Re: [PATCH v9 2/2] usb: dwc3: Add Google Tensor SoC DWC3 glue driver

[Krzysztof Kozlowski <krzk@kernel.org>]

On 18/12/2025 02:34, Roy Luo wrote:
> On Wed, Dec 17, 2025 at 12:14 PM Doug Anderson <dianders@google.com> wrote:
>>
>> Hi,
>>
>> On Wed, Dec 17, 2025 at 11:18 AM Roy Luo <royluo@google.com> wrote:
>>>
>>> On Wed, Dec 17, 2025 at 5:24 AM Greg Kroah-Hartman
>>> <gregkh@linuxfoundation.org> wrote:
>>>>
>>>> On Thu, Dec 04, 2025 at 11:14:39PM -0800, Roy Luo wrote:
>>>>> On Thu, Dec 4, 2025 at 10:05 PM Greg Kroah-Hartman
>>>>> <gregkh@linuxfoundation.org> wrote:
>>>>>>
>>>>>> On Fri, Dec 05, 2025 at 02:26:38AM +0000, Roy Luo wrote:
>>>>>>> +config USB_DWC3_GOOGLE
>>>>>>> +     tristate "Google Platform"
>>>>>>> +     depends on ARCH_GOOGLE || COMPILE_TEST
>>>>>>
>>>>>> There is no ARCH_GOOGLE in the tree now, so how is this supposed to
>>>>>> work?  Shouldn't tools that check for "invalid config options" trigger
>>>>>> on this?
>>>>>>
>>>>>> thanks,
>>>>>>
>>>>>> greg k-h
>>>>>
>>>>> Hi Greg,
>>>>>
>>>>> The menuconfig looks like the following and it doesn't complain:
>>>>> | Symbol: ARCH_GOOGLE [=ARCH_GOOGLE]
>>>>> | Type  : unknown
>>>>> |
>>>>> | Symbol: PHY_GOOGLE_USB [=y]
>>>>> | Type  : tristate
>>>>> | Defined at drivers/phy/Kconfig:104
>>>>> |     Prompt: Google Tensor SoC USB PHY driver
>>>>> |     Depends on: ARCH_GOOGLE || COMPILE_TEST [=y]
>>>>>
>>>>> According to Kconfig documentation [1], the unknown symbol
>>>>> would simply be evaluated as an "n", which is what we want.
>>>>> "Convert the symbol into an expression. Boolean and tristate
>>>>> symbols are simply converted into the respective expression
>>>>> values. All other symbol types result in ‘n’."
>>>>>
>>>>> In a different Kconfig documentation, an environment variable
>>>>> "KCONFIG_WARN_UNKNOWN_SYMBOLS" is there to detect
>>>>> undefined symbols in the "config input", but I can't find one that
>>>>> catches undefined symbols in the Kconfig tree itself.
>>>>>
>>>>> That is, the tool seems to allow this.
>>>>> However, if this turns out to be a major problem. I think we
>>>>> can either:
>>>>> - Remove ARCH_GOOGLE and leave COMPILE_TEST as
>>>>>   the only dependency. Then add ARCH_GOOGLE back
>>>>>   later once it's in the tree.
>>>>
>>>> Please do this.  I do not want to take patches that purposfully add
>>>> dependencies on config options that might, or might not, appear in the
>>>> future.  Please just remove all of the dependancies for now, as they are
>>>> not needed, right?
>>>>
>>>> thanks,
>>>>
>>>> greg k-h
>>>
>>> Greg,
>>>
>>> Yes, we can remove ARCH_GOOGLE for now.
>>> To clarify, we're not removing all of the dependencies, we still want
>>> to keep COMPILE_TEST for build tests, right?
>>> Please let me know if you think otherwise.
>>
>> I think you'd just remove all of them. Normally COMPILE_TEST just
>> allows folks to compile stuff even when they don't want the ARCH. We
>> can can add ARCH_GOOGLE back in later once the config exists.
>>
>> -Doug
> 
> Doug,
> 
> Thanks for chiming in. I'm hesitant to remove COMPILE_TEST
> because Greg specifically requested its inclusion in v7 [1].


What? No, that's not what Greg requested. Your COMPILE_TEST in current
form helps nothing in build testing. It makes absolutely no sense at all.

> Also it seems beneficial to get some build coverage before
> ARCH_GOOGLE is officially added to the tree.

And COMPILE_TEST like you wrote it does not give you that. Please first
read how this function works.

> Greg, could you clarify?

Can you first look at Linux kernel sources to see how this is properly
written?

> 
> [1] https://lore.kernel.org/linux-usb/2025112144-claw-recolor-49c3@gregkh/
> 
> Thanks,
> Roy


Best regards,
Krzysztof