Merge branch 'tracing-fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel...
[safe/jmp/linux-2.6] / net / ipv4 / ip_gre.c
index 0d5e35b..cb4a0f4 100644 (file)
@@ -126,8 +126,6 @@ static int ipgre_tunnel_bind_dev(struct net_device *dev);
 
 /* Fallback tunnel: no source, no destination, no key, no options */
 
-static int ipgre_fb_tunnel_init(struct net_device *dev);
-
 #define HASH_SIZE  16
 
 static int ipgre_net_id;
@@ -166,67 +164,124 @@ static DEFINE_RWLOCK(ipgre_lock);
 
 /* Given src, dst and key, find appropriate for input tunnel. */
 
-static struct ip_tunnel * ipgre_tunnel_lookup(struct net *net,
+static struct ip_tunnel * ipgre_tunnel_lookup(struct net_device *dev,
                                              __be32 remote, __be32 local,
                                              __be32 key, __be16 gre_proto)
 {
+       struct net *net = dev_net(dev);
+       int link = dev->ifindex;
        unsigned h0 = HASH(remote);
        unsigned h1 = HASH(key);
-       struct ip_tunnel *t;
-       struct ip_tunnel *t2 = NULL;
+       struct ip_tunnel *t, *cand = NULL;
        struct ipgre_net *ign = net_generic(net, ipgre_net_id);
        int dev_type = (gre_proto == htons(ETH_P_TEB)) ?
                       ARPHRD_ETHER : ARPHRD_IPGRE;
+       int score, cand_score = 4;
 
        for (t = ign->tunnels_r_l[h0^h1]; t; t = t->next) {
-               if (local == t->parms.iph.saddr && remote == t->parms.iph.daddr) {
-                       if (t->parms.i_key == key && t->dev->flags & IFF_UP) {
-                               if (t->dev->type == dev_type)
-                                       return t;
-                               if (t->dev->type == ARPHRD_IPGRE && !t2)
-                                       t2 = t;
-                       }
+               if (local != t->parms.iph.saddr ||
+                   remote != t->parms.iph.daddr ||
+                   key != t->parms.i_key ||
+                   !(t->dev->flags & IFF_UP))
+                       continue;
+
+               if (t->dev->type != ARPHRD_IPGRE &&
+                   t->dev->type != dev_type)
+                       continue;
+
+               score = 0;
+               if (t->parms.link != link)
+                       score |= 1;
+               if (t->dev->type != dev_type)
+                       score |= 2;
+               if (score == 0)
+                       return t;
+
+               if (score < cand_score) {
+                       cand = t;
+                       cand_score = score;
                }
        }
 
        for (t = ign->tunnels_r[h0^h1]; t; t = t->next) {
-               if (remote == t->parms.iph.daddr) {
-                       if (t->parms.i_key == key && t->dev->flags & IFF_UP) {
-                               if (t->dev->type == dev_type)
-                                       return t;
-                               if (t->dev->type == ARPHRD_IPGRE && !t2)
-                                       t2 = t;
-                       }
+               if (remote != t->parms.iph.daddr ||
+                   key != t->parms.i_key ||
+                   !(t->dev->flags & IFF_UP))
+                       continue;
+
+               if (t->dev->type != ARPHRD_IPGRE &&
+                   t->dev->type != dev_type)
+                       continue;
+
+               score = 0;
+               if (t->parms.link != link)
+                       score |= 1;
+               if (t->dev->type != dev_type)
+                       score |= 2;
+               if (score == 0)
+                       return t;
+
+               if (score < cand_score) {
+                       cand = t;
+                       cand_score = score;
                }
        }
 
        for (t = ign->tunnels_l[h1]; t; t = t->next) {
-               if (local == t->parms.iph.saddr ||
-                    (local == t->parms.iph.daddr &&
-                     ipv4_is_multicast(local))) {
-                       if (t->parms.i_key == key && t->dev->flags & IFF_UP) {
-                               if (t->dev->type == dev_type)
-                                       return t;
-                               if (t->dev->type == ARPHRD_IPGRE && !t2)
-                                       t2 = t;
-                       }
+               if ((local != t->parms.iph.saddr &&
+                    (local != t->parms.iph.daddr ||
+                     !ipv4_is_multicast(local))) ||
+                   key != t->parms.i_key ||
+                   !(t->dev->flags & IFF_UP))
+                       continue;
+
+               if (t->dev->type != ARPHRD_IPGRE &&
+                   t->dev->type != dev_type)
+                       continue;
+
+               score = 0;
+               if (t->parms.link != link)
+                       score |= 1;
+               if (t->dev->type != dev_type)
+                       score |= 2;
+               if (score == 0)
+                       return t;
+
+               if (score < cand_score) {
+                       cand = t;
+                       cand_score = score;
                }
        }
 
