Re: [RFC PATCH v2 19/19] hw/usb: Inline usb_bus_from_device()

["Philippe Mathieu-Daudé" <philmd@linaro.org>]

On 13/2/23 09:11, Thomas Huth wrote:
> On 13/02/2023 08.08, Philippe Mathieu-Daudé wrote:
>> Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
>> ---
>> RFC Other devices don't use such helper. Maybe it should
>>      be the other way around, introduce more bus_from_device()
>>      helpers?
>> ---
>>   hw/usb/bus.c        | 10 +++++-----
>>   hw/usb/core.c       |  6 +++---
>>   hw/usb/dev-hub.c    |  4 ++--
>>   hw/usb/dev-serial.c | 10 +++++-----
>>   hw/usb/hcd-xhci.c   |  2 +-
>>   include/hw/usb.h    |  5 -----
>>   6 files changed, 16 insertions(+), 21 deletions(-)
>>
>> diff --git a/hw/usb/bus.c b/hw/usb/bus.c
>> index d7c3c71435..4a1b67761c 100644
>> --- a/hw/usb/bus.c
>> +++ b/hw/usb/bus.c
>> @@ -427,7 +427,7 @@ void usb_unregister_port(USBBus *bus, USBPort *port)
>>   void usb_claim_port(USBDevice *dev, Error **errp)
>>   {
>> -    USBBus *bus = usb_bus_from_device(dev);
>> +    USBBus *bus = USB_BUS(qdev_get_parent_bus(DEVICE(dev)));
> 
> You're certainly missing a proper justification in the patch description 
> here. The "other devices don't use such a helper" does not sound like a 
> real justification to me, since the code lines rather get longer this 
> way. Thus this rather looks like unnecessary code churn to me --> rather 
> drop the patch?

The idea is to avoid having 7 different ways of implementing something
with 3 different APIs and 2 unfinished API conversions in flight.

I'm wondering if the QOM DECLARE_xxx() macros could also define some
xxx_BUS_FROM_DEV() or xxx_PARENT_BUS() macros. So here it would become:

     USBBus *bus = USB_PARENT_BUS(dev);