Re: CVS commit: src

[Kimmo Suominen <kim%netbsd.org@localhost>]

On Wed, Oct 14, 2020 at 10:13:21PM +0100, Roy Marples wrote:
> On 14/10/2020 20:07, Kimmo Suominen wrote:
> > - not interfere with (static) IPv4 configuration
> 
> How do you expect
> ifconfig_vioif0='dhcp rtsol'
> to work?

When I added the code for the rtsol keyword, I did not imagine that
it would be used with the dhcp keyword. In other words, I would have
considered that an invalid configuration (among the many other invalid
configurations you can make and that won't be checked against).

However, it would be cool if that worked.

Is there a better way to avoid having dhcpcd break a static IPv4
configuration?  Using the rtsol keyword should neither enable DHCP for
IPv4 nor enable IPv4LL address autoconfiguration.

I've adjusted the way dhcpcd gets started by rtsol: it was not necessary
to use nodhcp6 or ipv6ra_noautoconf. It was just part of how I got the
network configuration to work again after upgrading to a current that no
longer has in-kernel RA processing. Once it worked I didn't change it
further. So now the O and M flags should no longer be ignored.

> > I also think it is very good that with this change we once again
> > have backwards compatibility for configuring static network
> > addressing.
> 
> We've never lost it.

This is where we seem to have a difference in experience. For me the
IPv4 configuration was broken.

Kind regards,
+ Kimmo