On Mon, Jun 13, 2005 at 09:40:20PM +0100, Maciej W. Rozycki wrote:
> You may argue it's best to define a private copy of "cpu_has_llsc"
> expanding to a constant for selecting the right set of atomic operations
> at the compilation time and I would agree, but AFAIK the whole idea behind
> our current implementation is to provide a snail-speed fallback or perhaps
> to support more generic configurations at one point (e.g. one kernel for
> all DECstations).
> For most processor settings the current setup already works, as they
> provide ll/sc anyway, but not for MIPS I ones, like the R3k. Here's a
> patch that makes the affected code work for such processors as well.
> OK to apply?
Al had a bunch of other complaints which I'll try to take care of asap.