On 02/27/2013 10:26 PM, TJ wrote: > On 28/02/13 03:15, Eric Blake wrote: >> Does your series ever allow dnsmasq and dhcprelay to run at the same >> time, or can we use a single pid_t field that covers the mutually >> exclusive choice of which helper is running based on the rest of the config? >> > When local DHCP server services are disabled dnsmasq is still launched since there are several non-DHCP settings in the generated config. Right. a DNS server is run by default on any network that has an IP address defined. > > In my test-bed for these patches dnsmasq and dhcp-helper will be started alongside each other. > > If this series is accepted I was intending to propose adding <dns enable='(yes|no)'/> You can consider that one pre-approved, as we've already discussed it to death and decided it was a good idea; just nobody has gotten around to doing it that way yet :-) (in other words, go ahead and write it; you'll be getting it off the todo lists of at least two of us.) > in the same style used for <dhcp> so that dnsmasq can be Actually the enable attribute is totally unnecessary for <dhcp>, since an absence of <dhcp> implies that it is disabled (and indeed that is what already happens), and in the case of <dhcp relay='yes'/>, it is also implied (since a local dhcp server and dhcp relay are mutually exclusive). So, if there is no <dhcp> (or if it's <dhcp relay='yes'/>) and there is <dns enable='no'/>, then dnsmasq won't be run. The only reason we need an enable attribute for DNS is that the original default (when there is no <dns> element) is to enable the dns server, and we must preserve backward compatibility. -- libvir-list mailing list libvir-list@xxxxxxxxxx https://www.redhat.com/mailman/listinfo/libvir-list