[PATCH] ARM: multi_v7_defconfig: Select CONFIG_USB_ONBOARD_DEV as built-in

[Arnd Bergmann]

On Tue, Apr 30, 2024, at 21:53, Fabio Estevam wrote:
> On Tue, Apr 30, 2024 at 4:53 AM Arnd Bergmann <arnd at arndb.de> wrote:
>
>> It does sound like this is something the kernel should be able to
>> get to work properly in some form, but I don't think making it
>> a 'bool' symbol is the correct answer here: if CONFIG_USB is
>> set to =m, it would be impossible to include USB_ONBOARD_DEV
>> in this case.
>>
>> Fabio, can you explain how making it built-in addresses the
>> problem here? I assume this is related to probe order, so I
>> wonder if it's just a matter of making the usb hub driver
>> properly handle -EPROBE_DEFER until the onboard dev has been
>> initialized.
>
> From drivers/usb/misc/Kconfig:
>
> "config USB_ONBOARD_DEV
> tristate "Onboard USB device support"
> .....
>   This driver can be used as a module but its state (module vs
>   builtin) must match the state of the USB subsystem. Enabling
>   this config will enable the driver and it will automatically
>   match the state of the USB subsystem. If this driver is a
>   module it will be called onboard_usb_dev."

Ok, so there is some kind of design mistake here that this
is trying to paper over. Kbuild should always be powerful
enough to enforce these things without having a person read
a comment though.

> In multi_v7_defconfig:
> CONFIG_USB=y and CONFIG_USB_ONBOARD_DEV=m, so there is a mismatch.
>
> My patch enforces CONFIG_USB_ONBOARD_DEV=y to guarantee the matching.
>
> Is there any other way to solve this?

I can think of multiple ways to enforce this in Kbuild:

a) make CONFIG_USB_ONBOARD_DEV a 'bool' symbol and then add
   a hidden symbol that duplicates the state of CONFIG_USB when
   USB_ONBOARD_DEV is enabled:

config CONFIG_USB_ONBOARD_DEV_MODULE
     def_tristate USB
     depends on CONFIG_USB_ONBOARD_DEV

b) Do the same thing but use Makefile syntax instead of Kconfig
   syntax:

usb-onboard-dev-$(CONFIG_USB_ONBOARD_DEV) += onboard_usb_dev.o 
obj-$(CONFIG_USB) += usb-onboard-dev.o

(or something close to this, need to try).


However, I still feel there should be a better way to solve
the problem at the code level rather than the Kbuild level.

I don't know exactly how a "usb_device_driver" (as opposed
to a "usb_driver") works here, but from what I can tell,
the idea here is that the USB bus probe finds a device, matches
the driver and then continues probing it. If the onboard-dev
driver gets loaded after the initial bus probe, that driver
won't exist yet and it then runs into some error that
prevents it from being retried when after the onboard-dev
driver is there. If you can pinpoint exactly what error it
runs into, try to make it retry the same thing after
-EPROBE_DEFER.

    Arnd