Today, 15 people from the Genode community gathered in the very east of Germany to enjoy a long weekend of hacking fun projects, hiking through the Zittauer mountains, and chatting about everything and Genode. This year the hacking topics range from Sculpt TC related developments and packaging over retro-game library porting to ADA/SPARK based reliable utilities for XML handling. For me the Hack’n’Hike is a wonderful opportunity to revive the abandoned work on Cobalt.
This night, I had the first break-through. After I enabled the three NICs one-by-one I discovered that the MAC addresses equal and also look not like documented for the apu2.
[nit -> nic_drv] MAC address 19:00:00:0d:19:02
The reason was a bug in the iPXE intel NIC driver, which fortunately was already fixed upstream. The buggy version did always try to fetch the MAC address from EEPROM with a fall-back on a valid address in a device register. The proper logic is the other way round: Read from EEPROM if the MAC address was not already autoloaded by boot firmware.
[init -> lan1_drv] MAC address 00:0d:b9:47:83:48
[init -> lan2_drv] MAC address 00:0d:b9:47:83:49
[init -> lan3_drv] MAC address 00:0d:b9:47:83:4a
After this discovery I came to the conclusion that an update of our iPXE port dde_ipxe is overdue.