On Tue, Sep 29, 2020 at 09:49:01PM +0800, Xin Long wrote:
...
> +struct sctp_udpencaps {
> + sctp_assoc_t sue_assoc_id;
> + struct sockaddr_storage sue_address;
> + uint16_t sue_port;
> +};
...
> +static int sctp_setsockopt_encap_port(struct sock *sk,
> + struct sctp_udpencaps *encap,
> + unsigned int optlen)
> +{
> + struct sctp_association *asoc;
> + struct sctp_transport *t;
> +
> + if (optlen != sizeof(*encap))
> + return -EINVAL;
> +
> + /* If an address other than INADDR_ANY is specified, and
> + * no transport is found, then the request is invalid.
> + */
> + if (!sctp_is_any(sk, (union sctp_addr *)&encap->sue_address)) {
> + t = sctp_addr_id2transport(sk, &encap->sue_address,
> + encap->sue_assoc_id);
> + if (!t)
> + return -EINVAL;
> +
> + t->encap_port = encap->sue_port;
^^^^^^^^^^ ^^^^^^^^
encap_port is defined as __u16 is previous patch, but from RFC:
sue_port: The UDP port number in network byte order...
asoc->peer.port is stored in host order, so it makes sense to follow
it here. Then need a htons() here and its counter parts. It is right
in some parts of the patches already.
