Re: [Qemu-devel] [Qemu-ppc] [PATCH v2 0/5] QOM'ify Power Architecture CPU

[Andreas Färber]

Am 16.04.2012 19:51, schrieb Alexander Graf:
> 
> On 10.04.2012, at 16:09, Andreas Färber wrote:
> 
>> Am 10.04.2012 08:41, schrieb David Gibson:
>>> On Fri, Apr 06, 2012 at 06:17:07PM +0200, Andreas Färber wrote:
>>>> Hello,
>>>>
>>>> This series follows up on my PowerPC QOM'ification patches from the
>>>> qom-cpu-others.v1 RFC series and splits it into steps easier to review.
>>>> The finalizer is actually filled with life now. Subclasses are postponed.
>>>>
>>>> David and Scott, please review and test.
>>>
>>> Is there documentation about the object model somewhere?  I don't yet
>>> know enough about it to easily review these.
>>
>> No conclusive one that I'm aware of... There's the feature description:
>> http://wiki.qemu.org/Features/QOM
>>
>> You can look at your converted qdev devices and at previously committed
>> CPU conversions as reference:
>>
>> target-sparc, first one using multiple type names:
>> http://git.qemu.org/?p=qemu.git;a=commit;h=ab7ab3d74c357e73a37b241fba27ea7f0595c085
>>
>> target-s390x:
>> http://git.qemu.org/?p=qemu.git;a=commit;h=29e4bcb26b80f975920508c83a9f24f29eb6bc1a
>> http://git.qemu.org/?p=qemu.git;a=commit;h=1ac1a7499bcb44174735780e0bd0421a1ac7a323
>> http://git.qemu.org/?p=qemu.git;a=commit;h=8f22e0df803697c11e8b10c90cc2e67df6e42884
>>
>> target-unicore32:
>> http://git.qemu.org/?p=qemu.git;a=commit;h=ae0f5e9ea80de923ae1c11289cf6ac468f657880
>>
>> target-arm:
>> http://git.qemu.org/?p=qemu.git;a=commit;h=dec9c2d4306d7b4f8ffff482ac42dc468ed2a61d
>>
>> Basically if you verify that I'm not doing something terribly stupid
>> like introducing NULL dereferences, removing valid parts of unrelated
>> code, etc. and if you check that your KVM test cases don't assert/crash
>> at startup, I'm confident we're fine in light of the "mass conversion".
>> Maybe consider a Tested-by sufficient for this guest-invisible change?
> 
> Yeah, tested-by should be ok here. What's the status on this one?

David added Acked-bys after all, and I sent out a v3 merging conflicting
changes of ours which David agreed to, so I merged the core
QOM'ification already with my PULL.

Still TODO are the parts not in v3, namely David didn't like the
finalizer concept, Anthony asked why there's a malloc() as opposed to
g_malloc() or g_try_malloc() in create_new_table(), and subclasses were
left out in this v2 already. Not to mention the potential further
subclassing (e500 etc.) you and Paul were discussing and QOM properties.

And if anyone wants to volunteer for cleaning up the #include
"translate_init.c" mess so that we can have a cpu.c and speed up the
compilation that would be great, too. ;)

Andreas

-- 
SUSE LINUX Products GmbH, Maxfeldstr. 5, 90409 Nürnberg, Germany
GF: Jeff Hawn, Jennifer Guild, Felix Imendörffer; HRB 16746 AG Nürnberg