On 04/11/16(Fri) 07:15, Andreas Bartelt wrote: > On 11/02/16 17:41, Martin Pieuchot wrote: > > Hello Sebastien, > > > > On 01/11/16(Tue) 09:36, Sebastien Marie wrote: > > > Hi, > > > > > > I experiment problem with local connection on specific rdomain. > > > > > > I tried to make a simple and reproductible environment. > > > > Thanks for the nice report, could you confirm the diff below fixes your > > issue? > > > > The idea is to stop using lo0 for every routing domain. So with this > > diff a new loopback is created, and used, per rdomain. > > > > thanks, your patch works for me.
Thanks for testing, I just committed it. > I have removed the "rdomain 1" line from /etc/hostname.lo1 since this is set > automatically now when em2 is put into rdomain 1. Is this the intended way > to configure this or should "rdomain 1" still also be explicitly set for > lo1? The goal is to have the kernel setting 127.0.0.1 on the default lo(4) interface. If/when somebody does that you won't need a 'hostname.lo1' anymore but until then I suggest you keep adding 127.0.0.1 in there. > Another functional difference is that it's crucial now to "set skip lo" in > /etc/pf.conf which wasn't required for lo1 in rdomain 1 previously (i.e., I > had the skip rule explicitly set to lo0 instead of lo). I suppose the new > behaviour wrt pf(4) is the intended one. Indeed.
