[NET]: Fix the race between sk_filter_(de|at)tach and sk_clone()

The proposed fix is to delay the reference counter decrement
until the quiescent state pass. This will give sk_clone() a
chance to get the reference on the cloned filter.

Regular sk_filter_uncharge can happen from the sk_free() only
and there's no need in delaying the put - the socket is dead
anyway and is to be release itself.

Signed-off-by: Pavel Emelyanov <xemul@openvz.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

1 files changed, 21 insertions, 2 deletions

diff --git a/net/core/filter.c b/net/core/filter.c
index 54dddc92452d..1f0068eae501 100644
--- a/net/core/filter.c
+++ b/net/core/filter.c
@@ -387,6 +387,25 @@ int sk_chk_filter(struct sock_filter *filter, int flen)
 }
 
 /**
+ *      sk_filter_rcu_release: Release a socket filter by rcu_head
+ *      @rcu: rcu_head that contains the sk_filter to free
+ */
+static void sk_filter_rcu_release(struct rcu_head *rcu)
+{
+        struct sk_filter *fp = container_of(rcu, struct sk_filter, rcu);
+
+        sk_filter_release(fp);
+}
+
+static void sk_filter_delayed_uncharge(struct sock *sk, struct sk_filter *fp)
+{
+        unsigned int size = sk_filter_len(fp);
+
+        atomic_sub(size, &sk->sk_omem_alloc);
+        call_rcu_bh(&fp->rcu, sk_filter_rcu_release);
+}
+
+/**
  *      sk_attach_filter - attach a socket filter
  *      @fprog: the filter program
  *      @sk: the socket to use
@@ -428,7 +447,7 @@ int sk_attach_filter(struct sock_fprog *fprog, struct sock *sk)
         rcu_assign_pointer(sk->sk_filter, fp);
         rcu_read_unlock_bh();
 
-        sk_filter_uncharge(sk, old_fp);
+        sk_filter_delayed_uncharge(sk, old_fp);
         return 0;
 }
 
@@ -441,7 +460,7 @@ int sk_detach_filter(struct sock *sk)
         filter = rcu_dereference(sk->sk_filter);
         if (filter) {
                 rcu_assign_pointer(sk->sk_filter, NULL);
-                sk_filter_uncharge(sk, filter);
+                sk_filter_delayed_uncharge(sk, filter);
                 ret = 0;
         }
         rcu_read_unlock_bh();