->flags&IFF_UP))
                        return t;
        }
-       for (t = tunnels_r[h0]; t; t = t->next) {
+       for (t = ipn->tunnels_r[h0]; t; t = t->next) {
                if (remote == t->parms.iph.daddr && (t->dev->flags&IFF_UP))
                        return t;
        }
-       for (t = tunnels_l[h1]; t; t = t->next) {
+       for (t = ipn->tunnels_l[h1]; t; t = t->next) {
                if (local == t->parms.iph.saddr && (t->dev->flags&IFF_UP))
                        return t;
        }
-       if ((t = tunnels_wc[0]) != NULL && (t->dev->flags&IFF_UP))
+       if ((t = ipn->tunnels_wc[0]) != NULL && (t->dev->flags&IFF_UP))
                return t;
        return NULL;
 }
@@ -179,7 +178,7 @@ static struct ip_tunnel **__ipip_bucket(struct ipip_net *ipn,
                prio |= 1;
                h ^= HASH(local);
        }
-       return &tunnels[prio][h];
+       return &ipn->tunnels[prio][h];
 }
 
 static inline struct ip_tunnel **ipip_bucket(struct ipip_net *ipn,
@@ -238,6 +237,8 @@ static struct ip_tunnel * ipip_tunnel_locate(struct net *net,
        if (dev == NULL)
                return NULL;
 
+       dev_net_set(dev, net);
+
        if (strchr(name, '%')) {
                if (dev_alloc_name(dev, name) < 0)
                        goto failed_free;
@@ -266,7 +267,7 @@ static void ipip_tunnel_uninit(struct net_device *dev)
 
        if (dev == ipn->fb_tunnel_dev) {
                write_lock_bh(&ipip_lock);
-               tunnels_wc[0] = NULL;
+               ipn->tunnels_wc[0] = NULL;
                write_unlock_bh(&ipip_lock);
        } else
                ipip_tunnel_unlink(ipn, netdev_priv(dev));
@@ -275,9 +276,8 @@ static void ipip_tunnel_uninit(struct net_device *dev)
 
 static int ipip_err(struct sk_buff *skb, u32 info)
 {
-#ifndef I_WISH_WORLD_WERE_PERFECT
 
