Re: interesting things with PC164 and latest milo-980602

Alexander V. Lukyanov (lav@yars.free.net)
Sat, 5 Sep 1998 23:48:27 +0400


On Sat, Sep 05, 1998 at 11:57:37AM -0700, Richard Henderson wrote:
> On Sat, Sep 05, 1998 at 02:46:00PM +0400, Alexander V. Lukyanov wrote:
> > But the kernel compiled with it hung just after "jumping to kernel".
> > Everything returned to norm when I recompiled the kernel with -mcpu=ev5,
> > so I assume the problem was BWX on IO space.
>
> I very much doubt that. Look at how I/O is implemented in
> include/asm-alpha/[core_]cia.h and tell me how that would
> change based on BWX insns.

Oops, you are right. I was thinking about DENSE io, and linux uses SPARSE io.

And I tried to boot again that kernel compiled for rv56. Looks like I
was mistakenly booting with older milo, because with the new milo the
kernel starts to boot (and actually prints some messages) and then
hangs. So my theory is wrong. But it is still true that after rebuilding
compiler, kernel and some apps going to older milos is impossible on pc164.

The hang happens just after the line
SMC FDC37C93X Ultra I/O Controller found @ 0x370

Suspiciously looks like a hang I experienced a while ago with older kernels.
At that time the hang magically disappeared when I inserted a delay in a
function for the smc fdc init. Have not tried that now yet.

Alexander.