Subject: Re: ts102.c
To: Erik Berls <cyber@ono-sendai.com>
From: Martin Husemann <martin@duskware.de>
List: port-sparc
Date: 10/06/2002 19:07:46
On Sun, Oct 06, 2002 at 09:38:32AM -0700, Erik Berls wrote:

> I've read through the nell driver, both it and the i386 based chips
> like to do this funky double mapping to cope with a small footprint
> in the (normally) ISA or PCI space.  (*_alloc, *_map functions)

No, that's not the reason. You have to do some acounting to assign
different addresses to all PCMCIA windows the pcmcia card driver
requests. You could blame this on our pcmcia(4) framework, but it more
or less comes from the pcmcia spec (and how CIS is layed out).

The virtual mapping of the nell hardware windows is completely static
(thanks to sbus and OF already pre-allocating all needed resources)

> Yes, there is the crux, because the nell _looks_ to the driver to
> be very similar to the i386 ones.

Actually it does not. It more looks like some amiga pcmcia device that
just lineary maps pcmcia space into physical address space.

> My big question tho, is when is the sparc bus_* rototillage going to be
> completed?  I heard that there was something left still to do?

I can't comment on any ongoing work there (I'm not aware of any for sparc).
The nell would need some enhancements (others would call that fixing of bugs)
in the sparc bus_space_read*/bus_space_write* macros (on sparc64 it's easily
solvable, just a SMOP which I hope to get to soon), but I'm not sure if this
is ever going to happen - and I won't do it myself.


Martin