Re: [PATCH net 1/2] packet: do skb_probe_transport_header when we actually have data

[Eric Dumazet <eric.dumazet@gmail.com>]

On Fri, 2015-11-06 at 22:02 +0100, Daniel Borkmann wrote:
> In tpacket_fill_skb() commit c1aad275b029 ("packet: set transport
> header before doing xmit") and later on 40893fd0fd4e ("net: switch
> to use skb_probe_transport_header()") was probing for a transport
> header on the skb from a ring buffer slot, but at a time, where
> the skb has _not even_ been filled with data yet. So that call into
> the flow dissector is pretty useless. Lets do it after we've set
> up the skb frags.
> 
> Fixes: c1aad275b029 ("packet: set transport header before doing xmit")
> Reported-by: Eric Dumazet <edumazet@google.com>
> Cc: Jason Wang <jasowang@redhat.com>


> +	if (!packet_use_direct_xmit(po))
> +		skb_probe_transport_header(skb, 0);
> +

Thanks Daniel for working on this.

The if (!packet_use_direct_xmit(po)) test looks dubious.

Setting transport header has nothing to do with bypassing qdisc ?

This might lead to hard to debug problems, for drivers expecting
transport header being set ?

Maybe this needs a special socket flag, but this does not seem worth the pain.

(This would be different of course if trafgen was not defaulting to qdisc bypass)

Thanks.