Am 05.02.2014 17:38, schrieb Paolo Bonzini:
> Il 05/02/2014 15:09, Andreas Färber ha scritto:
>> diff --git a/hw/core/qdev.c b/hw/core/qdev.c
>> index 82a9123..14c8765 100644
>> --- a/hw/core/qdev.c
>> +++ b/hw/core/qdev.c
>> @@ -131,21 +131,27 @@ DeviceState *qdev_create(BusState *bus, const
>> char *name)
>> DeviceState *qdev_try_create(BusState *bus, const char *type)
>> {
>> DeviceState *dev;
>> + ObjectClass *oc;
>> + DeviceClass *dc;
>>
>> - if (object_class_by_name(type) == NULL) {
>> + oc = object_class_by_name(type);
>> + if (oc == NULL) {
>> return NULL;
>> }
>> + dc = DEVICE_CLASS(oc);
>> dev = DEVICE(object_new(type));
>> if (!dev) {
>> return NULL;
>> }
>>
>> - if (!bus) {
>> + if (!bus && dc->bus_type && strcmp(dc->bus_type, "System") == 0) {
>
> Should you check instead if dev is-a TYPE_SYSBUS_DEVICE?
Hmm, that would catch TYPE_SYS_BUS_DEVICE-derived types as well, which
might in theory override bus_type.
The ugly thing here is the cyclic dependency betwen qdev and sysbus,
therefore no TYPE_SYSTEM_BUS.
> Does this also leave the nand device out of info qtree, or is it still
> dumped?
As I pointed out several times already, devices that are not on a bus
are generally *not* printed by info qtree. It depends on the old qbus
concept. Therefore info qtree is not of much use anymore and should be
dropped in favor of the qom-list and qom-get operations, which get us
all devices and all properties. I had once attached a qom-tree for that
and need to put that into a proper patch. Best before the next question
pops up. :)
Andreas
--
SUSE LINUX Products GmbH, Maxfeldstr. 5, 90409 Nürnberg, Germany
GF: Jeff Hawn, Jennifer Guild, Felix Imendörffer; HRB 16746 AG Nürnberg
