Re: [PATCH 1/2] dt-bindings: linflexuart: add clock definitions

[Conor Dooley]

On Tue, Sep 24, 2024 at 06:32:30PM +0300, Ciprian Marian Costea wrote:
> On 9/24/2024 6:27 PM, Conor Dooley wrote:
> > On Tue, Sep 24, 2024 at 06:17:11PM +0300, Ciprian Marian Costea wrote:
> > > On 9/24/2024 6:01 PM, Conor Dooley wrote:
> > > > On Tue, Sep 24, 2024 at 05:52:13PM +0300, Ciprian Marian Costea wrote:
> > > > > On 9/24/2024 5:24 PM, Conor Dooley wrote:
> > > > > > On Tue, Sep 24, 2024 at 05:12:46PM +0300, Ciprian Costea wrote:
> > > > > > > From: Ciprian Marian Costea <ciprianmarian.costea@xxxxxxxxxxx>
> > > > > > > 
> > > > > > > Add clock definitions for NXP LINFlexD UART bindings
> > > > > > > and update the binding examples with S32G2 node.
> > > > > > > 
> > > > > > > Signed-off-by: Ciprian Marian Costea <ciprianmarian.costea@xxxxxxxxxxx>
> > > > > > > ---
> > > > > > >     .../bindings/serial/fsl,s32-linflexuart.yaml  | 21 +++++++++++++++++++
> > > > > > >     1 file changed, 21 insertions(+)
> > > > > > > 
> > > > > > > diff --git a/Documentation/devicetree/bindings/serial/fsl,s32-linflexuart.yaml b/Documentation/devicetree/bindings/serial/fsl,s32-linflexuart.yaml
> > > > > > > index 4171f524a928..45fcab9e186d 100644
> > > > > > > --- a/Documentation/devicetree/bindings/serial/fsl,s32-linflexuart.yaml
> > > > > > > +++ b/Documentation/devicetree/bindings/serial/fsl,s32-linflexuart.yaml
> > > > > > > @@ -34,6 +34,14 @@ properties:
> > > > > > >       interrupts:
> > > > > > >         maxItems: 1
> > > > > > > +  clocks:
> > > > > > > +    maxItems: 2
> > > > > > > +
> > > > > > > +  clock-names:
> > > > > > > +    items:
> > > > > > > +      - const: ipg
> > > > > > > +      - const: lin
> > > > > > 
> > > > > > Can all devices have 2 clocks, or just the s32g2?
> > > > > > 
> > > > > 
> > > > > All devices (S32G2, S32G3 and S32V234) have 2 clocks for LINFlexD module.
> > > > 
> > > > I see. How come the driver is capable of working without them?
> > > > 
> > > 
> > > The driver was working because the LINFlexD clocks were configured and kept
> > > enabled by the downstream bootloader (TF-A and U-Boot). This is not ideal
> > > since LINFlexD Linux driver should manage those clocks independently and not
> > > rely on a previous bootloader configuration (hence the need for this current
> > > patchset).
> > 
> > I'd also mark them as required in the binding too, but the driver will
> > still have to account for them being missing, for backwards
> > compatibility reasons. Add a comment explaining the history to the
> > commit message when you do that.
> 
> Already in the second patch from this patchset the clocking support was
> added in the LINFlexD driver as optional, indeed for backwards
> compatibility.

Oh great.

> I presumed that because of this optional clock enablement, I should not add
> the clocks as required in the bindings, but I will do so in V2. Thanks.

IMO required is correct, since it would not have worked properly if the
bootloader hadn't done the setup.

Attachment: signature.asc
Description: PGP signature