        for (t = ign->tunnels_wc[h1]; t; t = t->next) {
-               if (t->parms.i_key == key && t->dev->flags & IFF_UP) {
-                       if (t->dev->type == dev_type)
-                               return t;
-                       if (t->dev->type == ARPHRD_IPGRE && !t2)
-                               t2 = t;
+               if (t->parms.i_key != key ||
+                   !(t->dev->flags & IFF_UP))
+                       continue;
+
+               if (t->dev->type != ARPHRD_IPGRE &&
+                   t->dev->type != dev_type)
+                       continue;
+
+               score = 0;
+               if (t->parms.link != link)
+                       score |= 1;
+               if (t->dev->type != dev_type)
+                       score |= 2;
+               if (score == 0)
+                       return t;
+
+               if (score < cand_score) {
+                       cand = t;
+                       cand_score = score;
                }
        }
 
-       if (t2)
-               return t2;
+       if (cand != NULL)
+               return cand;
 
-       if (ign->fb_tunnel_dev->flags&IFF_UP)
+       if (ign->fb_tunnel_dev->flags & IFF_UP)
                return netdev_priv(ign->fb_tunnel_dev);
+
        return NULL;
 }
 
@@ -286,6 +341,7 @@ static struct ip_tunnel *ipgre_tunnel_find(struct net *net,
        __be32 remote = parms->iph.daddr;
        __be32 local = parms->iph.saddr;
        __be32 key = parms->i_key;
+       int link = parms->link;
        struct ip_tunnel *t, **tp;
        struct ipgre_net *ign = net_generic(net, ipgre_net_id);
 
@@ -293,6 +349,7 @@ static struct ip_tunnel *ipgre_tunnel_find(struct net *net,
                if (local == t->parms.iph.saddr &&
                    remote == t->parms.iph.daddr &&
                    key == t->parms.i_key &&
+                   link == t->parms.link &&
                    type == t->dev->type)
                        break;
 
@@ -371,7 +428,7 @@ static void ipgre_err(struct sk_buff *skb, u32 info)
    by themself???
  */
 
-       struct iphdr *iph = (struct iphdr*)skb->data;
+       struct iphdr *iph = (struct iphdr *)skb->data;
        __be16       *p = (__be16*)(skb->data+(iph->ihl<<2));
        int grehlen = (iph->ihl<<2) + 4;
        const int type = icmp_hdr(skb)->type;
@@ -423,7 +480,7 @@ static void ipgre_err(struct sk_buff *skb, u32 info)
        }
 
        read_lock(&ipgre_lock);
-       t = ipgre_tunnel_lookup(dev_net(skb->dev), iph->daddr, iph->saddr,
+       t = ipgre_tunnel_lookup(skb->dev, iph->daddr, iph->saddr,
                                flags & GRE_KEY ?
                                *(((__be32 *)p) + (grehlen / 4) - 1) : 0,
                                p[1]);
@@ -434,7 +491,7 @@ static void ipgre_err(struct sk_buff *skb, u32 info)
        if (t->parms.iph.ttl == 0 && type == ICMP_TIME_EXCEEDED)
                goto out;
 
