On Sun, Jan 15, 2017 at 05:49:51PM +0100, René Wagner wrote: > tags 766838 + patch > thanks > > > Dear maintainers, > > my testing indicates that there are two distinct problems. First, for > whatever reason, /etc/network/if-up.d/ntpdate is called with $IFACE = --all > in addition to legitimate interfaces. This causes a race condition: If the > $IFACE = --all instance wins, the other instances are prevented from doing > anything useful. > > Secondly, /etc/network/if-up.d/ntpdate is called somewhat early, i.e. when > the interface is already up but before an IP address has been obtained via > DHCP. > > The attached patch fixes both issues for me. It contains a whitelist of > interfaces /etc/network/if-up.d/ntpdate should act upon (eth*, wlan*) to > avoid the bogus $IFACE = --all case. In hindsight, perhaps this should be > turned into a blacklist containing "--all" only.
Yes, that would make more sense. I at least have a laptop were it's not called wlan0. > The second part of the patch builds upon Hartmut Buhrmester's suggestion to > insert a "sleep" call, but parses ifconfig output until an IP address has > been assigned. This part handles the IPv4 case only since I don't have an > IPv6 network to test with. I don't think we should have a while loop in there, or at least time out after some time. But we really shouldn't get called if it's not up in the first place. And calling us with "--all" doesn't make sense at all. Kurt
