Re: [PATCH 2/2] qom: Manage references to embedded child objects

On 15/06/2026 09:12, Daniel P. Berrangé wrote:

> On Mon, Jun 15, 2026 at 01:11:07PM +0900, Akihiko Odaki wrote:
>> Problem
>> =======
>>
>> The Rust wrapper for object_ref() is marked unsafe and has:
>>> The object must not be embedded in another unless the outer
>>> object is guaranteed to have a longer lifetime.
>>
>> In other words, object_ref() does not work for embedded objects and
>> does not keep embedded objects alive. MemoryRegion has its own
>> memory_region_ref() helper to call object_ref() on its owner for this
>> reason.
>>
>> However, this is insufficient to avoid calling object_ref() for all
>> embedded objects. For example, consider an embedded Device that has a
>> MemoryRegion. When referencing a MemoryRegion for guest memory access,
>> QEMU automatically references the owning Device to keep the MemoryRegion
>> alive. However, that reference is ineffective if the Device itself is
>> embedded, because object_ref() does not keep the containing storage
>> alive.
> 
> Do we know how many examples we have of embedding objects inside
> another ?
> 
> I would much prefer if we forbid the embedding of objects. It is
> horrible design practice to have some QOM objects which can be
> freed via reference count and some which cannot.

This is something I've been discussing with a few other people for some 
time now: shouldn't we deprecate object_initialize_child() and use 
references for everything instead? Not only would this help solve a lot 
of lifecycle problems, but it would help reviewers because at the moment 
we have too many different ways of doing the same thing.


ATB,

Mark.