After reporting on the problems with the ethernet hardware of my SN95 mainboard here and here, I got a mail from Timo Pylvanainen who tried to fix the same issue. His approach was to fix the VPD data in order to get a correct checksum. However, it appears that the checksum varies for each individual chipset. Therefore it is not possible to set a single byte that would fix the checksum...

Today I got this idea, why not allow all Marvell/Yukon hardware (supposedly with bad VPD checksums) to pass the checksum test? Actually this is a combination of the patch from Timo and my own patch, but does not depend on a specific checksum value, as it passes all bad checksums for Yukon hardware.

Although I've written a small patch which implements this workaround, I will (and cannot) release it, since I did not test it myself. Perhaps I will tomorrow...

By the way: merry christmas everybody...