When using dhcpcd, it should fork after configuring the interface and hang around. This is not happening currently, it exits, removing the pid file that is used to test for it during net.eth0 stop, causing the system to be left in a state where resolv.conf's aren't fixed, etc. I do not know what is causing this at present, but will investigate further. Reproducible: Always Steps to Reproduce: 1./etc/init.d/net.eth? start (for your dhcp interface) 2.ps aux | grep dhcpcd 3.ls -la /var/run/dhcpcd* Actual Results: No pid file and no dhcpcd process. Expected Results: There should be a pid file and dhcpcd process hanging around. I'm using the wireless.sh script from the forums as a preup. It appears to work properly and down's the interface before dhcpcd is called. Dhcpcd successfully configures interface but then exits where it should not.
this usually only happens when something is wrong check your log files to see why dhcpcd bailed does it bail if you run it yourself ?
I have figured it out. Whenever the dhcp server doesn't provide a lease period, the dhcpcd client assumes that it is infinite and as such assumes that it will not need to hang around to renew the information, thus exiting. 2 solutions: 1) Document this behaviour and leave it up to the user. 2) Have a default lease period specified in /etc/init.d/net.eth0. (option is -l time_in_seconds) I'm leaning towards 1) as 2) makes decision for the user. Note that this behaviour is relevant when like me you alternate between a fixed network and a wireless one. For me it must be shutdown with -z or when I get to work I sit with my resolv.conf from home.
agriffis: perhaps add a small note in conf.d/net right after where it mentions the timeout problem ? if not we can just punt this bug
I've documented this behaviour in net.example in CVS
Fixed in baselayout-1.11.7-r1
