[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, 1 insertions, 11 deletions

diff --git a/include/net/sock.h b/include/net/sock.h
index b9cfe125c9e6..43fc3fa50d62 100644
--- a/include/net/sock.h
+++ b/include/net/sock.h
@@ -905,16 +905,6 @@ static inline int sk_filter(struct sock *sk, struct sk_buff *skb)
 }
 
 /**
- *      sk_filter_rcu_free: Free a socket filter
- *      @rcu: rcu_head that contains the sk_filter to free
- */
-static inline void sk_filter_rcu_free(struct rcu_head *rcu)
-{
-        struct sk_filter *fp = container_of(rcu, struct sk_filter, rcu);
-        kfree(fp);
-}
-
-/**
  *      sk_filter_release: Release a socket filter
  *      @sk: socket
  *      @fp: filter to remove
@@ -925,7 +915,7 @@ static inline void sk_filter_rcu_free(struct rcu_head *rcu)
 static inline void sk_filter_release(struct sk_filter *fp)
 {
         if (atomic_dec_and_test(&fp->refcnt))
-                call_rcu_bh(&fp->rcu, sk_filter_rcu_free);
+                kfree(fp);
 }
 
 static inline void sk_filter_uncharge(struct sock *sk, struct sk_filter *fp)