Hi, After reading through #350746, I wonder why iptables-new isn't used by default for the ssh jail. I understand and agree with the arguments about possible interactions with other protocols and thus the reason for not using it per default for all jails, but at least for ssh, where it's clear new auth attempts will use new connections, this shouldn't have any sensible drawbacks while decreasing the "annoyance-factor" of locking yourself out of your own server (like I just did, while playing with the pubkeys used by automatic backups :D). Cheers
have you seen http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=438901 and http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=438901#15 in particular?
That's a very good point. I'm gonna try cooking up a version of iptables-multiport which wraps the actionban with a check to see if there's a utmp entry from that IP. Would you accept something like this to be included like iptables-new currently is? It would deal with the annoying part without hitting this particular issue (I'd probably test this on at least one server for a while before submitting anyway). Cheers
Hi Leo, sorry, my brain must be resting at the moment, but I would need more detailed description of what/how you are trying to achieve, to give any promise ;-)