Re: [Qemu-devel] [PATCH 4/4] vl: add -late-object to create QOM objects after machine init

Re: [Qemu-devel] [PATCH 4/4] vl: add -late-object to create QOM objects after machine init

[Paolo Bonzini]

Il 26/06/2012 00:43, Anthony Liguori ha scritto:
> In order to create qdev objects via -late-object, we almost always have to
> specify the parent_bus which is usually created during machine init.  Until we
> properly support two stage init, introduce a -late-object option that allows for
> creation of objects post-machine init.

This is only needed for -device, no?  So we can avoid introducing it and
just use -device.

Paolo

Re: [Qemu-devel] [PATCH 4/4] vl: add -late-object to create QOM objects after machine init

[Andreas Färber]

Am 02.07.2012 14:32, schrieb Paolo Bonzini:
> Il 26/06/2012 00:43, Anthony Liguori ha scritto:
>> In order to create qdev objects via -late-object, we almost always have to
>> specify the parent_bus which is usually created during machine init.  Until we
>> properly support two stage init, introduce a -late-object option that allows for
>> creation of objects post-machine init.
> 
> This is only needed for -device, no?  So we can avoid introducing it and
> just use -device.

+1, was thinking along the same lines...

Andreas

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

Re: [Qemu-devel] [PATCH 4/4] vl: add -late-object to create QOM objects after machine init

[Anthony Liguori]

Andreas Färber <afaerber@suse.de> writes:

> Am 02.07.2012 14:32, schrieb Paolo Bonzini:
>> Il 26/06/2012 00:43, Anthony Liguori ha scritto:
>>> In order to create qdev objects via -late-object, we almost always have to
>>> specify the parent_bus which is usually created during machine init.  Until we
>>> properly support two stage init, introduce a -late-object option that allows for
>>> creation of objects post-machine init.
>> 
>> This is only needed for -device, no?  So we can avoid introducing it and
>> just use -device.
>
> +1, was thinking along the same lines...

Sorry for such a delayed response...

The semantics of how -device is different in that it implies realize.

Regards,

Anthony Liguori

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

Re: [Qemu-devel] [PATCH 4/4] vl: add -late-object to create QOM objects after machine init

[Paolo Bonzini]

Il 19/09/2012 01:07, Anthony Liguori ha scritto:
>>>> In order to create qdev objects via -late-object, we almost always have to
>>>> >>> specify the parent_bus which is usually created during machine init.  Until we
>>>> >>> properly support two stage init, introduce a -late-object option that allows for
>>>> >>> creation of objects post-machine init.
>>> >> 
>>> >> This is only needed for -device, no?  So we can avoid introducing it and
>>> >> just use -device.
>> >
>> > +1, was thinking along the same lines...
> Sorry for such a delayed response...
> 
> The semantics of how -device is different in that it implies realize.

Yes, all I'm saying is, I don't see a need for -late-object since it is
only needed for devices and we have a separate method to create them.

Paolo

Re: [Qemu-devel] [PATCH 2/4] qdev: add realized property and make adding child bus implied by realize

[Paolo Bonzini]

Why does this lack recursive realization?

(I'm not top posting, I'm replying to the commit message. :))

Paolo

Il 26/06/2012 00:43, Anthony Liguori ha scritto:
> Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
> ---
>  hw/qdev.c |   36 +++++++++++++++++++++++++++++++++++-
>  1 files changed, 35 insertions(+), 1 deletions(-)
> 
> diff --git a/hw/qdev.c b/hw/qdev.c
> index a6c4c02..d305128 100644
> --- a/hw/qdev.c
> +++ b/hw/qdev.c
> @@ -101,7 +101,6 @@ static void bus_add_child(BusState *bus, DeviceState *child)
>  void qdev_set_parent_bus(DeviceState *dev, BusState *bus)
>  {
>      dev->parent_bus = bus;
> -    bus_add_child(bus, dev);
>  }
>  
>  /* Create a new device.  This only initializes the device state structure
> @@ -157,6 +156,10 @@ int qdev_init(DeviceState *dev)
>  
>      assert(dev->state == DEV_STATE_CREATED);
>  
> +    if (dev->parent_bus) {
> +        bus_add_child(dev->parent_bus, dev);
> +    }
> +
>      rc = dc->init(dev);
>      if (rc < 0) {
>          object_unparent(OBJECT(dev));
> @@ -663,6 +666,33 @@ void qdev_property_add_static(DeviceState *dev, Property *prop,
>      assert_no_error(local_err);
>  }
>  
> +static bool qdev_prop_get_realized(Object *obj, Error **errp)
> +{
> +    DeviceState *dev = DEVICE(obj);
> +
> +    return (dev->state == DEV_STATE_INITIALIZED);
> +}
> +
> +static void qdev_prop_set_realized(Object *obj, bool value, Error **errp)
> +{
> +    DeviceState *dev = DEVICE(obj);
> +    bool realized = (dev->state == DEV_STATE_INITIALIZED);
> +
> +    if (realized == value) {
> +        return;
> +    }
> +
> +    if (realized && !value) {
> +        error_set(errp, QERR_PERMISSION_DENIED);
> +        return;
> +    }
> +
> +    if (qdev_init(dev) < 0) {
> +        error_set(errp, QERR_DEVICE_INIT_FAILED, "");
> +        return;
> +    }
> +}
> +
>  static void device_initfn(Object *obj)
>  {
>      DeviceState *dev = DEVICE(obj);
> @@ -687,6 +717,10 @@ static void device_initfn(Object *obj)
>      } while (class != object_class_by_name(TYPE_DEVICE));
>      qdev_prop_set_globals(dev);
>  
> +    object_property_add_bool(obj, "realized",
> +                             qdev_prop_get_realized, qdev_prop_set_realized,
> +                             NULL);
> +
>      object_property_add_link(OBJECT(dev), "parent_bus", TYPE_BUS,
>                               (Object **)&dev->parent_bus, NULL);
>  }
>