On Tue, Mar 28, 2017 at 6:26 AM, Aviad Yehezkel <avia...@mellanox.com> wrote:
> This patch adds TLS_TX and TLS_RX TCP socket options.
>
> Setting these socket options will change the sk->sk_prot
> operations of the TCP socket. The user is responsible to
> prevent races between calls to the previous operations
> and the new operations. After successful return, data
> sent on this socket will be encapsulated in TLS.
>
> Signed-off-by: Aviad Yehezkel <avia...@mellanox.com>
> Signed-off-by: Boris Pismenny <bor...@mellanox.com>
> Signed-off-by: Ilya Lesokhin <il...@mellanox.com>
> ---
> include/uapi/linux/tcp.h | 2 ++
> net/ipv4/tcp.c | 32 ++++++++++++++++++++++++++++++++
> 2 files changed, 34 insertions(+)
>
> diff --git a/include/uapi/linux/tcp.h b/include/uapi/linux/tcp.h
> index c53de26..f9f0e29 100644
> --- a/include/uapi/linux/tcp.h
> +++ b/include/uapi/linux/tcp.h
> @@ -116,6 +116,8 @@ enum {
> #define TCP_SAVE_SYN 27 /* Record SYN headers for new
> connections */
> #define TCP_SAVED_SYN 28 /* Get SYN headers recorded for
> connection */
> #define TCP_REPAIR_WINDOW 29 /* Get/set window parameters */
> +#define TCP_TLS_TX 30
> +#define TCP_TLS_RX 31
>
> struct tcp_repair_opt {
> __u32 opt_code;
> diff --git a/net/ipv4/tcp.c b/net/ipv4/tcp.c
> index 302fee9..2d190e3 100644
> --- a/net/ipv4/tcp.c
> +++ b/net/ipv4/tcp.c
> @@ -273,6 +273,7 @@
> #include <net/icmp.h>
> #include <net/inet_common.h>
> #include <net/tcp.h>
> +#include <net/tls.h>
> #include <net/xfrm.h>
> #include <net/ip.h>
> #include <net/sock.h>
> @@ -2676,6 +2677,21 @@ static int do_tcp_setsockopt(struct sock *sk, int
> level,
> tp->notsent_lowat = val;
> sk->sk_write_space(sk);
> break;
> + case TCP_TLS_TX:
> + case TCP_TLS_RX: {
> + int (*fn)(struct sock *sk, int optname,
> + char __user *optval, unsigned int optlen);
> +
> + fn = symbol_get(tls_sk_attach);
> + if (!fn) {
> + err = -EINVAL;
> + break;
> + }
> +
> + err = fn(sk, optname, optval, optlen);
> + symbol_put(tls_sk_attach);
> + break;
> + }
> default:
> err = -ENOPROTOOPT;
> break;
> @@ -3064,6 +3080,22 @@ static int do_tcp_getsockopt(struct sock *sk, int
> level,
> }
> return 0;
> }
> + case TCP_TLS_TX:
> + case TCP_TLS_RX: {
> + int err;
> + int (*fn)(struct sock *sk, int optname,
> + char __user *optval, int __user *optlen);
> +
> + fn = symbol_get(tls_sk_query);
> + if (!fn) {
> + err = -EINVAL;
> + break;
> + }
> +
> + err = fn(sk, optname, optval, optlen);
> + symbol_put(tls_sk_query);
> + return err;
> + }
This mechanism should be generalized. If we can do this with TLS then
there will likely be other ULPs that we might want to set on a TCP
socket. Maybe something like TCP_ULP_PUSH, TCP_ULP_POP (borrowing from
STREAMS ever so slightly :-) ). I'd also suggest that the ULPs are
indicated by a text string in the socket option argument, then have
each ULP perform a registration for their service.
> default:
> return -ENOPROTOOPT;
> }
> --
> 2.7.4
>
