Short a moment ago NetworkManager performed successful DoS attack on remote Fedora 13 PC, which I administer. Well, it is my idiocy too, but should be NM behave so stupidly as just happened to me? This is whole story: - PC act as normal workstation in one firm several tens kilometers away, with GNOME GUI, but there are some KDE/Qt apps too. - it acts as well as small router/internet gateway with interfaces eth0(192.168.0.254/24), eth1(192.168.0.254/24), eth2(192.168.0.254/24), eth3(94.68.30.114/28). Gateway IP is 94.68.30.113. - I now performed system update. But what unfortunately I not noticed - new knetworkmanager version depends on NetworkManager (which wasn't previously installed) and the one was installed - and even worse, was set for run in runlevels 2/3/4/5. And when I restart this computer, system became inaccessible. When I call in some man to that PC and navigate him over phone, there was malformed routing tables, e.g default route was: "default via 94.68.30.113 dev eth0" I think there is probably something wrong in current NM arrangement, as in this case: - install script (according to NM init script) activate NM for all runlevels - NM will start control interfaces although there isn't any profile defined - and although no interface has stated "NM_CONTROLLED=yes" in its ifcfg - and finally start control them in worst possible case. In NM history it isn't first case when he destroyed my manually created network configuration. Yes, setting "NM_CONTROLLED=no" maybe solve much of this, but - should not be default NM installation configuration more restrictive (e.g. not set for run in any runlevels, not control interface when its BOOTPROTO is "static" etc)? At least until NM working logic will be at the level, when it wil not do such negative things. Franta Hanzlik -- users mailing list users@xxxxxxxxxxxxxxxxxxxxxxx To unsubscribe or change subscription options: https://admin.fedoraproject.org/mailman/listinfo/users Guidelines: http://fedoraproject.org/wiki/Mailing_list_guidelines
