dhcpcd-discuss

On 09/07/2020 13:57, Mike Crowe wrote:
> On Wednesday 08 July 2020 at 12:38:34 +0100, Roy Marples wrote:
> > On 08/07/2020 12:04, Roy Marples wrote:
> > > This is dhcpcd's fault.
> > > I've just added a patch to bump the route metric up for IPv4LL routes
> > > and ensure it's evaluated last on other OS's.
> > >
> > > https://roy.marples.name/cgit/dhcpcd.git/commit/?id=b7ede4d38fab12fd093dcb51416d83764de1557e
> >
> > This one also:
> > https://roy.marples.name/cgit/dhcpcd.git/commit/?id=3c0123be03d7315a3b35bc0fb90b6d3463393bcd
> >
> > And I tested with this patch on Linux .... seems to work fine.
>
> Thanks. That seems to solve the problem. I'm not really sure why it's worth
> adding a default route for interfaces with only a link-local address. The
> only reason I could come up with was that it allows communication with
> devices on the same link that have non-link-local addresses, but only if
> those hosts will also know to route their replies to link-local addresses
> back to the right interface. Is there another reason?

You're correct.

> > Are you running dhcpcd separately per interface?
>
> Separately for each interface since I need to cope with interfaces coming
> and going.
>
> > As you should not see that with one dhcpcd instance with the released
> > versions.
>
> I shall investigate whether it would be possible to run only a single
> instance.

It's possible - dhcpcd will spot interfaces coming and going.

> Thanks for the speedy response and fix.

Welcome.

Roy