11:46 p.m.
Hi Paolo -
On Thu, 2019-10-17 at 11:12 +0200, Paolo Abeni wrote:
Hi,
On Wed, 2019-10-16 at 17:02 -0700, Peter Krystad wrote:
> All routines exported already have their IPv4 counterparts exported.
> tcp_request_sock_ipv6_ops will be referenced during TCP subflow
> creation.
>
> Signed-off-by: Peter Krystad <peter.krystad(a)linux.intel.com>
I guess this could be squashed into "tcp: Export TCP functions and ops
struct", right?
Maybe, see my last comment.
[...]
> diff --git a/net/ipv6/tcp_ipv6.c b/net/ipv6/tcp_ipv6.c
> index e3d9f4559c99..1f3a87f4867e 100644
> --- a/net/ipv6/tcp_ipv6.c
> +++ b/net/ipv6/tcp_ipv6.c
> @@ -75,7 +75,7 @@ static void tcp_v6_reqsk_send_ack(const struct sock *sk, struct
sk_buff *skb,
> static int tcp_v6_do_rcv(struct sock *sk, struct sk_buff *skb);
>
> static const struct inet_connection_sock_af_ops ipv6_mapped;
> -static const struct inet_connection_sock_af_ops ipv6_specific;
> +const struct inet_connection_sock_af_ops ipv6_specific;
> #ifdef CONFIG_TCP_MD5SIG
> static const struct tcp_sock_af_ops tcp_sock_ipv6_specific;
> static const struct tcp_sock_af_ops tcp_sock_ipv6_mapped_specific;
> @@ -99,7 +99,7 @@ static struct ipv6_pinfo *tcp_inet6_sk(const struct sock *sk)
> return (struct ipv6_pinfo *)(((u8 *)sk) + offset);
> }
>
> -static void inet6_sk_rx_dst_set(struct sock *sk, const struct sk_buff *skb)
> +void inet6_sk_rx_dst_set(struct sock *sk, const struct sk_buff *skb)
> {
> struct dst_entry *dst = skb_dst(skb);
>
> @@ -111,6 +111,7 @@ static void inet6_sk_rx_dst_set(struct sock *sk, const struct
sk_buff *skb)
> tcp_inet6_sk(sk)->rx_dst_cookie = rt6_get_cookie(rt);
> }
> }
> +EXPORT_SYMBOL(inet6_sk_rx_dst_set);
I think this is not needed, as we don't build mptcp as a module.
Dropping the 'static' scope should suffice. This should apply also to
tcp_request_sock_ipv6_ops, tcp_v6_syn_recv_sock and ipv6_specific
below.
Correct, and I'll do the same for the others.
[...]
> diff --git a/net/mptcp/protocol.h b/net/mptcp/protocol.h
> index 791f2c19cfb8..6f09fdfdd523 100644
> --- a/net/mptcp/protocol.h
> +++ b/net/mptcp/protocol.h
> @@ -268,6 +268,9 @@ int mptcp_subflow_connect(struct sock *sk, struct sockaddr
*local,
> int mptcp_subflow_create_socket(struct sock *sk, struct socket **new_sock);
>
> extern const struct inet_connection_sock_af_ops ipv4_specific;
> +#if IS_ENABLED(CONFIG_IPV6)
> +extern const struct inet_connection_sock_af_ops ipv6_specific;
> +#endif
>
> void mptcp_proto_init(void);
>
uhmm.. _if_ we want to squash these patches, this chunk should likely
go to a different patch than all the above - possibly "mptcp: Handle
MP_CAPABLE options for outgoing connections". Perhaps squashing is not
a good idea in this case.
Squashing depends on if we want the patch sequence to look like we did v6 at
the same time as v4, i.e. as each feature is introduced it has both v4 and v6.
I'm in favor of not squashing this patch.
Peter.
Cheers,
Paolo