ipv6: Keep track of DST_NOCACHE routes in case of iface down/unregister
authorMartin KaFai Lau <kafai@fb.com>
Sat, 23 May 2015 03:56:04 +0000 (20:56 -0700)
committerDavid S. Miller <davem@davemloft.net>
Mon, 25 May 2015 17:25:34 +0000 (13:25 -0400)
This patch keeps track of the DST_NOCACHE routes in a list and replaces its
dev with loopback during the iface down/unregister event.

Signed-off-by: Martin KaFai Lau <kafai@fb.com>
Cc: Hannes Frederic Sowa <hannes@stressinduktion.org>
Cc: Steffen Klassert <steffen.klassert@secunet.com>
Cc: Julian Anastasov <ja@ssi.bg>
Signed-off-by: David S. Miller <davem@davemloft.net>
include/net/ip6_fib.h
net/ipv6/route.c

index 5556111..cc8f03c 100644 (file)
@@ -120,6 +120,9 @@ struct rt6_info {
        struct rt6key                   rt6i_src;
        struct rt6key                   rt6i_prefsrc;
 
+       struct list_head                rt6i_uncached;
+       struct uncached_list            *rt6i_uncached_list;
+
        struct inet6_dev                *rt6i_idev;
 
        u32                             rt6i_metric;
index 9e4c3c5..94ce1e0 100644 (file)
@@ -105,6 +105,67 @@ static struct rt6_info *rt6_get_route_info(struct net *net,
                                           const struct in6_addr *gwaddr, int ifindex);
 #endif
 
+struct uncached_list {
+       spinlock_t              lock;
+       struct list_head        head;
+};
+
+static DEFINE_PER_CPU_ALIGNED(struct uncached_list, rt6_uncached_list);
+
+static void rt6_uncached_list_add(struct rt6_info *rt)
+{
+       struct uncached_list *ul = raw_cpu_ptr(&rt6_uncached_list);
+
+       rt->dst.flags |= DST_NOCACHE;
+       rt->rt6i_uncached_list = ul;
+
+       spin_lock_bh(&ul->lock);
+       list_add_tail(&rt->rt6i_uncached, &ul->head);
+       spin_unlock_bh(&ul->lock);
+}
+
+static void rt6_uncached_list_del(struct rt6_info *rt)
+{
+       if (!list_empty(&rt->rt6i_uncached)) {
+               struct uncached_list *ul = rt->rt6i_uncached_list;
+
+               spin_lock_bh(&ul->lock);
+               list_del(&rt->rt6i_uncached);
+               spin_unlock_bh(&ul->lock);
+       }
+}
+
+static void rt6_uncached_list_flush_dev(struct net *net, struct net_device *dev)
+{
+       struct net_device *loopback_dev = net->loopback_dev;
+       int cpu;
+
+       for_each_possible_cpu(cpu) {
+               struct uncached_list *ul = per_cpu_ptr(&rt6_uncached_list, cpu);
+               struct rt6_info *rt;
+
+               spin_lock_bh(&ul->lock);
+               list_for_each_entry(rt, &ul->head, rt6i_uncached) {
+                       struct inet6_dev *rt_idev = rt->rt6i_idev;
+                       struct net_device *rt_dev = rt->dst.dev;
+
+                       if (rt_idev && (rt_idev->dev == dev || !dev) &&
+                           rt_idev->dev != loopback_dev) {
+                               rt->rt6i_idev = in6_dev_get(loopback_dev);
+                               in6_dev_put(rt_idev);
+                       }
+
+                       if (rt_dev && (rt_dev == dev || !dev) &&
+                           rt_dev != loopback_dev) {
+                               rt->dst.dev = loopback_dev;
+                               dev_hold(rt->dst.dev);
+                               dev_put(rt_dev);
+                       }
+               }
+               spin_unlock_bh(&ul->lock);
+       }
+}
+
 static u32 *ipv6_cow_metrics(struct dst_entry *dst, unsigned long old)
 {
        struct rt6_info *rt = (struct rt6_info *)dst;
@@ -262,6 +323,7 @@ static inline struct rt6_info *ip6_dst_alloc(struct net *net,
 
                memset(dst + 1, 0, sizeof(*rt) - sizeof(*dst));
                INIT_LIST_HEAD(&rt->rt6i_siblings);
+               INIT_LIST_HEAD(&rt->rt6i_uncached);
        }
        return rt;
 }
@@ -269,11 +331,14 @@ static inline struct rt6_info *ip6_dst_alloc(struct net *net,
 static void ip6_dst_destroy(struct dst_entry *dst)
 {
        struct rt6_info *rt = (struct rt6_info *)dst;
-       struct inet6_dev *idev = rt->rt6i_idev;
        struct dst_entry *from = dst->from;
+       struct inet6_dev *idev;
 
        dst_destroy_metrics_generic(dst);
 
+       rt6_uncached_list_del(rt);
+
+       idev = rt->rt6i_idev;
        if (idev) {
                rt->rt6i_idev = NULL;
                in6_dev_put(idev);
@@ -920,7 +985,7 @@ redo_rt6_select:
                dst_release(&rt->dst);
 
                if (uncached_rt)
-                       uncached_rt->dst.flags |= DST_NOCACHE;
+                       rt6_uncached_list_add(uncached_rt);
                else
                        uncached_rt = net->ipv6.ip6_null_entry;
                dst_hold(&uncached_rt->dst);
@@ -2367,6 +2432,7 @@ void rt6_ifdown(struct net *net, struct net_device *dev)
 
        fib6_clean_all(net, fib6_ifdown, &adn);
        icmp6_clean_all(fib6_ifdown, &adn);
+       rt6_uncached_list_flush_dev(net, dev);
 }
 
 struct rt6_mtu_change_arg {
@@ -3263,6 +3329,7 @@ static struct notifier_block ip6_route_dev_notifier = {
 int __init ip6_route_init(void)
 {
        int ret;
+       int cpu;
 
        ret = -ENOMEM;
        ip6_dst_ops_template.kmem_cachep =
@@ -3322,6 +3389,13 @@ int __init ip6_route_init(void)
        if (ret)
                goto out_register_late_subsys;
 
+       for_each_possible_cpu(cpu) {
+               struct uncached_list *ul = per_cpu_ptr(&rt6_uncached_list, cpu);
+
+               INIT_LIST_HEAD(&ul->head);
+               spin_lock_init(&ul->lock);
+       }
+
 out:
        return ret;