On Fri, Feb 16, 2024 at 04:54:29PM +0000, Conor Dooley wrote:
> On Fri, Feb 16, 2024 at 06:01:52PM +0200, Kalle Valo wrote:
> > (Adding devicetree list for comments)
> >
> > <Ajay.Kathat@xxxxxxxxxxxxx> writes:
> >
> > > On 2/13/24 09:58, Alexis Lothoré wrote:
> > >>
> > >> On 2/13/24 17:42, David Mosberger-Tang wrote:
> > >>> On Tue, 2024-02-13 at 16:22 +0100, Alexis Lothoré wrote:
> > >>>> When using a wilc1000 chip over a spi bus, users can optionally define a
> > >>>> reset gpio and a chip enable gpio. The reset line of wilc1000 is active
> > >>>> low, so to hold the chip in reset, a low (physical) value must be applied.
> > >>>>
> > >>>> The corresponding device tree binding documentation was introduced by
> > >>>> commit f31ee3c0a555 ("wilc1000: Document enable-gpios and reset-gpios
> > >>>> properties") and correctly indicates that the reset line is an active-low
> > >>>> signal. However, the corresponding driver part, brought by commit
> > >>>> ec031ac4792c ("wilc1000: Add reset/enable GPIO support to SPI driver"), is
> > >>>> misusing the gpiod APIs and apply an inverted logic when powering up/down
> > >>>> the chip (for example, setting the reset line to a logic "1" during power
> > >>>> up, which in fact asserts the reset line when device tree describes the
> > >>>> reset line as GPIO_ACTIVE_LOW).
> > >>>
> > >>> Note that commit ec031ac4792c is doing the right thing in regards to an
> > >>> ACTIVE_LOW RESET pin and the binding documentation is consistent with that code.
> > >>>
> > >>> It was later on that commit fcf690b0 flipped the RESET line polarity to treat it
> > >>> as GPIO_ACTIVE_HIGH. I never understood why that was done and, as you noted, it
> > >>> introduced in inconsistency with the binding documentation.
> > >>
> > >> Ah, you are right, and I was wrong citing your GPIOs patch as faulty
> > >> (git-blaming too fast !), thanks for the clarification. I missed this patch from
> > >> Ajay (fcf690b0) flipping the reset logic. Maybe he had issues while missing
> > >> proper device tree configuration and then submitted this flip ?
> > >
> > > Indeed, it was done to align the code as per the DT entry suggested in
> > > WILC1000/3000 porting guide[1 -page 18], which is already used by most
> > > of the existing users. This change has impact on the users who are using
> > > DT entry from porting guide. One approach is to retain the current code
> > > and document this if needed.
> >
> > So if I'm understanding the situation correctly Microchip's porting
> > guide[1] doesn't match with kernel.org documentation[2]? I'm not the
> > expert here but from my point of view the issue is clear: the code needs
> > to follow kernel.org documentation[2], not external documentation.
>
> My point of view would definitely be that drivers in the mainline kernel
> absolutely should respect the ABI defined in the dt-binding. What a vendor
> decides to do in their own tree I suppose is their problem, but I would
> advocate that vendor kernels would also respect the ABI from mainline.
>
> Looking a bit more closely at the porting guide, it contains other
> properties that are not present in the dt-binding - undocumented
> compatibles and a different enable gpio property for example.
> I guess it (and the vendor version of the driver) never got updated when
> wilc1000 supported landed in mainline?
>
> > I'll add devicetree list so hopefully people there can comment also,
> > full patch available in [3].
> >
> > Alexis, if there are no more comments I'm in favor submitting the revert
> > you mentioned.
>
> From a dt-bindings point of view, the aforementioned revert seems
> correct and would be
> Acked-by: Conor Dooley <conor.dooley@xxxxxxxxxxxxx>
Maybe an R-b is more suitable here, too used to acking trivial patches
that are dt related..
>
> Getting off my dt-binding maintainer high-horse, linux4microchip is going
> be updating to a 6.6 based kernel in the coming weeks - maybe that's a
> good time to update the vendor kernel wilc drivers (and therefore the
> porting guide?) to match the properties used by mainline Ajay?
>
> Cheers,
> Conor.
Attachment:
signature.asc
Description: PGP signature
