Subject: Re: structural bus_space_handles
To: Garrett D'Amore <>
From: Sam Leffler <>
List: tech-kern
Date: 06/05/2006 09:25:39
Garrett D'Amore wrote:
> Sam Leffler wrote:
>> Garrett D'Amore wrote:
>>> He passes a void * token (two actually, one for the tag and one for the
>>> handle) now.
>>> The point was that cleaning this up required all architectures to get an
>>> extra indirect access, whereas only sparc truly needs it.
>> The extra indirect was #ifdef'd by the arch.  It's only done for sparc64
>> where it's required.  Everywhere else there is no extra indirection; the
>> data are passed through as opaque values and cast/converted before use.
> Yes, I know that.  I had briefly put it back, but I've since changed
> it.   What I committed now stores the handle/tag in the softc, and pass
> a pointer to these to the HAL.  It works on SPARC, i386, and the
> AR5312.  I have not been able to make the HAL work on Au1550 (MIPS
> based) for some time, but at this point I don't know whether the problem
> is in the HAL or in the Au1550 port.  I'd love to have someone else with
> alternate MIPS hardware test this out for me.

Sounds like you've tied the definition of the softc structure to the hal
which is a step backward.