diff options
| author | Eric Dumazet <eric.dumazet@gmail.com> | 2009-12-03 03:49:01 +0300 |
|---|---|---|
| committer | David S. Miller <davem@davemloft.net> | 2009-12-04 03:17:44 +0300 |
| commit | 47e1c323069bcef0acb8a2b48921688573f5ca63 (patch) | |
| tree | bebfe86ae36055adae2100200e87727d21a7b762 /net/ipv4/inet_timewait_sock.c | |
| parent | 13475a30b66cdb9250a34052c19ac98847373030 (diff) | |
| download | linux-47e1c323069bcef0acb8a2b48921688573f5ca63.tar.xz | |
tcp: fix a timewait refcnt race
After TCP RCU conversion, tw->tw_refcnt should not be set to 1 in
inet_twsk_alloc(). It allows a RCU reader to get this timewait socket,
while we not yet stabilized it.
Only choice we have is to set tw_refcnt to 0 in inet_twsk_alloc(),
then atomic_add() it later, once everything is done.
Location of this atomic_add() is tricky, because we dont want another
writer to find this timewait in ehash, while tw_refcnt is still zero !
Thanks to Kapil Dakhane tests and reports.
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Diffstat (limited to 'net/ipv4/inet_timewait_sock.c')
| -rw-r--r-- | net/ipv4/inet_timewait_sock.c | 19 |
1 files changed, 16 insertions, 3 deletions
diff --git a/net/ipv4/inet_timewait_sock.c b/net/ipv4/inet_timewait_sock.c index 11a107a5af4f..0fdf45e4c90c 100644 --- a/net/ipv4/inet_timewait_sock.c +++ b/net/ipv4/inet_timewait_sock.c @@ -109,7 +109,6 @@ void __inet_twsk_hashdance(struct inet_timewait_sock *tw, struct sock *sk, tw->tw_tb = icsk->icsk_bind_hash; WARN_ON(!icsk->icsk_bind_hash); inet_twsk_add_bind_node(tw, &tw->tw_tb->owners); - atomic_inc(&tw->tw_refcnt); spin_unlock(&bhead->lock); spin_lock(lock); @@ -119,13 +118,22 @@ void __inet_twsk_hashdance(struct inet_timewait_sock *tw, struct sock *sk, * Should be done before removing sk from established chain * because readers are lockless and search established first. */ - atomic_inc(&tw->tw_refcnt); inet_twsk_add_node_rcu(tw, &ehead->twchain); /* Step 3: Remove SK from established hash. */ if (__sk_nulls_del_node_init_rcu(sk)) sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1); + /* + * Notes : + * - We initially set tw_refcnt to 0 in inet_twsk_alloc() + * - We add one reference for the bhash link + * - We add one reference for the ehash link + * - We want this refcnt update done before allowing other + * threads to find this tw in ehash chain. + */ + atomic_add(1 + 1 + 1, &tw->tw_refcnt); + spin_unlock(lock); } @@ -157,7 +165,12 @@ struct inet_timewait_sock *inet_twsk_alloc(const struct sock *sk, const int stat tw->tw_transparent = inet->transparent; tw->tw_prot = sk->sk_prot_creator; twsk_net_set(tw, hold_net(sock_net(sk))); - atomic_set(&tw->tw_refcnt, 1); + /* + * Because we use RCU lookups, we should not set tw_refcnt + * to a non null value before everything is setup for this + * timewait socket. + */ + atomic_set(&tw->tw_refcnt, 0); inet_twsk_dead_node_init(tw); __module_get(tw->tw_prot->owner); } |