-       if (jiffies - t->err_time < IPTUNNEL_ERR_TIMEO)
+       if (time_before(jiffies, t->err_time + IPTUNNEL_ERR_TIMEO))
                t->err_count++;
        else
                t->err_count = 1;
@@ -520,7 +577,7 @@ static int ipgre_rcv(struct sk_buff *skb)
        gre_proto = *(__be16 *)(h + 2);
 
        read_lock(&ipgre_lock);
-       if ((tunnel = ipgre_tunnel_lookup(dev_net(skb->dev),
+       if ((tunnel = ipgre_tunnel_lookup(skb->dev,
                                          iph->saddr, iph->daddr, key,
                                          gre_proto))) {
                struct net_device_stats *stats = &tunnel->dev->stats;
@@ -545,7 +602,7 @@ static int ipgre_rcv(struct sk_buff *skb)
 #ifdef CONFIG_NET_IPGRE_BROADCAST
                if (ipv4_is_multicast(iph->daddr)) {
                        /* Looped back packet, drop it! */
-                       if (skb->rtable->fl.iif == 0)
+                       if (skb_rtable(skb)->fl.iif == 0)
                                goto drop;
                        stats->multicast++;
                        skb->pkt_type = PACKET_BROADCAST;
@@ -586,8 +643,7 @@ static int ipgre_rcv(struct sk_buff *skb)
                stats->rx_packets++;
                stats->rx_bytes += len;
                skb->dev = tunnel->dev;
-               dst_release(skb->dst);
-               skb->dst = NULL;
+               skb_dst_drop(skb);
                nf_reset(skb);
 
                skb_reset_network_header(skb);
@@ -632,7 +688,7 @@ static int ipgre_tunnel_xmit(struct sk_buff *skb, struct net_device *dev)
 
        if (dev->header_ops && dev->type == ARPHRD_IPGRE) {
                gre_hlen = 0;
-               tiph = (struct iphdr*)skb->data;
+               tiph = (struct iphdr *)skb->data;
        } else {
                gre_hlen = tunnel->hlen;
                tiph = &tunnel->parms.iph;
@@ -641,13 +697,13 @@ static int ipgre_tunnel_xmit(struct sk_buff *skb, struct net_device *dev)
        if ((dst = tiph->daddr) == 0) {
                /* NBMA tunnel */
 
-               if (skb->dst == NULL) {
+               if (skb_dst(skb) == NULL) {
                        stats->tx_fifo_errors++;
                        goto tx_error;
                }
 
                if (skb->protocol == htons(ETH_P_IP)) {
-                       rt = skb->rtable;
+                       rt = skb_rtable(skb);
                        if ((dst = rt->rt_gateway) == 0)
                                goto tx_error_icmp;
                }
@@ -655,12 +711,12 @@ static int ipgre_tunnel_xmit(struct sk_buff *skb, struct net_device *dev)
                else if (skb->protocol == htons(ETH_P_IPV6)) {
                        struct in6_addr *addr6;
                        int addr_type;
-                       struct neighbour *neigh = skb->dst->neighbour;
+                       struct neighbour *neigh = skb_dst(skb)->neighbour;
 
                        if (neigh == NULL)
                                goto tx_error;
 
-                       addr6 = (struct in6_addr*)&neigh->primary_key;
+                       addr6 = (struct in6_addr *)&neigh->primary_key;
                        addr_type = ipv6_addr_type(addr6);
 
                        if (addr_type == IPV6_ADDR_ANY) {
@@ -679,10 +735,10 @@ static int ipgre_tunnel_xmit(struct sk_buff *skb, struct net_device *dev)
        }
 
        tos = tiph->tos;
-       if (tos&1) {
+       if (tos == 1) {
+               tos = 0;
                if (skb->protocol == htons(ETH_P_IP))
                        tos = old_iph->tos;
-               tos &= ~1;
        }
 
        {
@@ -709,10 +765,10 @@ static int ipgre_tunnel_xmit(struct sk_buff *skb, struct net_device *dev)
        if (df)
                mtu = dst_mtu(&rt->u.dst) - dev->hard_header_len - tunnel->hlen;
        else
-               mtu = skb->dst ? dst_mtu(skb->dst) : dev->mtu;
+               mtu = skb_dst(skb) ? dst_mtu(skb_dst(skb)) : dev->mtu;
 
-       if (skb->dst)
-               skb->dst->ops->update_pmtu(skb->dst, mtu);
+       if (skb_dst(skb))
+               skb_dst(skb)->ops->update_pmtu(skb_dst(skb), mtu);
 
        if (skb->protocol == htons(ETH_P_IP)) {
                df |= (old_iph->frag_off&htons(IP_DF));
@@ -726,14 +782,14 @@ static int ipgre_tunnel_xmit(struct sk_buff *skb, struct net_device *dev)
        }
 #ifdef CONFIG_IPV6
        else if (skb->protocol == htons(ETH_P_IPV6)) {
-               struct rt6_info *rt6 = (struct rt6_info*)skb->dst;
+               struct rt6_info *rt6 = (struct rt6_info *)skb_dst(skb);
 
