[coreboot] [patch][v2] AMD Fam10 rev B3 microcode patches

[Carl-Daniel Hailfinger]

On 22.07.2008 19:56, Marc Jones wrote:
> Carl-Daniel Hailfinger wrote:
> ...
>>> + (c) Advanced Micro Devices, Inc., 2004-2008
>>> +
>>> + The  enclosed microcode  is intended  to be  used  with AMD
>>> + Microprocessors.  You  may   copy,  view  and  install  the
>>> + enclosed microcode  only for development  and deployment of
>>> + firmware,  BIOS,  or  operating  system code  for  computer
>>> + systems   that  contain  AMD   processors.   You   are  not
>>> + authorized  to use  the  enclosed microcode  for any  other
>>> + purpose.
>>>   
>>
>> I trust that AMD is not going to hunt us down if we check out the
>> complete svn tree (copy) with the intent to develop for non-AMD systems.
>> Still, the legalese feels a bit weird.
>>
>>
>
> You bring this up every time I update microcode patches. This has been
> worked out in the past. The statement seems very straight forward to
> me (but I am not a lawyer).

I wish to apologize. I can't seem to remember having made a similar
statement in the past, but I'm not going to object the inclusion of this
code. Go ahead.


>>> +/* From the Revision Guide :
>>> + * Equivalent Processor Table for AMD Family 10h Processors
>>> + *
>>> + * Installed Processor   Equivalent Processor   Patch Level
>>> + * Revision ID           Revision ID
>>> + * 00100F00h             1000h                  01000020h
>>> + * 00100F01h             1000h                  01000020h
>>> + * 00100F02h             1000h                  01000020h
>>> + * 00100F20h             1020h                  01000084h
>>> + * 00100F21h             1020h                  01000084h
>>> + * 00100F2Ah             1020h                  01000084h
>>> + * 00100F22h             1022h                  01000083h
>>> + * 00100F23h             1022h                  01000083h
>>>   
>>
>> AFAICS it could happen that different "Equivalent Processor IDs" have
>> the same patch level. Naming the microcode files only after the patch
>> level would cause all sorts of interesting conflicts in that case.
>> How about a naming scheme like
>> mc_patch_$EQUIVALENTREVISION_$PATCHLEVEL.h
>
> I don't see your point. The code already handles equivalent processor
> ids. I put the table in so you didn't have to read the revision guide
> to understand what is going on. What you suggest only makes it more
> difficult if different equivalent ids have the same patch level.

Hm. Maybe I misunderstand the current scheme, but I think it breaks
exactly for the case you cite: "different equivalent ids have the same
patch level".


>     #include "mc_patch_01000065.h"
>>> +    /* Barcelona rev B2, B3 */
>>> +        #include "mc_patch_01000083.h"
>>>   
>>
>> This looks like manual source code editing is required to support
>> Barcelona processors before B2. May I suggest a Kconfig variable for
>> that?
>>
>
> Yes, it is manual editing so I will add a config option to v2. A
> Kconfig variable would be great when you port it to v3. :)

Heh. I live in the v3 world and forgot about the nonexistence of Kconfig
in v2. I'll take care of the conversion to Kconfig for a v3 port.

Does this changeset also include a removal of old microcode?

Regards,
Carl-Daniel

-- 
http://www.hailfinger.org/