Re: [PATCH] x86/boot: Fix BIOS memory corruption on certain IBM systems

[Keir Fraser <keir.xen@gmail.com>]

On 04/12/2013 10:47, "Jan Beulich" <JBeulich@suse.com> wrote:

>>>>           * Declare that our target operating mode is long mode.
>>>>           * Initialise 32-bit registers since some buggy BIOSes depend on
>>>> it.
>>>>           */
>>>> +        xor     %ecx,%ecx
>>>> +        xor     %edx,%edx
>>>> +        xor     %esi,%esi
>>>> +        xor     %edi,%edi
>>>> +        xor     %ebp,%ebp
>>>>          movl    $0xec00,%eax      # declare target operating mode
>>>>          movl    $0x0002,%ebx      # long mode
>>>>          int     $0x15
>>> ... I can't really see the value of the change here: If we're to
>>> work around theoretical BIOS bugs, we'd need to do this prior to
>>> each BIOS call. That's surely overkill. Therefore let's focus on
>>> what is needed to work around _known_ BIOS bugs.
>> 
>> I admit that I was leaning on the cautious side with these changes.
>> 
>> I can take them out if you think that would be better, but given this
>> int was already flagged as buggy in some BIOSes, and we have found
>> another case, I think covering all GPRs is the safer option.
> 
> As said - I doubt this would help much. I'd really prefer at least this
> part of the patch to be taken out again. Unless Keir is specifically of
> the opposite opinion...

The change does kind of fit with the comment immediately above though.
Overall I don't really care that much either way.

 -- Keir