Re: [PATCH 1/2] usb: gadget: restore tristate-choice for legacy gadgets

[Arnd Bergmann]

On Mon, Dec 11, 2017 at 5:09 PM, Bart Van Assche <Bart.VanAssche@xxxxxxx> wrote:
> On Mon, 2017-12-11 at 12:30 +0100, Arnd Bergmann wrote:
>> One patch that was meant as a cleanup apparently did more than it intended,
>> allowing all combinations of legacy gadget drivers to be built into the
>> kernel, and leaving an empty 'choice' statement behind:
>>
>> drivers/usb/gadget/Kconfig:487:warning: choice default symbol 'USB_ETH' is not contained in the choice
>>
>> The description of commit 7a9618a22aad ("usb: gadget: allow to enable legacy
>> drivers without USB_ETH") was a bit cryptic, as it did not change the
>> behavior of USB_ETH other than allowing it to be built into the kernel
>> alongside other legacy gadgets, which is not a valid configuration.
>>
>> As Felipe explained in the description for commit bc49d1d17dcf ("usb:
>> gadget: don't couple configfs to legacy gadgets"), the configfs based
>> gadgets can be freely configured as loadable modules or built-in
>> drivers, but the legacy gadgets can only be modules if there is more
>> than one of them, so we require the 'choice' statement here.
>>
>> This leaves the added USB_GADGET_LEGACY menuconfig symbol in place,
>> but then restores the 'choice' below it, so we can enforce the
>> single-legacy-gadget rule as before.
>
> Hello Arnd,
>
> A discussion is ongoing about whether or not commit 7a9618a22aad should be reverted.
> Please drop this patch until a conclusion has been reached.

Ok. I'll use a revert of 7a9618a22aad in my local test tree then.
Reverting that is probably good, I thought about suggesting that
instead, but couldn't tell whether you had a bigger plan behind that
commit.

      Arnd