Re: [PATCH 05/19] netfilter: nf_conntrack_ipv6: improve fragmentation handling

On Mon, 27 Aug 2012, Jesper Dangaard Brouer wrote:

quoted

quoted

When loading "nf_conntrack_ipv6" I run into issues.

When sending a fragmented UDP packet.  With these patches, the IPVS
stack will no longer see the fragmented packets, but instead see one
large SKB.  This will trigger a MTU path check in e.g.
ip_vs_dr_xmit_v6() and an ICMPv6 too big packet is send back.

 IPVS: ip_vs_dr_xmit_v6(): frag needed

Perhaps we could change/fix the MTU check in IPVS?
(This would also solve issues I've seen with TSO/GSO frames, hitting
this code path).

I guess this should use the same check as in IPv6 output, check
whether IP6CB(skb)->max_frag_size is != 0 and > MTU and only send
an ICMPv6 error in that case.

Hans have (already) proposed this solution:

 if ((!skb->local_df && skb->len > mtu && !skb_is_gso(skb)) ||
     (IP6CB(skb)->frag_max_size && IP6CB(skb)->frag_max_size > mtu)) {

And I have tested it works.
But I'm not sure about, if we really need the "!skb->local_df" check ?

Not necessarily, alternatively you could use:

if ((!IP6CB(skb)->frag_max_size &&
      (skb->len > mtu && !skb_is_gso(skb)) ||
     IP6CB(skb)->frag_max_size > mtu)

We just need to make sure that if frag_max_size is set that skb->len
is ignored. Either way is fine, but I'd suggest to keep it similar to
the variant used in IPv6.

Thus, we should extend you patchset with a patch, that also address the
MTU checks in IPVS.

Can you send over a patch for inclusion in the patchset? Besides this, its
ready for submission.