On Mon, Apr 12, 2004 at 11:49:20PM +0900, Izumi Tsutsui wrote:

> Umm, the latter is not streamline but just backout...

Er, isn't arch/mips/mips/* supposed to be canonical for all mips targets?
Looking at the code involved, I can't see any good reason to prefer the
contortions in the previous sgimips-specific routine ... hence the term
"streamline".

> (though I wouldn't object if the previous code causes any problem)

Well, I'm trying to track down weird problems when one uses a CD-ROM on IP32.
My working hypothesis was that the bus_space DMA code was a bit chigau, so
I was doing a line-by-line diff between the MD and MI code.  The changes as
committed unfortunately have not fixed the core problem, but one must fix
bugs as they are encountered ...

(btw, there are several instances in the MD code where we're declaring
variables as vaddr where in the MI equivalent they are paddr, which indicates
a bit of confusion as to what that code was doing on a given platform although
the typedefs involved result in the same type)

I'm now thinking that it's best to deprecate sgimips/bus.c in favor of
mips/bus_dma.c, with function overrides where necessary.

-- Chris
	GPG key FEB9DE7F (91AF 4534 4529 4BCC 31A5  938E 023E EEFB FEB9 DE7F)