[Dnsmasq-discuss] dnsmasq: dhcp-server with disabled dns-server: /etc/hosts ignored
Geert Stappers
stappers at stappers.nl
Tue Nov 1 22:43:25 UTC 2022
On Mon, Oct 31, 2022 at 09:07:40PM +0000, Simon Kelley wrote:
> On 21/10/2022 22:23, Michael Traxler wrote:
> > Hello,
> >
> > dnsmasq-2.87:
> >
> > It took me quite some time to figure this out :-)
> >
> > When you disable the DNS-server via
> > port=0
> >
> > the /etc/hosts file is ignored and also the
> > addn-hosts=<fn>
> > is ignored.
> >
> > This then results in the problem, that the
> > DHCP-server still works, but if you use the
> >
> > dhcp-host=da:7a:3a:62:7d:d2,trb
> >
> > where trb is a name in /etc/hosts
> > dnsmasq reacts with:
> >
> > DHCPDISCOVER(enp2s0) da:7a:3a:62:7d:d2 no address available
> >
> > I checked via strace that it really doesn't even open
> > /etc/hosts nor the given file via the "addn-hosts" option.
> >
> > Looks for me like a bug. I can not find a logical reason, why a dhcp-server
> > *needs* to have a running DNS-server.
> >
> > Or is it a feature? :-)
>
> It's at least a documentation bug. I think the reasoning behind the
> behaviour is that setting port=0 disables all the DNS functionality, which
> can be useful for its own sake, but also saves machine resources; no memory
> is allocated for cache and all the configuration infrastructure for DNS is
> not initialised, hence /etc/hosts doesn't get read.
>
> It would be trivial to remove half a dozen "if (port != 0) { }" lines from
> the code, and have all of the DNS stuff still live apart from actually
> listening on port 53,
Nice
> but that's not necessarily an improvement. There's not increase in
> functionality, since the same behaviour can be configured simply by
> adding the the addresses to the dhcp-host lines and that works with
> the existing code even when port=0 is set.
An already mentioned improvement is /etc/hosts content being respected.
The
> > dhcp-host=da:7a:3a:62:7d:d2,trb
> >
> > where trb is a name in /etc/hosts dnsmasq reacts with:
from above.
> My feeling is that the best fix to this is a heads-up in the man page,
> but I'm happy to entertain arguments to the contrary.
> Simon.
Groeten
Geert Stappers
--
Silence is hard to parse
More information about the Dnsmasq-discuss
mailing list