-               if (rt6 && mtu < dst_mtu(skb->dst) && mtu >= IPV6_MIN_MTU) {
+               if (rt6 && mtu < dst_mtu(skb_dst(skb)) && mtu >= IPV6_MIN_MTU) {
                        if ((tunnel->parms.iph.daddr &&
                             !ipv4_is_multicast(tunnel->parms.iph.daddr)) ||
                            rt6->rt6i_dst.plen == 128) {
                                rt6->rt6i_flags |= RTF_MODIFIED;
-                               skb->dst->metrics[RTAX_MTU-1] = mtu;
+                               skb_dst(skb)->metrics[RTAX_MTU-1] = mtu;
                        }
                }
 
@@ -746,7 +802,8 @@ static int ipgre_tunnel_xmit(struct sk_buff *skb, struct net_device *dev)
 #endif
 
        if (tunnel->err_count > 0) {
-               if (jiffies - tunnel->err_time < IPTUNNEL_ERR_TIMEO) {
+               if (time_before(jiffies,
+                               tunnel->err_time + IPTUNNEL_ERR_TIMEO)) {
                        tunnel->err_count--;
 
                        dst_link_failure(skb);
@@ -779,8 +836,8 @@ static int ipgre_tunnel_xmit(struct sk_buff *skb, struct net_device *dev)
        memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
        IPCB(skb)->flags &= ~(IPSKB_XFRM_TUNNEL_SIZE | IPSKB_XFRM_TRANSFORMED |
                              IPSKB_REROUTED);
-       dst_release(skb->dst);
-       skb->dst = &rt->u.dst;
+       skb_dst_drop(skb);
+       skb_dst_set(skb, &rt->u.dst);
 
        /*
         *      Push down and install the IPIP header.
@@ -800,7 +857,7 @@ static int ipgre_tunnel_xmit(struct sk_buff *skb, struct net_device *dev)
                        iph->ttl = old_iph->ttl;
 #ifdef CONFIG_IPV6
                else if (skb->protocol == htons(ETH_P_IPV6))
-                       iph->ttl = ((struct ipv6hdr*)old_iph)->hop_limit;
+                       iph->ttl = ((struct ipv6hdr *)old_iph)->hop_limit;
 #endif
                else
                        iph->ttl = dst_metric(&rt->u.dst, RTAX_HOPLIMIT);
@@ -962,7 +1019,7 @@ ipgre_tunnel_ioctl (struct net_device *dev, struct ifreq *ifr, int cmd)
                                        break;
                                }
                        } else {
-                               unsigned nflags=0;
+                               unsigned nflags = 0;
 
                                t = netdev_priv(dev);
 
@@ -1104,7 +1161,7 @@ static int ipgre_header(struct sk_buff *skb, struct net_device *dev,
 
 static int ipgre_header_parse(const struct sk_buff *skb, unsigned char *haddr)
 {
-       struct iphdr *iph = (struct iphdr*) skb_mac_header(skb);
+       struct iphdr *iph = (struct iphdr *) skb_mac_header(skb);
        memcpy(haddr, &iph->saddr, 4);
        return 4;
 }
@@ -1142,6 +1199,7 @@ static int ipgre_open(struct net_device *dev)
 static int ipgre_close(struct net_device *dev)
 {
        struct ip_tunnel *t = netdev_priv(dev);
+
        if (ipv4_is_multicast(t->parms.iph.daddr) && t->mlink) {
                struct in_device *in_dev;
                in_dev = inetdev_by_index(dev_net(dev), t->mlink);
@@ -1155,14 +1213,22 @@ static int ipgre_close(struct net_device *dev)
 
