On Sun, Oct 05, 2025 at 09:39:37AM -0400, Jon Maloy wrote:
On 2025-10-03 01:33, David Gibson wrote:
On Thu, Oct 02, 2025 at 08:34:03PM -0400, Jon Maloy wrote:
Bug #120 asks us to use the true MAC addresses of LAN local remote hosts, since some programs need this information. These commits introduces this for ARP, NDP, UDP, TCP and ICMP.
I have a bunch of more detailed comments on the patches, but they're a bit confused, because for the earlier ones I hadn't read the whole series to get the whole context.
There is a problem here in that you're not consistent about whether the neighbour table is indexed by host side addresses or guest side addresses. You'll mostly get away with this, because they're usually the same. But it will result in bugs on the edge cases.
Having read the whole series now, I think it will be simplest to standardize on indexing by guest side addresses. That invalidates some of my earlier comments, but instead you'll need to use nat_inbound() when you populate the table from netlink updates.
I think I agree. After all this is all about what we are showing the guest in the form of IP and MAC addresses, so it is conceptually clearer.
It doesn't change the implementation much, but it makes it easier to understand which special cases we need to consider.
Agreed. I think it will also make this easier to extend for guest-side ARP lookups in future, which we'll want for handling multiple guests.
See my next email.
-- David Gibson (he or they) | I'll have my music baroque, and my code david AT gibson.dropbear.id.au | minimalist, thank you, not the other way | around. http://www.ozlabs.org/~dgibson