>This is a bit unfortunate? no?
>
>Whats the root cause of this? the 21285's PCI implementation? or the SA?
>
>Doesn't this mean that most (if not all) IO/DMA device drivers need
>patching to run on ARMLinux?
The cause is simply the way the StrongARM and the 21285 are designed. To
provide a fully coherent cache without involvement from software, logic must
be provided in the chipset to snoop cycles on the bus and take appropriate
action to synchronise caches when necessary. Some processor architects felt
this was a worthwhile feature to implement in silicon; others didn't. The ARM
isn't by any means the only processor to be this way. Some MIPS systems have
the same behaviour; I think some 68000s do as well and I'm sure there are
others.
The only drivers that are affected are those performing DMA to main memory.
Yes, they need patching to work with Linux/ARM (or similar systems) but the
changes are usually trivial.
Programmed I/O isn't affected.
p.
unsubscribe: body of `unsubscribe linux-arm' to [EMAIL PROTECTED]
