PXE bootup issues on Multi-Ethernet CPU

[ARTDGuy]

So,
I have been experimenting with the Porteus PXE boot option (3.1 I believe) , which is really great.
My issue is that the CPU has multiple Ethernet ports. If I boot Porteus from the "primary" Ethernet port on the booting CPU (the lowest one in PCI space), everything works fine and great.

If I choose another Ethernet port on the bootin CPU, the Porteus Boot menu shows up but when it attempts to load in the startup script but after the
searching for porteus-v3.1-x86_64.sgn file
udhcpc (v1.21.1) started

I just get a bunch of
Sending discover...
Sending discover...


So it mostly comes up but when it attempts to get the rest of the system, I believe it goes for the first found Ethernet port and fails to continue. I can prove this by moving the Connection to the first port of the booting CPU and everything continues on fine.
Basically, it boots the startup menu and script from a secondary Ethernet , but then finishes the download from the primary Ethernet.

I hope this makes sense.

I am looking at the initrd package and attempting to modify linuxrc but I don't see where the control section to select the ethernet really comes into play, and I don't have the experience to figure it out.
Anyone know of an easy hack?

[Bogomips]

Welcome to Porteus. The network expert should be turning up on next but one TZ, and hopefully you will get some light thrown on this issue.

[ARTDGuy]

From what I can figure out , looking at the code in linuxrc, it would seem that the code scans for all the Ethernet ports in lspci, then attaches the driver to them. Then I can't figure out how one is picked to become the one, and udhcpc is executed.

There is an ifconfig in the same area, but it is after the udhcpc call, it kind of seems to be called before it is required.

There is a part where it appears to be going through every network card and trying to attach a driver to them. That seems to be some kind of catch-all.

[ARTDGuy]

Ok , there is another script for udhcpc under /usr/share/udhcpc/default.script that I wasn't aware of, so it looks like my issue may reside in there.

[burdi01]

I do not have a Porteus 3.1 version anymore - 3.2 is current for quite some time now.
When looking at a running Porteus Mate 3.2.2 I do not think that the scripts are very different though.
If you are technically up to it have a look at the /opt/porteus-scripts/pxe-server script and especially the /srv/pxe/boot/pxelinux.cfg one: the server puts the "additionals" channels on the first nic.
If things go over your head then do not boot your clients from a non-first nic. 8)

[ARTDGuy]

Hey,
The /opt/porteus-scripts/pxe-server and the /srv/pxe/boot/pxelinux.cfg seem to be server side. It is interesting to see how the /opt/porteus-scripts/pxe-server creates and packages the initrd.xz, but my issue is client side.

The client has multiple Ethernet ports, but the script just attaches to eth0 it appears.
I was hoping to find some iterative script that could perhaps check the /sys/class/net/*/ properties on the client to find which one is actually connected to Ethernet , which is sufficient at this point because only one port is in use for this case.
Its too bad there isn't some passed in option from the PXE BIOS that says, "i was using PCI 01:00:00 ethernet port" and then it would be easy.

Anyways, thanks for pointing me at the server side stuff.

Tony