net: remove INIT_RCU_HEAD() usage
[safe/jmp/linux-2.6] / net / ipv6 / xfrm6_policy.c
index a31dd53..dbdc696 100644 (file)
 #include <net/mip6.h>
 #endif
 
-static struct dst_ops xfrm6_dst_ops;
 static struct xfrm_policy_afinfo xfrm6_policy_afinfo;
 
-static struct dst_entry *xfrm6_dst_lookup(int tos, xfrm_address_t *saddr,
+static struct dst_entry *xfrm6_dst_lookup(struct net *net, int tos,
+                                         xfrm_address_t *saddr,
                                          xfrm_address_t *daddr)
 {
        struct flowi fl = {};
@@ -38,7 +38,7 @@ static struct dst_entry *xfrm6_dst_lookup(int tos, xfrm_address_t *saddr,
        if (saddr)
                memcpy(&fl.fl6_src, saddr, sizeof(fl.fl6_src));
 
-       dst = ip6_route_output(NULL, &fl);
+       dst = ip6_route_output(net, NULL, &fl);
 
        err = dst->error;
        if (dst->error) {
@@ -49,16 +49,20 @@ static struct dst_entry *xfrm6_dst_lookup(int tos, xfrm_address_t *saddr,
        return dst;
 }
 
-static int xfrm6_get_saddr(xfrm_address_t *saddr, xfrm_address_t *daddr)
+static int xfrm6_get_saddr(struct net *net,
+                          xfrm_address_t *saddr, xfrm_address_t *daddr)
 {
        struct dst_entry *dst;
+       struct net_device *dev;
 
-       dst = xfrm6_dst_lookup(0, NULL, daddr);
+       dst = xfrm6_dst_lookup(net, 0, NULL, daddr);
        if (IS_ERR(dst))
                return -EHOSTUNREACH;
 
-       ipv6_get_saddr(dst, (struct in6_addr *)&daddr->a6,
-                      (struct in6_addr *)&saddr->a6);
+       dev = ip6_dst_idev(dst)->dev;
+       ipv6_dev_get_saddr(dev_net(dev), dev,
+                          (struct in6_addr *)&daddr->a6, 0,
+                          (struct in6_addr *)&saddr->a6);
        dst_release(dst);
        return 0;
 }
@@ -98,6 +102,20 @@ static int xfrm6_get_tos(struct flowi *fl)
        return 0;
 }
 
+static int xfrm6_init_path(struct xfrm_dst *path, struct dst_entry *dst,
+                          int nfheader_len)
+{
+       if (dst->ops->family == AF_INET6) {
+               struct rt6_info *rt = (struct rt6_info*)dst;
+               if (rt->rt6i_node)
+                       path->path_cookie = rt->rt6i_node->fn_sernum;
+       }
+
+       path->u.rt6.rt6i_nfheader_len = nfheader_len;
+
+       return 0;
+}
+
 static int xfrm6_fill_dst(struct xfrm_dst *xdst, struct net_device *dev)
 {
        struct rt6_info *rt = (struct rt6_info*)xdst->route;
@@ -115,6 +133,8 @@ static int xfrm6_fill_dst(struct xfrm_dst *xdst, struct net_device *dev)
                                                   RTF_LOCAL);
        xdst->u.rt6.rt6i_metric = rt->rt6i_metric;
        xdst->u.rt6.rt6i_node = rt->rt6i_node;
+       if (rt->rt6i_node)
+               xdst->route_cookie = rt->rt6i_node->fn_sernum;
        xdst->u.rt6.rt6i_gateway = rt->rt6i_gateway;
        xdst->u.rt6.rt6i_dst = rt->rt6i_dst;
        xdst->u.rt6.rt6i_src = rt->rt6i_src;
@@ -123,8 +143,9 @@ static int xfrm6_fill_dst(struct xfrm_dst *xdst, struct net_device *dev)
 }
 
 static inline void
-_decode_session6(struct sk_buff *skb, struct flowi *fl)
+_decode_session6(struct sk_buff *skb, struct flowi *fl, int reverse)
 {
+       int onlyproto = 0;
        u16 offset = skb_network_header_len(skb);
        struct ipv6hdr *hdr = ipv6_hdr(skb);
        struct ipv6_opt_hdr *exthdr;
@@ -132,14 +153,17 @@ _decode_session6(struct sk_buff *skb, struct flowi *fl)
        u8 nexthdr = nh[IP6CB(skb)->nhoff];
 