 #endif
 
+static const struct net_device_ops ipgre_netdev_ops = {
+       .ndo_init               = ipgre_tunnel_init,
+       .ndo_uninit             = ipgre_tunnel_uninit,
+#ifdef CONFIG_NET_IPGRE_BROADCAST
+       .ndo_open               = ipgre_open,
+       .ndo_stop               = ipgre_close,
+#endif
+       .ndo_start_xmit         = ipgre_tunnel_xmit,
+       .ndo_do_ioctl           = ipgre_tunnel_ioctl,
+       .ndo_change_mtu         = ipgre_tunnel_change_mtu,
+};
+
 static void ipgre_tunnel_setup(struct net_device *dev)
 {
-       dev->init               = ipgre_tunnel_init;
-       dev->uninit             = ipgre_tunnel_uninit;
+       dev->netdev_ops         = &ipgre_netdev_ops;
        dev->destructor         = free_netdev;
-       dev->hard_start_xmit    = ipgre_tunnel_xmit;
-       dev->do_ioctl           = ipgre_tunnel_ioctl;
-       dev->change_mtu         = ipgre_tunnel_change_mtu;
 
        dev->type               = ARPHRD_IPGRE;
        dev->needed_headroom    = LL_MAX_HEADER + sizeof(struct iphdr) + 4;
@@ -1171,6 +1237,7 @@ static void ipgre_tunnel_setup(struct net_device *dev)
        dev->iflink             = 0;
        dev->addr_len           = 4;
        dev->features           |= NETIF_F_NETNS_LOCAL;
+       dev->priv_flags         &= ~IFF_XMIT_DST_RELEASE;
 }
 
 static int ipgre_tunnel_init(struct net_device *dev)
@@ -1194,8 +1261,6 @@ static int ipgre_tunnel_init(struct net_device *dev)
                                return -EINVAL;
                        dev->flags = IFF_BROADCAST;
                        dev->header_ops = &ipgre_header_ops;
-                       dev->open = ipgre_open;
-                       dev->stop = ipgre_close;
                }
 #endif
        } else
@@ -1204,7 +1269,7 @@ static int ipgre_tunnel_init(struct net_device *dev)
        return 0;
 }
 
-static int ipgre_fb_tunnel_init(struct net_device *dev)
+static void ipgre_fb_tunnel_init(struct net_device *dev)
 {
        struct ip_tunnel *tunnel = netdev_priv(dev);
        struct iphdr *iph = &tunnel->parms.iph;
@@ -1220,7 +1285,6 @@ static int ipgre_fb_tunnel_init(struct net_device *dev)
 
        dev_hold(dev);
        ign->tunnels_wc[0]      = tunnel;
-       return 0;
 }
 
 
@@ -1264,9 +1328,9 @@ static int ipgre_init_net(struct net *net)
                err = -ENOMEM;
                goto err_alloc_dev;
        }
-
-       ign->fb_tunnel_dev->init = ipgre_fb_tunnel_init;
        dev_net_set(ign->fb_tunnel_dev, net);
+
+       ipgre_fb_tunnel_init(ign->fb_tunnel_dev);
        ign->fb_tunnel_dev->rtnl_link_ops = &ipgre_link_ops;
 
        if ((err = register_netdev(ign->fb_tunnel_dev)))
@@ -1345,7 +1409,7 @@ out:
 static void ipgre_netlink_parms(struct nlattr *data[],
                                struct ip_tunnel_parm *parms)
 {
-       memset(parms, 0, sizeof(parms));
+       memset(parms, 0, sizeof(*parms));
 
        parms->iph.protocol = IPPROTO_GRE;
 
@@ -1368,10 +1432,10 @@ static void ipgre_netlink_parms(struct nlattr *data[],
                parms->o_key = nla_get_be32(data[IFLA_GRE_OKEY]);
 
        if (data[IFLA_GRE_LOCAL])
-               memcpy(&parms->iph.saddr, nla_data(data[IFLA_GRE_LOCAL]), 4);
+               parms->iph.saddr = nla_get_be32(data[IFLA_GRE_LOCAL]);
 
        if (data[IFLA_GRE_REMOTE])
-               memcpy(&parms->iph.daddr, nla_data(data[IFLA_GRE_REMOTE]), 4);
+               parms->iph.daddr = nla_get_be32(data[IFLA_GRE_REMOTE]);
 
        if (data[IFLA_GRE_TTL])
                parms->iph.ttl = nla_get_u8(data[IFLA_GRE_TTL]);
@@ -1397,16 +1461,22 @@ static int ipgre_tap_init(struct net_device *dev)
        return 0;
 }
 
