Daniel P. Berrangé <[email protected]> writes:
> On Wed, Jul 08, 2026 at 11:16:30AM +0100, Peter Maydell wrote:
>> On Fri, 3 Jul 2026 at 10:09, Mark Cave-Ayland
>> <[email protected]> wrote:
>> >
>> > Signed-off-by: Mark Cave-Ayland <[email protected]>
>> > ---
>> > include/hw/i386/pc.h | 2 ++
>> > hw/i386/pc.c | 8 ++++++++
>> > hw/i386/pc_sysfw.c | 7 +------
>> > 3 files changed, 11 insertions(+), 6 deletions(-)
>> >
>> > diff --git a/include/hw/i386/pc.h b/include/hw/i386/pc.h
>> > index 309de2eda1..16cb0ec01f 100644
>> > --- a/include/hw/i386/pc.h
>> > +++ b/include/hw/i386/pc.h
>> > @@ -40,6 +40,8 @@ typedef struct PCMachineState {
>> >
>> > Object *alias_pcspk;
>> > Object *alias_rtc_time;
>> > + Object *alias_pflash0;
>> > + Object *alias_pflash1;
>> >
>> > /* Configuration options: */
>> > uint64_t max_ram_below_4g;
>> > diff --git a/hw/i386/pc.c b/hw/i386/pc.c
>> > index 13c9c1bfc8..11eef176f4 100644
>> > --- a/hw/i386/pc.c
>> > +++ b/hw/i386/pc.c
>> > @@ -1762,6 +1762,14 @@ static void pc_machine_class_init(ObjectClass *oc,
>> > const void *data)
>> > offsetof(X86MachineState, acpi_dev),
>> > object_property_allow_set_link,
>> > OBJ_PROP_LINK_STRONG);
>> > + object_class_property_add_alias(oc, "pflash0",
>> > + offsetof(PCMachineState,
>> > alias_pflash0),
>> > + TYPE_PFLASH_CFI01,
>> > + "drive");
>> > + object_class_property_add_alias(oc, "pflash1",
>> > + offsetof(PCMachineState,
>> > alias_pflash1),
>> > + TYPE_PFLASH_CFI01,
>> > + "drive");
>>
>> Does making these class properties fix
>> https://gitlab.com/qemu-project/qemu/-/work_items/3254
>> (which is a report that "qemu-system-x86_64 -machine pc-q35,help"
>> doesn't list these properties) ?
Not showing object properties is a bug.
Cannot be fixed for object properties that are created outside the
object's instance_init().
Now, object properties created in instance_init() should almost
certainly be class properties instead. Converting them all fixes the
fixable part of the bug:
> Yes, it ought to, as that CLI logic iterates over class properties.
One property conversion at a time.
There's a quicker fix, though.
-machine TYPE.help uses type_print_class_properties() to show
properties. Here's its loop:
object_class_property_iter_init(&iter, klass);
while ((prop = object_property_iter_next(&iter))) {
if (!prop->set) {
continue;
}
g_ptr_array_add(array,
object_property_help(prop->name, prop->type,
prop->defval, prop->description));
}
-device TYPE,help uses qmp_device_list_properties() to show both.
Here's its loop:
obj = object_new_with_class(klass);
object_property_iter_init(&iter, obj);
while ((prop = object_property_iter_next(&iter))) {
ObjectPropertyInfo *info;
[...]
info = g_new0(ObjectPropertyInfo, 1);
info->name = g_strdup(prop->name);
info->type = g_strdup(prop->type);
info->description = g_strdup(prop->description);
info->default_value = qobject_ref(prop->defval);
QAPI_LIST_PREPEND(prop_list, info);
}
object_unref(obj);
Is there any excuse not to do it this way?
