On Tue, Nov 15, 2022 at 11:29 AM Ryota Ozaki <ozak...@netbsd.org> wrote: > > On Mon, Nov 14, 2022 at 8:52 PM Roy Marples <r...@marples.name> wrote: > > > > On 14/11/2022 11:04, Martin Husemann wrote: > > > This clearly is a layering/abstraction violation and would have been > > > good to discuss upfront. > > > > > > Where do you make use of that information? What about other packet > > > injection > > > paths? > > > > The next commit uses it in if_arp to ensure that the DaD probe sending > > interface > > hardware address matches the sending hardware address in the ARP packet as > > specified in RFC 5227 section 1.1 > > > > I couldn't think of a better way of achieving this. > > RFC 5227 says senders must follow the spec but doesn't say receivers' > check is must IIUC. > > I don't think it is a good idea to increase the mbuf size just for > broken clients. > I think it is better to make the strict check optional (by sysctl or > something) and use mtag, > so the change doesn't impact on most of us.
Well... another possible option is to unionize l2_* with pattr_*. This is possible (IIUC) because l2_* are used only on receiving packets while pattr_* are used only on sending packets. Am I missing something? ozaki-r