[OLPC-devel] Re: Why ACPI DSDT tables?

[Eric W. Biederman]

David Woodhouse <dwmw2 at infradead.org> writes:

> On Mon, 2006-07-10 at 16:11 -0600, Eric W. Biederman wrote:
>> The part where firmware is almost always useful is to report board
>> specific details so we don't need board specific drivers.  Anything
>> that is board specific can't be reused (except as an example).  If
>> something can't be reused it has a questionable amount of purpose in
>> the kernel.  But so far that doesn't seem to be where the conversation
>> is about. 
>
> None of the DCON work can be reused either -- should we write all its
> support in ASL just so that it doesn't go into the kernel?

C is certainly better than ASL, for maintainability.

> Seriously, there's _plenty_ of board-specific stuff in the Linux kernel;
> just not a huge amount of it in the i386 architecture. 

Right.  The point is that is when things get board specific is when
the line gets fuzzy.  Ideally if the firmware can provide a board
specific table and the rest can be generic we are in better shape.

Currently i386 does not have a good infrastructure for boards and
their magic drivers, so board specific code is something we need
to be very careful with so we don't wind up with unmaintainable code.

Eric