Re: [PATCH 50/55] s390x/event-facility: Simplify creation of SCLP event devices

[Paolo Bonzini]

On 25/05/20 09:01, Markus Armbruster wrote:
>> Just remembered that we fail creating the machine and therefore abort. So 
>> not necessary :)
> True.
> 
> But let's review briefly what happens when a realize method fails.
> 
> In theory, realize fails cleanly, i.e. doing nothing.  Another attempt
> could be made then.
> 
> In practice, realize failure is always followed by destruction, unless
> preempted by outright exit(1).
> 
> Destroying a device must also destroy its components.
> 
> Paolo, is destroying a realized device okay, or does it have to be
> unrealized first?  I can't see automatic unrealize on destruction...

It cannot happen, because a device must be unparented before it's
destroyed and unparenting takes care of unrealizing the device.  So the
stageobject lifetime should always proceed in this order:

   created
   created, with parent
   created, with parent, with bus (if applicable)
   realizing
   realized
   unrealizing
   unrealized
   unrealized, without parent, with bus (if applicable)
   unrealized, without parent, without bus
   finalizing (without references)
   finalized
   freed

Where the second and third would be fixed by moving /machine/unattached
from device_set_realized to qdev_realize.

Paolo