        memset(fl, 0, sizeof(struct flowi));
-       ipv6_addr_copy(&fl->fl6_dst, &hdr->daddr);
-       ipv6_addr_copy(&fl->fl6_src, &hdr->saddr);
+       ipv6_addr_copy(&fl->fl6_dst, reverse ? &hdr->saddr : &hdr->daddr);
+       ipv6_addr_copy(&fl->fl6_src, reverse ? &hdr->daddr : &hdr->saddr);
 
-       while (pskb_may_pull(skb, nh + offset + 1 - skb->data)) {
+       while (nh + offset + 1 < skb->data ||
+              pskb_may_pull(skb, nh + offset + 1 - skb->data)) {
                nh = skb_network_header(skb);
                exthdr = (struct ipv6_opt_hdr *)(nh + offset);
 
                switch (nexthdr) {
+               case NEXTHDR_FRAGMENT:
+                       onlyproto = 1;
                case NEXTHDR_ROUTING:
                case NEXTHDR_HOP:
                case NEXTHDR_DEST:
@@ -153,17 +177,18 @@ _decode_session6(struct sk_buff *skb, struct flowi *fl)
                case IPPROTO_TCP:
                case IPPROTO_SCTP:
                case IPPROTO_DCCP:
-                       if (pskb_may_pull(skb, nh + offset + 4 - skb->data)) {
+                       if (!onlyproto && (nh + offset + 4 < skb->data ||
+                            pskb_may_pull(skb, nh + offset + 4 - skb->data))) {
                                __be16 *ports = (__be16 *)exthdr;
 
-                               fl->fl_ip_sport = ports[0];
-                               fl->fl_ip_dport = ports[1];
+                               fl->fl_ip_sport = ports[!!reverse];
+                               fl->fl_ip_dport = ports[!reverse];
                        }
                        fl->proto = nexthdr;
                        return;
 
                case IPPROTO_ICMPV6:
-                       if (pskb_may_pull(skb, nh + offset + 2 - skb->data)) {
+                       if (!onlyproto && pskb_may_pull(skb, nh + offset + 2 - skb->data)) {
                                u8 *icmp = (u8 *)exthdr;
 
                                fl->fl_icmp_type = icmp[0];
@@ -174,7 +199,7 @@ _decode_session6(struct sk_buff *skb, struct flowi *fl)
 
 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
                case IPPROTO_MH:
-                       if (pskb_may_pull(skb, nh + offset + 3 - skb->data)) {
+                       if (!onlyproto && pskb_may_pull(skb, nh + offset + 3 - skb->data)) {
                                struct ip6_mh *mh;
                                mh = (struct ip6_mh *)exthdr;
 
@@ -196,10 +221,12 @@ _decode_session6(struct sk_buff *skb, struct flowi *fl)
        }
 }
 
-static inline int xfrm6_garbage_collect(void)
+static inline int xfrm6_garbage_collect(struct dst_ops *ops)
 {
-       xfrm6_policy_afinfo.garbage_collect();
-       return (atomic_read(&xfrm6_dst_ops.entries) > xfrm6_dst_ops.gc_thresh*2);
+       struct net *net = container_of(ops, struct net, xfrm.xfrm6_dst_ops);
+
+       xfrm6_policy_afinfo.garbage_collect(net);
+       return (atomic_read(&ops->entries) > ops->gc_thresh * 2);
 }
 
 static void xfrm6_update_pmtu(struct dst_entry *dst, u32 mtu)
@@ -229,7 +256,8 @@ static void xfrm6_dst_ifdown(struct dst_entry *dst, struct net_device *dev,
 
        xdst = (struct xfrm_dst *)dst;
        if (xdst->u.rt6.rt6i_idev->dev == dev) {
-               struct inet6_dev *loopback_idev = in6_dev_get(init_net.loopback_dev);
+               struct inet6_dev *loopback_idev =
+                       in6_dev_get(dev_net(dev)->loopback_dev);
                BUG_ON(!loopback_idev);
 
