[v17,7/7] usb: Specify dependencies on USB_XHCI_PLATFORM with 'depends on'

Some USB controller drivers that depend on the xhci-plat driver
specify this dependency using 'select' in Kconfig. This is not
recommended for symbols that have other dependencies as it may
lead to invalid configurations. Use 'depends on' to specify the
dependency instead of 'select'.

For dwc3 specify the dependency on USB_XHCI_PLATFORM in
USB_DWC3_HOST and USB_DWC3_DUAL_ROLE. Also adjust the
dependencies of USB_DWC3_CORE to make sure that at least one
of USB_DWC3_HOST, USB_DWC3_GADGET or USB_DWC3_DUAL_ROLE can be
selected.

Signed-off-by: Matthias Kaehlcke <mka@chromium.org>
Reviewed-by: Roger Quadros <rogerq@kernel.org>
Reviewed-by: Douglas Anderson <dianders@chromium.org>
---

Changes in v17:
- removed explicit dependency on USB from USB_DWC3
- added 'Reviewed-by' tags from Roger and Doug

Changes in v16:
- none

Changes in v15:
- adjusted dependencies of USB_DWC3_CORE to make sure it can only
  be enabled when at least one of USB_DWC3_HOST, USB_DWC3_GADGET
  or USB_DWC3_DUAL_ROLE is selectable
- updated commit message

Changes in v14:
- none

Changes in v13:
- patch added to the series

 drivers/usb/cdns3/Kconfig | 2 +-
 drivers/usb/dwc3/Kconfig  | 5 +++--
 drivers/usb/host/Kconfig  | 4 ++--
 3 files changed, 6 insertions(+), 5 deletions(-)

On Tue, Nov 16, 2021 at 12:07:39PM -0800, Matthias Kaehlcke wrote:
> Some USB controller drivers that depend on the xhci-plat driver
> specify this dependency using 'select' in Kconfig. This is not
> recommended for symbols that have other dependencies as it may
> lead to invalid configurations. Use 'depends on' to specify the
> dependency instead of 'select'.
> 
> For dwc3 specify the dependency on USB_XHCI_PLATFORM in
> USB_DWC3_HOST and USB_DWC3_DUAL_ROLE. Also adjust the
> dependencies of USB_DWC3_CORE to make sure that at least one
> of USB_DWC3_HOST, USB_DWC3_GADGET or USB_DWC3_DUAL_ROLE can be
> selected.
> 
> Signed-off-by: Matthias Kaehlcke <mka@chromium.org>
> Reviewed-by: Roger Quadros <rogerq@kernel.org>
> Reviewed-by: Douglas Anderson <dianders@chromium.org>

Note: This patch has been removed from the onboard_usb_hub series,
together with "ARM: configs: Explicitly enable USB_XHCI_PLATFORM
where needed" and "arm64: defconfig: Explicitly enable
USB_XHCI_PLATFORM". These patches aren't any longer needed for the
series. If maintainers think they are useful independently from
the series please pick them or let me know what needs to be
changed to get them landed.

17.12.2021 02:56, Matthias Kaehlcke пишет:
> On Tue, Nov 16, 2021 at 12:07:39PM -0800, Matthias Kaehlcke wrote:
>> Some USB controller drivers that depend on the xhci-plat driver
>> specify this dependency using 'select' in Kconfig. This is not
>> recommended for symbols that have other dependencies as it may
>> lead to invalid configurations. Use 'depends on' to specify the
>> dependency instead of 'select'.
>>
>> For dwc3 specify the dependency on USB_XHCI_PLATFORM in
>> USB_DWC3_HOST and USB_DWC3_DUAL_ROLE. Also adjust the
>> dependencies of USB_DWC3_CORE to make sure that at least one
>> of USB_DWC3_HOST, USB_DWC3_GADGET or USB_DWC3_DUAL_ROLE can be
>> selected.
>>
>> Signed-off-by: Matthias Kaehlcke <mka@chromium.org>
>> Reviewed-by: Roger Quadros <rogerq@kernel.org>
>> Reviewed-by: Douglas Anderson <dianders@chromium.org>
> 
> Note: This patch has been removed from the onboard_usb_hub series,
> together with "ARM: configs: Explicitly enable USB_XHCI_PLATFORM
> where needed" and "arm64: defconfig: Explicitly enable
> USB_XHCI_PLATFORM". These patches aren't any longer needed for the
> series. If maintainers think they are useful independently from
> the series please pick them or let me know what needs to be
> changed to get them landed.
> 

Hi,

I don't know what this all is about, perhaps I'm CC'ed semi-randomly
because touched that Kconfig once.

All I can say here is that the commit message tells us "This is not
recommended" and doesn't explain what's the actual problem is being
solved. If there is no real problem, why bother?

