RESOLVED: Finally! rtl8139 nic pxe/gpxe/grub boot after alpha 2.15

[pigdog]

Hi,

I've tried different stuff over the last couple of days and I'm getting nowhere.

Everything worked prior to alpha 2.15 and now, no matter what I have tried, it won't boot.

I know the hardware works under ubuntu, I'm out of ideas.

I'll just have to wait and see if anyone else hits the same snag.

Cheers.

[pigdog]

Hi,

Well I was going thru my parts bin and I found a D-Link DFE-538TX 10/100.

It had a rtl8139 on board but instead of 10ec:8139 it was 1186:1300.

I tried it but same kernel panic and junk.

[pigdog]

O.K.  I managed to get the RTL8139 running but I changed a bunch of stuff, even MTU junk, so I don't know if it is a combination of things or just the last resort wtf thing I tried.

So, I'm going to reload from scratch and see if it can be replicated.

Cheers.

[pigdog]

O.K. now this is getting frustrating.

I rebuilt from scratch loaded the latest and when I went to try I got a completely different error.

So I tried a couple of things and got nowhere.

So I reloaded from scratch and tried things again.

System (A)
EP-8KDA3+ NVIDIA nForce3-250 Socket 754 1GbE LAN - not using the on board Nvidia chipset/BootROM because it's toast.  Using RTL8139 card.

System (B)
ECS-P4S5A/DX+ Socket 478 RTL8201BL 10/100 LAN - not using the on board Realtek because it doesn't PXE boot and testing with the RTL8139 card to match above for troubleshooting.

System (C)
Foxconn Winfast 6150K8MA-8EKRS and Nvidia MCP51 LAN - boots from Core no problem.

System A & B where tested with 8139C and 8139D Realtek LAN Cards  ( I have 2 of each type).

On PXE Boot using ROM-O-Matic Etherboot 5.4.4 both reported same error:

0200 AX:0212 BX:2400 CX:001 DX:0100

I've rebuilt the 5.4.4 version twice ans tried the 5.4.3 variant with the same result.

I've built a grub pxe boot disk as per the wiki and the original error type was returned (traceback, kernel panic, etc).

I'm going to stick with the grub pxe boot disk for testing.  I don't need this other variable.

Cheers.

[Marie.O]

pigdog,

make sure you disable your onboard NICs completely when you are using plugin NICs (PCI or USB) on a media director.

[pigdog]

Hi posde,

Yes it's been checked, checked, double-checked, what happens if I change this, what happens with that, what about this?

I'm shaking my head here.  What is it with the 8139 chipset? 

I could easily walk away and buy another NIC card but it would not help someone else.

So once more into the breach - and it'll probably turn out to be something really dumb, but, it worked prior to 2.15 and not afterwards.

What changed?  I thought maybe I was getting mapped to the wrong driver at one time but someone else should have reported problems by now if that was the case.

I will keep at it.

Thanks for the input - Cheers.

[pigdog]

Hi,

The saga continues...

System (B) The ECS-P4S5A/DX+ Socket 478 I put a rtl8139 into. 

In the BIOS settings of the AMIBIOS Version 1.21.12 I changed Plug and Play Aware OS to no.

Using the grub pxe boot disk (as per the wiki) I selected new media director and managed to receive "We announce ourselves to the router" at which point I shut it off and checked the Core/Hybrid.

The Core/Hybrid shows #31 as arch i386, the IP and MAC.

I rebooted with the grub pxe boot disk and no joy - typical error message.

I swapped the NIC to System (A) EP-8KDA3+ NVIDIA nForce3-250 Socket 754 and no joy.

I created an etherboot ROM-O-Matic disk version 5.4.4 and tried it in system B but I get RX errors and the system reboots.

In system A with etherboot ROM-O-Matic I get

-Trying to load:pxelinux.cfg /01-00-08-54-11-98-e3
-could not find kernel image:31/vmlinux
boot:

So this is looking a little better.

Time for sleepy.

[pigdog]

O.K.  I've been going nuts the last couple of days trying to figure out why System B had RX and UDP errors.

Why System A didn't have the errors but had the traceback, kernel init, eth0, blah, blah.

I've used Wireshark, ethtools, looked at MTU's and other junk.

I've been through four different NIC cards, four different CAT5 cables, different pxe etherboot and grub gpxe disk versions.

I eliminated the 8139cp driver complaint by forcing the 8139too driver to run.

Eventually I read about the difference between "noacpi": "pci=noacpi".

I went into /tftpboot/pxelinux.cfg and edited the default boot script to remove acpi=off.

I booted System A and away we go!

So now that it has booted up and is running (moon48) I changed the default back to acpi=off.

My MD hooked up to my LCD (moon31) did not have any issues with acpi=off.

The BIOS in system B has no settings for turning on or off acpi.

I'm glad that's over with.