On Sat, 17 Jun 2000, matthew green wrote:

> but, even so, NetBSD is designed such that any MI driver can (should,
> and is in many cases) be endian-independant... we do the smarts in
> the backends.

Not quite.  Register access is covered quite well but DMA is still
roll-your-own.  (Jason how are the bus_dma_*() accessors coming
along?  Need help?)  

Anyway, PCI DMA seems to be working pretty well in 64-bit SPARC although
32-bit seems to have some problems still.  Have any other architectures
used PCI HME successfully?  I wouln't be surprised it some htole*() or
htobe*() macros may be required when accessing DMA memory areas with PCI
HME.

Eduardo Horvath