                do {
@@ -247,14 +275,14 @@ static void xfrm6_dst_ifdown(struct dst_entry *dst, struct net_device *dev,
 
 static struct dst_ops xfrm6_dst_ops = {
        .family =               AF_INET6,
-       .protocol =             __constant_htons(ETH_P_IPV6),
+       .protocol =             cpu_to_be16(ETH_P_IPV6),
        .gc =                   xfrm6_garbage_collect,
        .update_pmtu =          xfrm6_update_pmtu,
        .destroy =              xfrm6_dst_destroy,
        .ifdown =               xfrm6_dst_ifdown,
        .local_out =            __ip6_local_out,
        .gc_thresh =            1024,
-       .entry_size =           sizeof(struct xfrm_dst),
+       .entries =              ATOMIC_INIT(0),
 };
 
 static struct xfrm_policy_afinfo xfrm6_policy_afinfo = {
@@ -265,12 +293,13 @@ static struct xfrm_policy_afinfo xfrm6_policy_afinfo = {
        .find_bundle =          __xfrm6_find_bundle,
        .decode_session =       _decode_session6,
        .get_tos =              xfrm6_get_tos,
+       .init_path =            xfrm6_init_path,
        .fill_dst =             xfrm6_fill_dst,
 };
 
-static void __init xfrm6_policy_init(void)
+static int __init xfrm6_policy_init(void)
 {
-       xfrm_policy_register_afinfo(&xfrm6_policy_afinfo);
+       return xfrm_policy_register_afinfo(&xfrm6_policy_afinfo);
 }
 
 static void xfrm6_policy_fini(void)
@@ -278,14 +307,65 @@ static void xfrm6_policy_fini(void)
        xfrm_policy_unregister_afinfo(&xfrm6_policy_afinfo);
 }
 
-void __init xfrm6_init(void)
+#ifdef CONFIG_SYSCTL
+static struct ctl_table xfrm6_policy_table[] = {
+       {
+               .procname       = "xfrm6_gc_thresh",
+               .data           = &init_net.xfrm.xfrm6_dst_ops.gc_thresh,
+               .maxlen         = sizeof(int),
+               .mode           = 0644,
+               .proc_handler   = proc_dointvec,
+       },
+       { }
+};
+
+static struct ctl_table_header *sysctl_hdr;
+#endif
+
+int __init xfrm6_init(void)
 {
-       xfrm6_policy_init();
-       xfrm6_state_init();
+       int ret;
+       unsigned int gc_thresh;
+
+       /*
+        * We need a good default value for the xfrm6 gc threshold.
+        * In ipv4 we set it to the route hash table size * 8, which
+        * is half the size of the maximaum route cache for ipv4.  It
+        * would be good to do the same thing for v6, except the table is
+        * constructed differently here.  Here each table for a net namespace
+        * can have FIB_TABLE_HASHSZ entries, so lets go with the same
+        * computation that we used for ipv4 here.  Also, lets keep the initial
+        * gc_thresh to a minimum of 1024, since, the ipv6 route cache defaults
+        * to that as a minimum as well
+        */
+       gc_thresh = FIB6_TABLE_HASHSZ * 8;
+       xfrm6_dst_ops.gc_thresh = (gc_thresh < 1024) ? 1024 : gc_thresh;
+
+       ret = xfrm6_policy_init();
+       if (ret)
+               goto out;
+
+       ret = xfrm6_state_init();
+       if (ret)
+               goto out_policy;
+
+#ifdef CONFIG_SYSCTL
+       sysctl_hdr = register_net_sysctl_table(&init_net, net_ipv6_ctl_path,
+                                               xfrm6_policy_table);
+#endif
+out:
+       return ret;
+out_policy:
+       xfrm6_policy_fini();
+       goto out;
 }
 
 void xfrm6_fini(void)
 {
+#ifdef CONFIG_SYSCTL
+       if (sysctl_hdr)
+               unregister_net_sysctl_table(sysctl_hdr);
+#endif
        //xfrm6_input_fini();
        xfrm6_policy_fini();
        xfrm6_state_fini();