Re: [PATCH net-next 04/11] udp: Handle VRF device in sendmsg

[Tom Herbert <tom@herbertland.com>]

On Thu, Aug 13, 2015 at 1:59 PM, David Ahern <dsa@cumulusnetworks.com> wrote:
> For unconnected UDP sockets using a VRF device lookup source address
> based on VRF table. This allows the UDP header to be properly setup
> before showing up at the VRF device via the dst.
>
> Signed-off-by: Shrijeet Mukherjee <shm@cumulusnetworks.com>
> Signed-off-by: David Ahern <dsa@cumulusnetworks.com>
> ---
>  net/ipv4/udp.c | 22 +++++++++++++++++++++-
>  1 file changed, 21 insertions(+), 1 deletion(-)
>
> diff --git a/net/ipv4/udp.c b/net/ipv4/udp.c
> index 83aa604f9273..7af5052e3b1f 100644
> --- a/net/ipv4/udp.c
> +++ b/net/ipv4/udp.c
> @@ -1013,11 +1013,31 @@ int udp_sendmsg(struct sock *sk, struct msghdr *msg, size_t len)
>
>         if (!rt) {
>                 struct net *net = sock_net(sk);
> +               __u8 flow_flags = inet_sk_flowi_flags(sk);
>
>                 fl4 = &fl4_stack;
> +
> +               /* unconnected socket. If output device is enslaved to a VRF
> +                * device lookup source address from VRF table. This mimics
> +                * behavior of ip_route_connect{_init}.
> +                */
> +               if (netif_index_is_vrf(net, ipc.oif)) {
> +                       flowi4_init_output(fl4, ipc.oif, sk->sk_mark, tos,
> +                                          RT_SCOPE_UNIVERSE, sk->sk_protocol,
> +                                          (flow_flags | FLOWI_FLAG_VRFSRC),
> +                                          faddr, saddr, dport,
> +                                          inet->inet_sport);
> +
> +                       rt = ip_route_output_flow(net, fl4, sk);
> +                       if (!IS_ERR(rt)) {
> +                               saddr = fl4->saddr;
> +                               ip_rt_put(rt);
> +                       }
> +               }
> +

I really don't like this. It seems like you're putting device specific
code in a critical L4 data path function. Also, does ipv6/udp.c need
be updated similarly? Why can't VRF be abstracted out in routing
lookups?

Tom

>                 flowi4_init_output(fl4, ipc.oif, sk->sk_mark, tos,
>                                    RT_SCOPE_UNIVERSE, sk->sk_protocol,
> -                                  inet_sk_flowi_flags(sk),
> +                                  flow_flags,
>                                    faddr, saddr, dport, inet->inet_sport);
>
>                 security_sk_classify_flow(sk, flowi4_to_flowi(fl4));
> --
> 2.3.2 (Apple Git-55)
>
> --
> To unsubscribe from this list: send the line "unsubscribe netdev" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html