[PATCH] usb: gadget: Compile USB ethernet gadget only if NET is enabled

[Marek Vasut]

On 5/1/23 23:18, Tom Rini wrote:
> On Mon, May 01, 2023 at 10:49:37PM +0200, Marek Vasut wrote:
>> On 5/1/23 20:53, Tom Rini wrote:
>>> On Mon, May 01, 2023 at 07:40:57PM +0200, Marek Vasut wrote:
>>>> On 5/1/23 19:23, Tom Rini wrote:
>>>>> On Mon, May 01, 2023 at 06:53:52PM +0200, Marek Vasut wrote:
>>>>>> On 5/1/23 15:47, Tom Rini wrote:
>>>>>>> On Sun, Apr 30, 2023 at 11:20:35PM +0200, Marek Vasut wrote:
>>>>>>>
>>>>>>>> In case NET networking is not enabled, it is not possible to compile
>>>>>>>> the USB ethernet gadget. Protect the symbols in Makefile to avoid build
>>>>>>>> failure. Such build failure may occur e.g. in case NET and USB ethernet
>>>>>>>> gadget is enabled in U-Boot proper, but not in SPL.
>>>>>>>>
>>>>>>>> Signed-off-by: Marek Vasut <marex at denx.de>
>>>>>>>> ---
>>>>>>>> Cc: Lukasz Majewski <lukma at denx.de>
>>>>>>>> ---
>>>>>>>>      drivers/usb/gadget/Makefile | 2 ++
>>>>>>>>      1 file changed, 2 insertions(+)
>>>>>>>>
>>>>>>>> diff --git a/drivers/usb/gadget/Makefile b/drivers/usb/gadget/Makefile
>>>>>>>> index 6cfe0f3a041..36f65e7eb95 100644
>>>>>>>> --- a/drivers/usb/gadget/Makefile
>>>>>>>> +++ b/drivers/usb/gadget/Makefile
>>>>>>>> @@ -34,8 +34,10 @@ endif
>>>>>>>>      obj-$(CONFIG_CI_UDC) += ci_udc.o
>>>>>>>> +ifeq ($(CONFIG_$(SPL_TPL_)NET),y)
>>>>>>>>      obj-$(CONFIG_USB_ETHER) += ether.o
>>>>>>>>      obj-$(CONFIG_USB_ETH_RNDIS) += rndis.o
>>>>>>>> +endif
>>>>>>>>      # Devices not related to the new gadget layer depend on CONFIG_USB_DEVICE
>>>>>>>>      # This is really only N900 and USBTTY now.
>>>>>>>
>>>>>>> Why can't we just enforce this via Kconfig?
>>>>>>
>>>>>> Because there is no SPL/TPL USB_ETHER Kconfig .
>>>>>> Do we want to grow the Kconfig file with those instead ?
>>>>>
>>>>> Ah right.  Yes, we have SPL_USB_ETHER today
>>>>
>>>> This is exactly the opposite of what I wrote.
>>>>
>>>> And no, we do NOT have this symbol, see:
>>>>
>>>> $ git grep SPL_USB_ETHER drivers/usb | wc -l
>>>> 0
>>>
>>> Yes, it resides in common/spl/Kconfig
>>
>> Uhhhhh, such USB symbol is not supposed to be there in the first place.
> 
> There's long running debate on where some symbols should be for a better
> overall experience of enabling features.

Putting my USB maintainer hat on, USB drivers related symbols should be 
in drivers/usb/ .

>> However, looking at the meaning of that symbol, it seems it governs
>> something else -- USB ethernet device(s) (like the USB-ethernet adapter
>> which you plug into USB host port). So, the SPL_USB_ETHER symbol name is
>> incorrectly named too and should be renamed to something else first I think
>> ?
>>
>> Do I read it right ?
>>
>> And if so, then, back to my original reply -- there is no SPL_USB_ETHER
>> symbol. Does it make more sense to really add one (not misnamed one) or not
>> ?
> 
> We should re-work things as needed so that we have more Kconfig symbols,
> as needed.  There's some overlap / overloading of things today as
> platforms have been able to (and I think am335x_evm is still one that
> builds for) supporting making a USB RNDIS gadget device happen in SPL,
> so that we can then be fed the next stage of U-Boot.

I have a hardware where I only want USB ethernet support in U-Boot, not 
in SPL, hence this patch.

But the symbol which is currently named SPL_USB_ETHER as defined in 
common/ does NOT ungate SPL CDC ethernet support, it ungates SPL USB 
host ethernet (like those ASIX USB-ethernet adapters), do you agree ?

If yes, then the symbol itself should be renamed .