Dell Dimension 2400 LAN issues

[GiGaBiTe]

I was given a couple of Dell Dimension machines awhile back, and I've had some time to play with them recently. One is a 2350 and works fine, the other is a 2400 and the Broadcom BCM4401 Ethernet Controller doesn't work.

Long story short, it appears the 1k ROM for the BCM4401, which is an Atmel 93C46 is corrupt. The MAC address was lost and is just FFs, and some other configuration data was as well. Just by chance, the 2350 also uses the same BCM4401, so I was able to use the ROM to figure out how to insert the MAC address into the corrupt ROM, but it still doesn't work. I'm also not able to verbatim copy the ROM from the 2350 on to the 2400, the machine gets very upset and will say the NIC has some unspecified failure. Windows won't boot without crashing to a corrupt display..

This machine is probably too new and too undesirable for anyone to have, but if anyone does have it, would you be able to dump the BCM4401 ROM and post it here? I'm not able to find it anywhere, and I'd rather not have to buy an entire second board to get it. The ROM is a small SOIC-8 chip right next to the BCM4401 behind the PCI slots.

I have a sneaking suspicion that the BCM4401 itself may have some sort of failure, but I won't know until I can get a known good ROM. If it is bad, I can't replace the chip without having a good ROM.

 

[graceanderson]

Does the b44 driver under Linux still bind with the blank MAC? If it does, you can read and write the 93C46 through it with ethtool -e / ethtool -E, which lets you try a few config layouts without pulling the chip each time. On these Broadcom parts the EEPROM holds the MAC plus a handful of config words, and a checksum or length field elsewhere is a plausible reason a verbatim copy from another board gives an unspecified failure. Comparing the 2350 dump word by word, ignoring the all-FF regions, might show which fields matter. If the chip itself is dead, a cheap PCI NIC is the practical fix.