FCode-utils 1.0.2 toke checksum issue

[Asif Haswarey]

Hi,

We have run into an issue with our FCodes tokenized using FCode-utils 1.0.2 toke,
on specific SPARC (Enterprise) platforms: M4000, M9000.

Power-up device probe yields the following warning message on the console:

WARNING: /pci@20,600000/...: FCode checksum error; expected: 280d result: 2837

After looking at the OBP sources, I found that the routine:
fcode-sum-ok? : obp/dev/pci/fcode-rom.fth

calculates the checksum by adding every FCode image (excluding header) byte and doing:

lwsplit + lwsplit + h# ffff and

on the sum. The
lwsplit + lwsplit +
apparently breaks the OpenFirmware spec.

OpenFirmware-1275 section 5.2.2.5 describes the checksum algorithm as:

Checksum is the doublet size sum of the bytes of the program body (i.e.,
excluding the header), calculated using two's complement addition and
ignoring overflow.

FCode-utils 1.0.2 toke calculates the checksum in function:

finish_fcodehdr : toke/emit.c

as described in the spec.

How do we deal with this?
Any suggestions?

Thanks for any response(s)!

-Asif

[Stefan Reinauer]

On 1/28/10 4:37 AM, Asif Haswarey wrote:
> Hi,
>
> We have run into an issue with our FCodes tokenized using FCode-utils 1.0.2 toke,
> on specific SPARC (Enterprise) platforms: M4000, M9000.
>
> Power-up device probe yields the following warning message on the console:
>
> WARNING: /pci@20,600000/...: FCode checksum error; expected: 280d result: 2837

Dear Asif,

could you please try the attached patch and see if the error goes away?

Best regards,

Stefan