In the readonly nfs root case we don't want to down the interface before requesting an ip, as that kills the root filesystem. We need a flag that disables this behaviour.
Created attachment 100869 [details] patch to implement -ns
This patch works for me if i put DHCLIENTARGS=-ns in /etc/sysconfig/network.
Hmmm, in fact the interface isn't downed, the hang is just when doing "ifconfig eth0 0 up". This should be a nop since the interface is already up. I dunno whats causing this, will investigate. Hold on the patch until this has been further analyzed.
Ah, so the "0" is whats causing it. It sets the interface ip to 0.0.0.0, which is needed when you broadcast for the initial dhcp ip. What we need to do is to make it do a renew instead, and carry over the lease from the old dhcp call.
Created attachment 100941 [details] Don't set ip to 0.0.0.0 if we have a lease for the current ip already
With this patch things work if i carry over the dhcp client leases file from the initrd.
Patch applied to dhcp-3_0_1rc12-8 Dan
Closing, this is working fine
An advisory has been issued which should help the problem described in this bug report. This report is therefore being closed with a resolution of ERRATA. For more information on the solution and/or where to find the updated files, please follow the link below. You may reopen this bug report if the solution does not work for you. http://rhn.redhat.com/errata/RHBA-2004-566.html