Re: [PATCH] flow disector: check if arp_eth is null rather than arp

[Eric Dumazet]

On Fri, 2017-01-13 at 18:25 +0000, Colin Ian King wrote:
> On 13/01/17 18:24, Eric Dumazet wrote:

> > It looks that we try very hard to add critical bugs in flow dissector.
> > 
> > This is embarrassing really.
> > 
> > I am questioning if the __skb_header_pointer() is correct
> > 
> > Why using hlen - sizeof(_arp) ?
> > 
> >                arp_eth = __skb_header_pointer(skb, nhoff + sizeof(_arp),
> >                                               sizeof(_arp_eth), data,
> >                                               hlen - sizeof(_arp),
> >                                               &_arp_eth);
> > 
> 
> Yep, the sizeof maybe dubious too, I overlooked that one; if somebody
> can clarify that then I'll send a V2 if it needs fixing up too.

I am pretty sure we should use hlen instead of (hlen - sizeof(_arp))

A V2 would be nice ;)