On 08/23/2016 06:22 AM, Vasiliy Tolstov wrote: ACCKKK!!! (and I mean that in the "Bill the Cat" manner, not the "Yes, I approve" manner) That is obviously wrong, and the only reason for it that I can think of is some odd merge problem during one of the final rebases of the patches, since I tested it innumerable times prior to that. Yes, but I wonder if anything else got screwed up at the same time. Interesting question. I guess the routes should be set whenever the device is ifup'ed, and maybe deleted when it's taken offline (if the tap itself isn't going to be deleted) I think this would have been less likely to happen if I was actually *using* type='ethernet' myself :-) If you want to send a patch to fix start/stop device I can look at it later today or tomorrow. After that I suppose it would e useful to go through a few tests and see what else is broken.