Thanks. The patch probably fixes it on x86, but it's still the wrong thing to
do (you really _have_ to use readb/writeb to be portable, because accessing
IO memory is not just a matter of getting the right address, on some
architectures you have to do special IO cycles or coherency stuff instead of
just a pointer dereference).
> And: pre-2.1.0 survived 9 hours crashme.
Ok, good to know that I didn't do anything overly stupid with the system call
entry point optimization etc.. Btw, has anybody tested whether "dosemu" still
works in pre-2.1.0?
Anyway, wrt the aha152x driver: could you try the _newest_ pre-patch
(currently #4) to see if the aha152x driver works with my version of your fix
too? I've written the code to do it portably, but I obviously cannot test if
it actually works..
Linus