> But, I would NOT like the pckbd to be the console keyboard.
> From looking at the code this doesn't seem possible, 
> pckbc_cnattach() unconditionally calls pckbd_cnattach() which
> will turn the pckbd into the console keyboard.

The code "doesn't know better".

At first, a way has to be found to find out that there
is an USB keyboard present in early system startup, and to
get the physical attachment parameters (eg PCI device number
of the USB interface). Is the BIOS of any use for this?
If the console uses an USB keyboard, pckbc_cnattach() should
probably not be called at all; there should be an
uhci/ohci_kbdattach() which can be called from machdep
consinit().
To get this working, it will need a number of tweaks to the
i386 vm and bus_space code to be able to map/access memory
ranges in early system startup.

I'm quite clueless about the legacy hardware emulation
facilities of USB controllers. Is this worth using?

best regards
Matthias