The kudzu code loads the megaraid driver when it sees the supertrsk100. The megaraid code should be using a subvendorid list. In fact what I suspect should have happened is i960 ... humm, subid isnt known class i2o .. ah load i2o_core, config, block, scsi, net Bug 32082 now has a comprehensive list of subids attached
i960 is now 'unknown' as of 0.99-1. Should we actively be trying to load i2o stuff for anything that claims to be i2o, or is that not stable enough yet?
For the current -ac i2o should be solid for everything. There is one catch, some boards provide both i2o and 'native' interfaces. Notably the AMI Megaraid, so match on PCI idents, only take an unknown of class i2o through the i2o drivers