Re: [Qemu-devel] [RFC PATCH 02/15] qdev: store DeviceState's canonical path to use when unparenting

[Michael Roth <mdroth@linux.vnet.ibm.com>]

Quoting Paolo Bonzini (2015-05-04 04:35:11)
> 
> 
> On 02/05/2015 00:54, Michael Roth wrote:
> >> > 
> >> > What about unparenting children devices in the device's unrealize
> >> > callback?  It sucks that you have to do it manually, but using stale
> >> > canonical paths isn't the nicest thing either.
> > That does seems to do the trick. It felt wrong when I first looked at
> > it because in some cases the children attach themselves to the parent
> > without making making the parent aware,
> 
> Can you point to an example?  The parent "owns" its property namespace,
> so it would be wrong for a child to attach itself unbeknownst to the parent.

Well, I was referring to:

  sPAPRTCETable *spapr_tce_new_table(DeviceState *owner, uint32_t liobn)

and:

  sPAPRDRConnector *spapr_dr_connector_new(Object *owner,
                                           sPAPRDRConnectorType type,
                                           uint32_t id)

It wasn't immediately obvious to me that this would result in the
resulting objects attaching themselves as children, but in retrospect
that does seem to be implied by the 'owner' parameter.

If there are cases where this is done with a parameter that isn't explicitly
named 'owner' though it might be somewhat ambiguous (could be pulling out
individual fields as opposed to attaching itself), but I don't see any
examples outside of local functions or qdev-managed devices.

> 
> There are a couple cases where an object adds a property to another
> object, e.g. the rtc-time property of /machine, but in that case the
> property is a well-known name whose setting is "outsourced" by /machine
> to the device.
> 
> Paolo
>