Il 15/05/2013 18:22, Markus Armbruster ha scritto:
> Paolo Bonzini <[email protected]> writes:
> 
>> Il 15/05/2013 17:09, Markus Armbruster ha scritto:
>>>>>>> @@ -209,7 +209,7 @@ void pc_system_firmware_init(MemoryRegion 
>>>>>>> *rom_memory)
>>>>>>>       * TODO This device exists only so that users can switch between
>>>>>>>       * use of flash and ROM for the BIOS.  The ability to switch was
>>>>>>>       * created because flash doesn't work with KVM.  Once it does, we
>>>>>>> -     * should drop this device for new machine types.
>>>>>>> +     * should drop this device.
>>>>>>>       */
>>>>>>>      sysfw_dev = (PcSysFwDevice*) qdev_create(NULL, "pc-sysfw");
>>>>>>>  
>>>>>
>>>>> Why did you change the comment?
>>>>
>>>> Because we agreed on the way forward for the flash patches, and it will
>>>> remove the need for (a) changes to machine types; (b) pc_sysfw in
>>>> general.  The device will be created iff a -pflash or -drive if=pflash
>>>> option is provided.  Thus in principle you could use -M pc-0.12 with
>>>> -pflash and it will work.
>>>
>>> Yes, that's the way forward, and yes, that means we'll have no use for
>>> the "pc-sysfw" dummy device on new machine types.  But why can we
>>> retroactively delete it from existing machine types?
>>
>> Because it would only affect TCG and people probably don't care much
>> about backwards-compatible machine types with TCG.  I'd rather remove
>> the misfeature completely and start from scratch with a sane design, now
>> that we have it.
> 
> I'm fine with limiting our backward compatibility promise to KVM, I just
> didn't expect it.

I think in general there's no reason to distinguish TCG from KVM, but in
this case we've been doing that all along so...

Paolo


Reply via email to