It makes sense that a lower metric should work, and as you say that then relieves the burden of saving/restoring the original route. One thing that is not clear to me in this regard is the relationship, if any, between metric and protocol...should the new route in any case be marked as "protocol static"? I'm happy to test any changes, so let me know if I can help. [ On a related note, since I originally reported this issue, I have found that more-or-less consistently, my attempt to set the new route *from the script* simply fails. I invariably end up setting it by hand once vpnc has finished the setup. Since I can set the new route by hand, it seems like a race, but adding delays here and there into the script seemed to make no difference. I cannot think why this should be...my kernel is whatever kernel Ubuntu "wily" thinks it should be...but I mention it here in case it is somehow relevant ]. Thanks, Shaheed On 6 April 2016 at 12:10, David Woodhouse <dwmw2 at infradead.org> wrote: > On Fri, 2016-02-12 at 20:43 +0000, Shaheed Haque wrote: >> [ This report concerns a defect originally reported at >> https://bugs.launchpad.net/ubuntu/+source/vpnc-scripts/+bug/1544802 ] >> >> Bug Description >> >> The /usr/share/vpnc-scripts/vpnc-script handles updates to the >> default >> route using two different codepaths. In one codepath, the command "ip >> route replace" is used to update the original default route with new >> one (and to restore it later). The replace command in the update case >> does not work if the original route default route has non-standard >> attributes. > > Hi Shaheed, thanks for looking at this. > > I wonder if it might be better just to *leave* the original default > route in place. Just add the new default route with a lower metric, so > the old route doesn't get used? > > Then we don't have to worry about preserving the details of the old > route at all, do we? Or the fact that there might be *multiple* default > routes pre-existing? All we need to do is ensure that our new default > route has a lower metric than any which already existed. > > -- > David Woodhouse Open Source Technology Centre > David.Woodhouse at intel.com Intel Corporation >
