netns xfrm: deal with dst entries in netns
[safe/jmp/linux-2.6] / net / ipv6 / xfrm6_policy.c
index d92d1fc..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(&init_net, NULL, &fl);
+       dst = ip6_route_output(net, NULL, &fl);
 
        err = dst->error;
        if (dst->error) {
@@ -49,15 +49,18 @@ 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_dev_get_saddr(ip6_dst_idev(dst)->dev,
+       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);
@@ -142,6 +145,7 @@ 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, int reverse)
 {
+       int onlyproto = 0;
        u16 offset = skb_network_header_len(skb);
        struct ipv6hdr *hdr = ipv6_hdr(skb);
        struct ipv6_opt_hdr *exthdr;
@@ -152,11 +156,14 @@ _decode_session6(struct sk_buff *skb, struct flowi *fl, int reverse)
        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:
@@ -170,7 +177,8 @@ _decode_session6(struct sk_buff *skb, struct flowi *fl, int reverse)
                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[!!reverse];
@@ -180,7 +188,7 @@ _decode_session6(struct sk_buff *skb, struct flowi *fl, int reverse)
                        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];
@@ -191,7 +199,7 @@ _decode_session6(struct sk_buff *skb, struct flowi *fl, int reverse)
 
 #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;
 
@@ -215,8 +223,10 @@ _decode_session6(struct sk_buff *skb, struct flowi *fl, int reverse)
 
 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)
@@ -247,7 +257,7 @@ 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(dev->nd_net->loopback_dev);
+                       in6_dev_get(dev_net(dev)->loopback_dev);
                BUG_ON(!loopback_idev);
 
                do {
@@ -265,14 +275,13 @@ 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),
 };
 
@@ -298,9 +307,39 @@ static void xfrm6_policy_fini(void)
        xfrm_policy_unregister_afinfo(&xfrm6_policy_afinfo);
 }
 
+#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)
 {
        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)
@@ -309,6 +348,11 @@ int __init xfrm6_init(void)
        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:
@@ -318,6 +362,10 @@ out_policy:
 
 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();