2.3.36-pre on SPARC

[Dominik Kubla]

On Wed, Dec 22, 1999 at 07:28:47PM -0800, David S. Miller wrote:
...
> It is hoped that sparc32 can be in working order by mid January
> but no promises.

Hi Dave,

from looking at the CVS tree and the latest pre-patch it looks as if
the two are back in sync WRT the SPARC architecture. Am i correct?

Now while trying to compile the latest pre-patch i got a failure in
ide-geometry.c (Why is that compiled unconditionally into the block device
driver code in the first place i wonder, after all not all systems have
IDE interfaces/drives...).  I got over that by simply copying the sparc64
version of hdregs.h and ide.h over to sparc and change were the compiler
bailed out.  Now here are my questions:

1. Does ide_ioreg_t need to be a long on sparc as well or will a short do
   as on the other archs?

2. In ide.h the functions inw_be() and outw_be() set ASI_PHYS_BYPASS_EC_E,
   which according to asi.h is "PADDR, E-cachable, E-bit". Now the v8 MMU
   apparently does not have this condition instead i used ASI_M_BYPASS on
   a wild guess.  Is that correct?

3. In ide.h the functions ide_insw() and ide_outsw() both finish with
   "__flush_dcache_range()". That function does not exist on sparc. Is that
   not possible on older SPARC chips or simply not (yet) implemented?

Simply omitting "__flush_dcache_range()" and applying step 1 and 2 gives me
a clean compile up to srmmu.c...

Yours,
  Dominik Kubla