UDHCPC modifications?
-
Have there been any changes in the FOS kernel that would affect how UDHCPC operates, or how long it waits for a DHCP lease?
We’re on kernel 4.7.1
-
No.
The 4.7.1 kernels actually have IP loading back as a default part of the kernel load up (meaning 4.7.1 and up should work again for 1.2.0 and earlier too).
-
Just spoke with @Tom-Elliott
He now takes the IP that iPXE gets from DHCP, and passes it as a kernel argument - and the latest FOG kernel 4.7.1 accepts this IP and uses it. So if the IP is already configured in the kernel, UDHCPC doesn’t need to get a new IP - but it still gets DNS addresses. If for whatever reason the kernel doesn’t get passed an IP, UDHCPC will request one. It operates as a sort of failsafe now.
This also means the latest kernel should work with 1.2.0 in theory - no promises.
Also - This change allows FOG to work with PortFast turned OFF. Yes, you heard me right. Tested working at work.
-
I feel like a middle man.
Tom says - the kernel loads the NIC as part of the kernel loadup now.
The difference - the time the nic is down is way, way shorter. So short, the switches apparently don’t even notice. So you can get away with portfast being off.
-
It seems that this change would only affect certain switches with certain configurations - with portfast being off. I think older switches would perhaps work but newer ones won’t, too fast of a CPU in them. That’s what I’m seeing right now.
-
Sorry to pop this bubble… Due to communication issues, I thought PortFast was turned off for this case. I’ve found out that PortFast was in fact turned on.
