Hi,
On Sat, 7 May 2016 08:53:44 -0600 David Ahern <d...@cumulusnetworks.com> wrote:
> >> @@ -1193,7 +1193,12 @@ void ipv4_pktinfo_prepare(const struct sock *sk,
> >> struct sk_buff *skb)
> >> ipv6_sk_rxinfo(sk);
> >>
> >> if (prepare && skb_rtable(skb)) {
> >> - pktinfo->ipi_ifindex = inet_iif(skb);
> >> + /* skb->cb is overloaded: prior to this point it is IP{6}CB
> >> + * which has interface index (iif) as the first member of the
> >> + * underlying inet{6}_skb_parm struct. This code then overlays
> >> + * PKTINFO_SKB_CB and in_pktinfo also has iif as the first
> >> + * element so the iif is picked up from the prior IPCB
> >> + */
> >
> > Better if there was a guarantee in the code that inet_skb_parm layout stays
> > that way. Or instead just explicitly assign the iif.
>
> At this point inet_iif points to the vrf device so can't use it.
Initially I was thinking about explicitly getting the iif out of the
IPCB first, then assign to ipi_ifindex. Seems more readable, and less
fragile. However this depends on the IPCB/IP6CB layout relationship as
well (iif being first on both).
I assume documenting the IP{6}CB/PKTINFO_SKB_CB layout relationship at the
struct definitions would be beneficial.
Regards,
Shmulik
