On Jul 8, 12:24pm, Scott Stevens wrote:
> Subject: Re: PowerROM support in latest kernel
> Thomas Boroske writes:
>  >
>  > > We are talking to alsystems about a way of identifying PowerROM'd
>  > > cards in a different way... this could get messy tho'.
>  >
>  > I'm not sure. AFAIK, although the modules actually on the ROMs are
>  > identical (I believe) for all types of cards, I think that there are
>  > 'Morley' PowerRoms, 'Castle' PowerRoms etc... So the roms are different
>  > anyway (perhaps it's just the type of eprom used, not the content),
>  > so perhaps it would be easy for them to change the ID.
>
> This is what we thought...
>
>  > Back to the actual problem: RiscBSD uses podule identification to select
>  > the correct driver for any card found. In this case this breaks the
support,
>  > so wouldn't it be possible to make a kernel where you tell in the config
>  > to just assume a card type x in slot y, not to test for it ?
>
> Well you can get the kernel to only check specific slots. But all that
> really does is only *present* a limited range of podules to the driver
> for identification.
>
> ie the probe process is something like this:
>
> podulebus attaches and scans podules creating podule info strcutures.
> Each unattached podules is passed to *every* possible driver that could
> have a card in that slot.
>
> The drivers simply check the ID and say whether they can drive it or
> not. So the identification needs to be in the driver's match routine,
> not in the parent podulebus routines.

I have a piece of code which I wrote for manually configuring the console which
I could hack such that it asks you a series of yes/no questions to override the
configuration on certain podules. Would this be of any use? Of course this
means that you'd have to sit there and answer questions every time you boot the
machine.

Cheers

Rob