Commit 9e17f8a4 authored by Eric Dumazet's avatar Eric Dumazet Committed by David S. Miller
Browse files

net: make skb_set_owner_w() more robust 



skb_set_owner_w() is called from various places that assume
skb->sk always point to a full blown socket (as it changes
sk->sk_wmem_alloc)

We'd like to attach skb to request sockets, and in the future
to timewait sockets as well. For these kind of pseudo sockets,
we need to take a traditional refcount and use sock_edemux()
as the destructor.

It is now time to un-inline skb_set_owner_w(), being too big.

Fixes: ca6fb065 ("tcp: attach SYNACK messages to request sockets instead of listener")
Signed-off-by: default avatarEric Dumazet <edumazet@google.com>
Bisected-by: default avatarHaiyang Zhang <haiyangz@microsoft.com>
Signed-off-by: default avatarDavid S. Miller <davem@davemloft.net>
parent eca1e006

include/net/sock.h

+2 −15
Original line number Diff line number Diff line
@@ -1951,6 +1951,8 @@ static inline void skb_set_hash_from_sk(struct sk_buff *skb, struct sock *sk) 
	}

}



void skb_set_owner_w(struct sk_buff *skb, struct sock *sk);



/*

 *	Queue a received datagram if it will fit. Stream and sequenced

 *	protocols can't normally use this as they need to fit buffers in
@@ -1959,21 +1961,6 @@ static inline void skb_set_hash_from_sk(struct sk_buff *skb, struct sock *sk) 
 *	Inlined as it's very short and called for pretty much every

 *	packet ever received.

 */



static inline void skb_set_owner_w(struct sk_buff *skb, struct sock *sk)

{

	skb_orphan(skb);

	skb->sk = sk;

	skb->destructor = sock_wfree;

	skb_set_hash_from_sk(skb, sk);

	/*

	 * We used to take a refcount on sk, but following operation

	 * is enough to guarantee sk_free() wont free this sock until

	 * all in-flight packets are completed

	 */

	atomic_add(skb->truesize, &sk->sk_wmem_alloc);

}



static inline void skb_set_owner_r(struct sk_buff *skb, struct sock *sk)

{

	skb_orphan(skb);

net/core/sock.c

+22 −0
Original line number Diff line number Diff line
@@ -1656,6 +1656,28 @@ void sock_wfree(struct sk_buff *skb) 
}

EXPORT_SYMBOL(sock_wfree);



void skb_set_owner_w(struct sk_buff *skb, struct sock *sk)

{

	skb_orphan(skb);

	skb->sk = sk;

#ifdef CONFIG_INET

	if (unlikely(!sk_fullsock(sk))) {

		skb->destructor = sock_edemux;

		sock_hold(sk);

		return;

	}

#endif

	skb->destructor = sock_wfree;

	skb_set_hash_from_sk(skb, sk);

	/*

	 * We used to take a refcount on sk, but following operation

	 * is enough to guarantee sk_free() wont free this sock until

	 * all in-flight packets are completed

	 */

	atomic_add(skb->truesize, &sk->sk_wmem_alloc);

}

EXPORT_SYMBOL(skb_set_owner_w);



void skb_orphan_partial(struct sk_buff *skb)

{

	/* TCP stack sets skb->ooo_okay based on sk_wmem_alloc,

net/ipv4/tcp_output.c

+1 −3
Original line number Diff line number Diff line
@@ -2963,9 +2963,7 @@ struct sk_buff *tcp_make_synack(const struct sock *sk, struct dst_entry *dst, 
	skb_reserve(skb, MAX_TCP_HEADER);



	if (attach_req) {

		skb->destructor = sock_edemux;

		sock_hold(req_to_sk(req));

		skb->sk = req_to_sk(req);

		skb_set_owner_w(skb, req_to_sk(req));

	} else {

		/* sk is a const pointer, because we want to express multiple

		 * cpu might call us concurrently.

CRA Git | Maintained and supported by SUSTech CRA and CCSE