I ended up just using the symbios driver, I don't know the history
behind the two but the symbios looked to be almost identical (though
it had the base_io member defined :). I think I had to change a few
other things as well, mincore was not defined anywhere and sys_madvise
was not defined either. Both of those are in the entry.S file and
ended up as undefined references at link time. I didn't have time to
look into what either of those are before leaving work. I did get it
compiled and running on an ES40 before leaving, though. This is the
first kernel I've used after the IRQ rewrite, but it seemed like
response time is a little slower when there is heavy disk activity.
Also, is there a reason that there is never an IRQ handled on CPU 3
(the last CPU)?
BAPper
On Thu, Mar 16, 2000 at 09:20:37PM -0800, Richard Henderson wrote:
> On Thu, Mar 16, 2000 at 03:21:12PM -0800, Brian Pomerantz wrote:
> > I was still unable to compile this. First I was bombing with an error
> > in include/asm/delay.h where it was trying to dereference a pointer of
> > unknown type in the cpu_data array.
>
> Bother. Here's a fix for it.
>
> > Then I came across the NCR driver which bombed. The macro INB_OFF(o)
> > is trying to reference the member 'base_io' in variable np which is of type
> > 'struct ncb *' and doesn't have this member in its definition.
>
> Yes, I saw this too. I worked around it, but it appears that
> something else is wrong because it doesn't pass cache tests.
>
> I was able to the get old 53c7,8xx driver working though...
>
