Diagnostics support

[Robert Reif]

How much diagnostics support is already in OpenBIOS and what would it 
take to add something like sun's OBDiag.

Has anyone ever tried running a sparc32 OpenBIOS image on real hardware?

[Blue Swirl]

On 3/19/09, Robert Reif  wrote:
> How much diagnostics support is already in OpenBIOS and what would it take
> to add something like sun's OBDiag.

Very little to none. If you mean something like this:

http://tracker.coreboot.org/trac/openboot/browser/obp/os/bootprom/diagmode.f...

it would mean touching a lot of places.
> Has anyone ever tried running a sparc32 OpenBIOS image on real hardware?

This has been on my TODO for a long time. Standard image will fail, it
probes for QEMU. But it should be possible to create a special entry.S
for real hardware. There are some other places where we just use the
Qemu facilities directly without probing for the real alternative,
these should be fixed.

[Robert Reif]

Blue Swirl wrote:
> On 3/19/09, Robert Reif  wrote:
> > How much diagnostics support is already in OpenBIOS and what would it take
> > to add something like sun's OBDiag.
>
> Very little to none. If you mean something like this:
>
> http://tracker.coreboot.org/trac/openboot/browser/obp/os/bootprom/diagmode.f...
>
> it would mean touching a lot of places.

A simple hook for POST would be a good start.  I just want a way of running
some code for testing the hardware (or hardware emulation).
> > Has anyone ever tried running a sparc32 OpenBIOS image on real hardware?
>
> This has been on my TODO for a long time. Standard image will fail, it
> probes for QEMU. But it should be possible to create a special entry.S
> for real hardware. There are some other places where we just use the
> Qemu facilities directly without probing for the real alternative,
> these should be fixed.

QEMU and OpenBIOS and linux are too interdependent.  QEMU doesn't
come close to running a real OpenBoot image because the hardware emulation
isn't complete.  Granted, some stuff will never be emulated properly but 
there
is room for improvement.  Running OpenBIOS on real hardware validates
it much better than on QEMU because the quality of emulation factor is
removed.  QEMU/OpenBIOS is not capable of booting any version of solaris
and some versions of bsd due to deficiencies in both.

I'm just trying to tackle these issues from any angle that improves QEMU
and OpenBIOS by reducing the dependencies and number of variables by
using known working code or hardware.