On Fri, Dec 17, 2021 at 03:47:24AM +0300, Dmitry Osipenko wrote:
> 17.12.2021 02:56, Matthias Kaehlcke пишет:
> > On Tue, Nov 16, 2021 at 12:07:39PM -0800, Matthias Kaehlcke wrote:
> >> Some USB controller drivers that depend on the xhci-plat driver
> >> specify this dependency using 'select' in Kconfig. This is not
> >> recommended for symbols that have other dependencies as it may
> >> lead to invalid configurations. Use 'depends on' to specify the
> >> dependency instead of 'select'.
> >>
> >> For dwc3 specify the dependency on USB_XHCI_PLATFORM in
> >> USB_DWC3_HOST and USB_DWC3_DUAL_ROLE. Also adjust the
> >> dependencies of USB_DWC3_CORE to make sure that at least one
> >> of USB_DWC3_HOST, USB_DWC3_GADGET or USB_DWC3_DUAL_ROLE can be
> >> selected.
> >>
> >> Signed-off-by: Matthias Kaehlcke <mka@chromium.org>
> >> Reviewed-by: Roger Quadros <rogerq@kernel.org>
> >> Reviewed-by: Douglas Anderson <dianders@chromium.org>
> > 
> > Note: This patch has been removed from the onboard_usb_hub series,
> > together with "ARM: configs: Explicitly enable USB_XHCI_PLATFORM
> > where needed" and "arm64: defconfig: Explicitly enable
> > USB_XHCI_PLATFORM". These patches aren't any longer needed for the
> > series. If maintainers think they are useful independently from
> > the series please pick them or let me know what needs to be
> > changed to get them landed.
> > 
> 
> Hi,
> 
> I don't know what this all is about, perhaps I'm CC'ed semi-randomly
> because touched that Kconfig once.

Yes, it seems tools select you based on their heuristics because you
made changes to that file.

> All I can say here is that the commit message tells us "This is not
> recommended" and doesn't explain what's the actual problem is being
> solved. If there is no real problem, why bother?

Earlier versions of the onboard_usb_hub series [1] which had a dependency
involving USB_XHCI_PLATFORM had an issue with invalid (rand)configs
that was related with the 'selects'.

The series doesn't depend on USB_XHCI_PLATFORM any longer, hence the
original issue doesn't exist anymore, however it might re-surface in
the future.

Personally I have no vested interest at this point in getting the
config changes landed, I just wanted to make clear what the status
is (split off from the series, no future versions unless someone
requests them), rather than abandoning them silently.

[1]: https://patchwork.kernel.org/project/linux-usb/list/?series=531343

20.12.2021 21:44, Matthias Kaehlcke пишет:
> On Fri, Dec 17, 2021 at 03:47:24AM +0300, Dmitry Osipenko wrote:
>> 17.12.2021 02:56, Matthias Kaehlcke пишет:
>>> On Tue, Nov 16, 2021 at 12:07:39PM -0800, Matthias Kaehlcke wrote:
>>>> Some USB controller drivers that depend on the xhci-plat driver
>>>> specify this dependency using 'select' in Kconfig. This is not
>>>> recommended for symbols that have other dependencies as it may
>>>> lead to invalid configurations. Use 'depends on' to specify the
>>>> dependency instead of 'select'.
>>>>
>>>> For dwc3 specify the dependency on USB_XHCI_PLATFORM in
>>>> USB_DWC3_HOST and USB_DWC3_DUAL_ROLE. Also adjust the
>>>> dependencies of USB_DWC3_CORE to make sure that at least one
>>>> of USB_DWC3_HOST, USB_DWC3_GADGET or USB_DWC3_DUAL_ROLE can be
>>>> selected.
>>>>
>>>> Signed-off-by: Matthias Kaehlcke <mka@chromium.org>
>>>> Reviewed-by: Roger Quadros <rogerq@kernel.org>
>>>> Reviewed-by: Douglas Anderson <dianders@chromium.org>
>>>
>>> Note: This patch has been removed from the onboard_usb_hub series,
>>> together with "ARM: configs: Explicitly enable USB_XHCI_PLATFORM
>>> where needed" and "arm64: defconfig: Explicitly enable
>>> USB_XHCI_PLATFORM". These patches aren't any longer needed for the
>>> series. If maintainers think they are useful independently from
>>> the series please pick them or let me know what needs to be
>>> changed to get them landed.
>>>
>>
>> Hi,
>>
>> I don't know what this all is about, perhaps I'm CC'ed semi-randomly
>> because touched that Kconfig once.
> 
> Yes, it seems tools select you based on their heuristics because you
> made changes to that file.
> 
>> All I can say here is that the commit message tells us "This is not
>> recommended" and doesn't explain what's the actual problem is being
>> solved. If there is no real problem, why bother?
> 
> Earlier versions of the onboard_usb_hub series [1] which had a dependency
> involving USB_XHCI_PLATFORM had an issue with invalid (rand)configs
> that was related with the 'selects'.
> 
> The series doesn't depend on USB_XHCI_PLATFORM any longer, hence the
> original issue doesn't exist anymore, however it might re-surface in
> the future.
> 
> Personally I have no vested interest at this point in getting the
> config changes landed, I just wanted to make clear what the status
> is (split off from the series, no future versions unless someone
> requests them), rather than abandoning them silently.
> 
> [1]: https://patchwork.kernel.org/project/linux-usb/list/?series=531343
> 

My point is that you should always explain in a commit message what
problem is being solved and how it's solved. There is no explanation of
the problem in the commit of this patch. If there is no real problem and
patch is a minor improvement, then it also should be explained explicitly.