+static const struct net_device_ops ipgre_tap_netdev_ops = {
+       .ndo_init               = ipgre_tap_init,
+       .ndo_uninit             = ipgre_tunnel_uninit,
+       .ndo_start_xmit         = ipgre_tunnel_xmit,
+       .ndo_set_mac_address    = eth_mac_addr,
+       .ndo_validate_addr      = eth_validate_addr,
+       .ndo_change_mtu         = ipgre_tunnel_change_mtu,
+};
+
 static void ipgre_tap_setup(struct net_device *dev)
 {
 
        ether_setup(dev);
 
-       dev->init               = ipgre_tap_init;
-       dev->uninit             = ipgre_tunnel_uninit;
+       dev->netdev_ops         = &ipgre_netdev_ops;
        dev->destructor         = free_netdev;
-       dev->hard_start_xmit    = ipgre_tunnel_xmit;
-       dev->change_mtu         = ipgre_tunnel_change_mtu;
 
        dev->iflink             = 0;
        dev->features           |= NETIF_F_NETNS_LOCAL;
@@ -1539,10 +1609,10 @@ static int ipgre_fill_info(struct sk_buff *skb, const struct net_device *dev)
        NLA_PUT_U32(skb, IFLA_GRE_LINK, p->link);
        NLA_PUT_BE16(skb, IFLA_GRE_IFLAGS, p->i_flags);
        NLA_PUT_BE16(skb, IFLA_GRE_OFLAGS, p->o_flags);
-       NLA_PUT_BE32(skb, IFLA_GRE_IFLAGS, p->i_flags);
-       NLA_PUT_BE32(skb, IFLA_GRE_OFLAGS, p->o_flags);
-       NLA_PUT(skb, IFLA_GRE_LOCAL, 4, &p->iph.saddr);
-       NLA_PUT(skb, IFLA_GRE_REMOTE, 4, &p->iph.daddr);
+       NLA_PUT_BE32(skb, IFLA_GRE_IKEY, p->i_key);
+       NLA_PUT_BE32(skb, IFLA_GRE_OKEY, p->o_key);
+       NLA_PUT_BE32(skb, IFLA_GRE_LOCAL, p->iph.saddr);
+       NLA_PUT_BE32(skb, IFLA_GRE_REMOTE, p->iph.daddr);
        NLA_PUT_U8(skb, IFLA_GRE_TTL, p->iph.ttl);
        NLA_PUT_U8(skb, IFLA_GRE_TOS, p->iph.tos);
        NLA_PUT_U8(skb, IFLA_GRE_PMTUDISC, !!(p->iph.frag_off & htons(IP_DF)));
@@ -1559,8 +1629,8 @@ static const struct nla_policy ipgre_policy[IFLA_GRE_MAX + 1] = {
        [IFLA_GRE_OFLAGS]       = { .type = NLA_U16 },
        [IFLA_GRE_IKEY]         = { .type = NLA_U32 },
        [IFLA_GRE_OKEY]         = { .type = NLA_U32 },
-       [IFLA_GRE_LOCAL]        = { .len = 4 },
-       [IFLA_GRE_REMOTE]       = { .len = 4 },
+       [IFLA_GRE_LOCAL]        = { .len = FIELD_SIZEOF(struct iphdr, saddr) },
+       [IFLA_GRE_REMOTE]       = { .len = FIELD_SIZEOF(struct iphdr, daddr) },
        [IFLA_GRE_TTL]          = { .type = NLA_U8 },
        [IFLA_GRE_TOS]          = { .type = NLA_U8 },
        [IFLA_GRE_PMTUDISC]     = { .type = NLA_U8 },
@@ -1643,5 +1713,5 @@ static void __exit ipgre_fini(void)
 module_init(ipgre_init);
 module_exit(ipgre_fini);
 MODULE_LICENSE("GPL");
-MODULE_ALIAS("rtnl-link-gre");
-MODULE_ALIAS("rtnl-link-gretap");
+MODULE_ALIAS_RTNL_LINK("gre");
+MODULE_ALIAS_RTNL_LINK("gretap");