Re: Bus space handle type for mips

[David Holland <dholland-mips%netbsd.org@localhost>]

On Sun, Jun 06, 2010 at 11:17:38PM +0200, Martin Husemann wrote:
 > This patch makes bus_space_handle_t unsigned again (as it was before the 
 > partial mips64 merges where it changed from u_long to intptr_t).
 > 
 > This unbreaks PCI on alchemy boards for me. I don't see why the handle
 > should be a signed value - am I missing something?

To some extent mips addresses are signed; in particular I believe the
mapping of the 32-bit address space into the 64-bit address space is
signed, in the sense of using sign extension for addresses 0x80000000
to 0xffffffff. (Someone who's actually touched mips64 can probably
confirm or refute this instantly.)

So it may make sense; however, for it to actually work someone's going
to have to check through all uses for sign issues...

-- 
David A. Holland
dholland%netbsd.org@localhost