-/* It is not :-( All the routers (except for Linux) return only
+/* All the routers (except for Linux) return only
    8 bytes of packet payload. It means, that precise relaying of
    ICMP in the real Internet is absolutely infeasible.
  */
@@ -334,133 +334,6 @@ static int ipip_err(struct sk_buff *skb, u32 info)
 out:
        read_unlock(&ipip_lock);
        return err;
-#else
-       struct iphdr *iph = (struct iphdr*)dp;
-       int hlen = iph->ihl<<2;
-       struct iphdr *eiph;
-       const int type = icmp_hdr(skb)->type;
-       const int code = icmp_hdr(skb)->code;
-       int rel_type = 0;
-       int rel_code = 0;
-       __be32 rel_info = 0;
-       __u32 n = 0;
-       struct sk_buff *skb2;
-       struct flowi fl;
-       struct rtable *rt;
-
-       if (len < hlen + sizeof(struct iphdr))
-               return 0;
-       eiph = (struct iphdr*)(dp + hlen);
-
-       switch (type) {
-       default:
-               return 0;
-       case ICMP_PARAMETERPROB:
-               n = ntohl(icmp_hdr(skb)->un.gateway) >> 24;
-               if (n < hlen)
-                       return 0;
-
-               /* So... This guy found something strange INSIDE encapsulated
-                  packet. Well, he is fool, but what can we do ?
-                */
-               rel_type = ICMP_PARAMETERPROB;
-               rel_info = htonl((n - hlen) << 24);
-               break;
-
-       case ICMP_DEST_UNREACH:
-               switch (code) {
-               case ICMP_SR_FAILED:
-               case ICMP_PORT_UNREACH:
-                       /* Impossible event. */
-                       return 0;
-               case ICMP_FRAG_NEEDED:
-                       /* And it is the only really necessary thing :-) */
-                       n = ntohs(icmp_hdr(skb)->un.frag.mtu);
-                       if (n < hlen+68)
-                               return 0;
-                       n -= hlen;
-                       /* BSD 4.2 MORE DOES NOT EXIST IN NATURE. */
-                       if (n > ntohs(eiph->tot_len))
-                               return 0;
-                       rel_info = htonl(n);
-                       break;
-               default:
-                       /* All others are translated to HOST_UNREACH.
-                          rfc2003 contains "deep thoughts" about NET_UNREACH,
-                          I believe, it is just ether pollution. --ANK
-                        */
-                       rel_type = ICMP_DEST_UNREACH;
-                       rel_code = ICMP_HOST_UNREACH;
-                       break;
-               }
-               break;
-       case ICMP_TIME_EXCEEDED:
-               if (code != ICMP_EXC_TTL)
-                       return 0;
-               break;
-       }
-
-       /* Prepare fake skb to feed it to icmp_send */
-       skb2 = skb_clone(skb, GFP_ATOMIC);
-       if (skb2 == NULL)
-               return 0;
-       dst_release(skb2->dst);
-       skb2->dst = NULL;
-       skb_pull(skb2, skb->data - (u8*)eiph);
-       skb_reset_network_header(skb2);
-
-       /* Try to guess incoming interface */
-       memset(&fl, 0, sizeof(fl));
-       fl.fl4_daddr = eiph->saddr;
-       fl.fl4_tos = RT_TOS(eiph->tos);
-       fl.proto = IPPROTO_IPIP;
-       if (ip_route_output_key(&init_net, &rt, &key)) {
-               kfree_skb(skb2);
-               return 0;
-       }
-       skb2->dev = rt->u.dst.dev;
-
-       /* route "incoming" packet */
-       if (rt->rt_flags&RTCF_LOCAL) {
-               ip_rt_put(rt);
-               rt = NULL;
-               fl.fl4_daddr = eiph->daddr;
-               fl.fl4_src = eiph->saddr;
-               fl.fl4_tos = eiph->tos;
-               if (ip_route_output_key(&init_net, &rt, &fl) ||
-                   rt->u.dst.dev->type != ARPHRD_TUNNEL) {
-                       ip_rt_put(rt);
-                       kfree_skb(skb2);
-                       return 0;
-               }
-       } else {
-               ip_rt_put(rt);
-               if (ip_route_input(skb2, eiph->daddr, eiph->saddr, eiph->tos, skb2->dev) ||
-                   skb2->dst->dev->type != ARPHRD_TUNNEL) {
-                       kfree_skb(skb2);
-                       return 0;
-               }
-       }
-
-       /* change mtu on this route */
-       if (type == ICMP_DEST_UNREACH && code == ICMP_FRAG_NEEDED) {
-               if (n > dst_mtu(skb2->dst)) {
-                       kfree_skb(skb2);
-                       return 0;
-               }
-               skb2->dst->ops->update_pmtu(skb2->dst, n);
-       } else if (type == ICMP_TIME_EXCEEDED) {
-               struct ip_tunnel *t = netdev_priv(skb2->dev);
-               if (t->parms.iph.ttl) {
-                       rel_type = ICMP_DEST_UNREACH;
-                       rel_code = ICMP_HOST_UNREACH;
-               }
-       }
-
-       icmp_send(skb2, rel_type, rel_code, rel_info);
-       kfree_skb(skb2);
-       return 0;
-#endif
 }
 
 static inline void ipip_ecn_decapsulate(const struct iphdr *outer_iph,
@@ -493,8 +366,8 @@ static int ipip_rcv(struct sk_buff *skb)
                skb->protocol = htons(ETH_P_IP);
                skb->pkt_type = PACKET_HOST;
 
-               tunnel->stat.rx_packets++;
-               tunnel->stat.rx_bytes += skb->len;
+               tunnel->dev->stats.rx_packets++;
+               tunnel->dev->stats.rx_bytes += skb->len;
                skb->dev = tunnel->dev;
                dst_release(skb->dst);
                skb->dst = NULL;
@@ -517,7 +390,7 @@ static int ipip_rcv(struct sk_buff *skb)
 static int ipip_tunnel_xmit(struct sk_buff *skb, struct net_device *dev)
 {
        struct ip_tunnel *tunnel = netdev_priv(dev);
-       struct net_device_stats *stats = &tunnel->stat;
+       struct net_device_stats *stats = &tunnel->dev->stats;
        struct iphdr  *tiph = &tunnel->parms.iph;
        u8     tos = tunnel->parms.iph.tos;
        __be16 df = tiph->frag_off;
@@ -530,7 +403,7 @@ static int ipip_tunnel_xmit(struct sk_buff *skb, struct net_device *dev)
        int    mtu;
 
        if (tunnel->recursion++) {
-               tunnel->stat.collisions++;
+               stats->collisions++;
                goto tx_error;
        }
 
@@ -543,7 +416,7 @@ static int ipip_tunnel_xmit(struct sk_buff *skb, struct net_device *dev)
        if (!dst) {
                /* NBMA tunnel */
                if ((rt = skb->rtable) == NULL) {
-                       tunnel->stat.tx_fifo_errors++;
+                       stats->tx_fifo_errors++;
                        goto tx_error;
                }
                if ((dst = rt->rt_gateway) == 0)
@@ -557,8 +430,8 @@ static int ipip_tunnel_xmit(struct sk_buff *skb, struct net_device *dev)
                                                .saddr = tiph->saddr,
                                                .tos = RT_TOS(tos) } },
                                    .proto = IPPROTO_IPIP };
-               if (ip_route_output_key(&init_net, &rt, &fl)) {
-                       tunnel->stat.tx_carrier_errors++;
+               if (ip_route_output_key(dev_net(dev), &rt, &fl)) {
+                       stats->tx_carrier_errors++;
                        goto tx_error_icmp;
                }
        }
@@ -566,7 +439,7 @@ static int ipip_tunnel_xmit(struct sk_buff *skb, struct net_device *dev)
 
        if (tdev == dev) {
                ip_rt_put(rt);
-               tunnel->stat.collisions++;
+               stats->collisions++;
                goto tx_error;
        }
 
@@ -576,7 +449,7 @@ static int ipip_tunnel_xmit(struct sk_buff *skb, struct net_device *dev)
                mtu = skb->dst ? dst_mtu(skb->dst) : dev->mtu;
 
        if (mtu < 68) {
-               tunnel->stat.collisions++;
+               stats->collisions++;
                ip_rt_put(rt);
                goto tx_error;
        }
@@ -678,7 +551,7 @@ static void ipip_tunnel_bind_dev(struct net_device *dev)
                                                .tos = RT_TOS(iph->tos) } },
                                    .proto = IPPROTO_IPIP };
                struct rtable *rt;
-               if (!ip_route_output_key(&init_net, &rt, &fl)) {
+               if (!ip_route_output_key(dev_net(dev), &rt, &fl)) {
                        tdev = rt->u.dst.dev;
                        ip_rt_put(rt);
                }
@@ -686,7 +559,7 @@ static void ipip_tunnel_bind_dev(struct net_device *dev)
        }
 
        if (!tdev && tunnel->parms.link)
-               tdev = __dev_get_by_index(&init_net, tunnel->parms.link);
+               tdev = __dev_get_by_index(dev_net(dev), tunnel->parms.link);
 
        if (tdev) {
                dev->hard_header_len = tdev->hard_header_len + sizeof(struct iphdr);
@@ -810,11 +683,6 @@ done:
        return err;
 }
 
-static struct net_device_stats *ipip_tunnel_get_stats(struct net_device *dev)
-{
-       return &(((struct ip_tunnel*)netdev_priv(dev))->stat);
-}
-
 static int ipip_tunnel_change_mtu(struct net_device *dev, int new_mtu)
 {
        if (new_mtu < 68 || new_mtu > 0xFFF8 - sizeof(struct iphdr))
@@ -827,7 +695,6 @@ static void ipip_tunnel_setup(struct net_device *dev)
 {
        dev->uninit             = ipip_tunnel_uninit;
        dev->hard_start_xmit    = ipip_tunnel_xmit;
-       dev->get_stats          = ipip_tunnel_get_stats;
        dev->do_ioctl           = ipip_tunnel_ioctl;
        dev->change_mtu         = ipip_tunnel_change_mtu;
        dev->destructor         = free_netdev;
@@ -838,6 +705,7 @@ static void ipip_tunnel_setup(struct net_device *dev)
        dev->flags              = IFF_NOARP;
        dev->iflink             = 0;
        dev->addr_len           = 4;
+       dev->features           |= NETIF_F_NETNS_LOCAL;
 }
 
 static int ipip_tunnel_init(struct net_device *dev)
@@ -861,6 +729,7 @@ static int ipip_fb_tunnel_init(struct net_device *dev)
 {
        struct ip_tunnel *tunnel = netdev_priv(dev);
        struct iphdr *iph = &tunnel->parms.iph;
+       struct ipip_net *ipn = net_generic(dev_net(dev), ipip_net_id);
 
        tunnel->dev = dev;
        strcpy(tunnel->parms.name, dev->name);
@@ -870,7 +739,7 @@ static int ipip_fb_tunnel_init(struct net_device *dev)
        iph->ihl                = 5;
 
        dev_hold(dev);
-       tunnels_wc[0]           = tunnel;
+       ipn->tunnels_wc[0]      = tunnel;
        return 0;
 }
 
@@ -883,13 +752,27 @@ static struct xfrm_tunnel ipip_handler = {
 static char banner[] __initdata =
        KERN_INFO "IPv4 over IPv4 tunneling driver\n";
 
+static void ipip_destroy_tunnels(struct ipip_net *ipn)
+{
+       int prio;
+
+       for (prio = 1; prio < 4; prio++) {
+               int h;
+               for (h = 0; h < HASH_SIZE; h++) {
+                       struct ip_tunnel *t;
+                       while ((t = ipn->tunnels[prio][h]) != NULL)
+                               unregister_netdevice(t->dev);
+               }
+       }
+}
+
 static int ipip_init_net(struct net *net)
 {
        int err;
        struct ipip_net *ipn;
 
        err = -ENOMEM;
-       ipn = kmalloc(sizeof(struct ipip_net), GFP_KERNEL);
+       ipn = kzalloc(sizeof(struct ipip_net), GFP_KERNEL);
        if (ipn == NULL)
                goto err_alloc;
 
@@ -897,6 +780,11 @@ static int ipip_init_net(struct net *net)
        if (err < 0)
                goto err_assign;
 
+       ipn->tunnels[0] = ipn->tunnels_wc;
+       ipn->tunnels[1] = ipn->tunnels_l;
+       ipn->tunnels[2] = ipn->tunnels_r;
+       ipn->tunnels[3] = ipn->tunnels_r_l;
+
        ipn->fb_tunnel_dev = alloc_netdev(sizeof(struct ip_tunnel),
                                           "tunl0",
                                           ipip_tunnel_setup);
@@ -929,6 +817,7 @@ static void ipip_exit_net(struct net *net)
 
        ipn = net_generic(net, ipip_net_id);
        rtnl_lock();
+       ipip_destroy_tunnels(ipn);
        unregister_netdevice(ipn->fb_tunnel_dev);
        rtnl_unlock();
        kfree(ipn);
@@ -957,29 +846,11 @@ static int __init ipip_init(void)
        return err;
 }
 
-static void __exit ipip_destroy_tunnels(void)
-{
-       int prio;
-
-       for (prio = 1; prio < 4; prio++) {
-               int h;
-               for (h = 0; h < HASH_SIZE; h++) {
-                       struct ip_tunnel *t;
-                       while ((t = tunnels[prio][h]) != NULL)
-                               unregister_netdevice(t->dev);
-               }
-       }
-}
-
 static void __exit ipip_fini(void)
 {
        if (xfrm4_tunnel_deregister(&ipip_handler, AF_INET))
                printk(KERN_INFO "ipip close: can't deregister tunnel\n");
 
-       rtnl_lock();
-       ipip_destroy_tunnels();
-       rtnl_unlock();
-
        unregister_pernet_gen_device(ipip_net_id, &ipip_net_ops);
 }