Re: [PATCH v3 1/5] dt-bindings: aspeed-sgpio: Convert txt bindings to yaml.

[Andrew Jeffery]

On Fri, 4 Jun 2021, at 13:00, Steven Lee wrote:
> The 06/04/2021 07:25, Andrew Jeffery wrote:
> > Hi Steven,
> > 
> > On Thu, 3 Jun 2021, at 19:48, Steven Lee wrote:
> > > sgpio-aspeed bindings should be converted to yaml format.
> > > 
> > > Signed-off-by: Steven Lee <steven_lee at aspeedtech.com>
> > > ---
> > >  .../bindings/gpio/aspeed,sgpio.yaml           | 78 +++++++++++++++++++
> > >  .../devicetree/bindings/gpio/sgpio-aspeed.txt | 46 -----------
> > >  2 files changed, 78 insertions(+), 46 deletions(-)
> > >  create mode 100644 Documentation/devicetree/bindings/gpio/aspeed,sgpio.yaml
> > >  delete mode 100644 Documentation/devicetree/bindings/gpio/sgpio-aspeed.txt
> > > 
> > > diff --git a/Documentation/devicetree/bindings/gpio/aspeed,sgpio.yaml 
> > > b/Documentation/devicetree/bindings/gpio/aspeed,sgpio.yaml
> > > new file mode 100644
> > > index 000000000000..e7c2113cc096
> > > --- /dev/null
> > > +++ b/Documentation/devicetree/bindings/gpio/aspeed,sgpio.yaml
> > > @@ -0,0 +1,78 @@
> > > +# SPDX-License-Identifier: (GPL-2.0-only OR BSD-2-Clause)
> > > +%YAML 1.2
> > > +---
> > > +$id: http://devicetree.org/schemas/gpio/aspeed,sgpio.yaml#
> > > +$schema: http://devicetree.org/meta-schemas/core.yaml#
> > > +
> > > +title: Aspeed SGPIO controller
> > > +
> > > +maintainers:
> > > +  - Andrew Jeffery <andrew at aj.id.au>
> > > +
> > > +description:
> > > +  This SGPIO controller is for ASPEED AST2400, AST2500 and AST2600 SoC,
> > > +  AST2600 have two sgpio master one with 128 pins another one with 80 
> > > pins,
> > > +  AST2500/AST2400 have one sgpio master with 80 pins. Each of the 
> > > Serial
> > > +  GPIO pins can be programmed to support the following options
> > > +  - Support interrupt option for each input port and various interrupt
> > > +    sensitivity option (level-high, level-low, edge-high, edge-low)
> > > +  - Support reset tolerance option for each output port
> > > +  - Directly connected to APB bus and its shift clock is from APB bus 
> > > clock
> > > +    divided by a programmable value.
> > > +  - Co-work with external signal-chained TTL components 
> > > (74LV165/74LV595)
> > > +
> > > +properties:
> > > +  compatible:
> > > +    enum:
> > > +      - aspeed,ast2400-sgpio
> > > +      - aspeed,ast2500-sgpio
> > > +      - aspeed,ast2600-sgpiom1
> > > +      - aspeed,ast2600-sgpiom2
> > 
> > You should have followed Rob's request here and made two patches for 
> > the binding document:
> > 
> > 1. A 1-to-1 conversion of the text file to dt-schema
> > 2. Add your new compatibles for the 2600.
> > 
> 
> Sorry I forgot to remove compatibles and move them to a new patch.
> 
> > From a cursory glance it looks okay except for the new compatibles.
> > 
> > Regarding the compatibles, I'd prefer we use something a bit more 
> > meaningful. What do you think of these?
> > 
> > - aspeed,ast2600-sgpiom-80
> > - aspeed,ast2600-sgpiom-128
> > 
> 
> Ok, I will change the name as you suggested.
> 
> BTW, I and development team have an internal discussion about the
> current sgpio design.
> 
> In the current design, the base offset of gpio input and output
> are calculated by the maximum number of gpio pins that SoC supported.
> For instance, in AST2500, max_ngpios is 80(defined in MAX_NR_HW_SGPIO),
> if ngpios is 16 in dts, gpio input pin id is from 0 to 15 and
> gpio output pin id is from 80 to 95.
> 
> We are thinking of removing max_ngpios(and MAX_NR_HW_SGPIO) and
> corresponding design to make the gpio input and output pin base
> are determined by ngpios.
> For instance, in any AST SoC, if ngpios is 16 in dts,
> gpio input pin id is from 0 to 15 and gpio output pin id is from 16 to 31.
> Thus we don't need to care about the max_ngpios of SoCs, and needn't to
> add 2 compatibles for ast2600.
> 
> However, it might affect users who update kernel/driver from the
> old kernel/driver as they may expect the gpio output pin base is start
> from 80(MAX_NR_HW_SGPIO).
> I was wondering if it is better to change the design as above.
> It would be great to have your suggestion.

Right, this breaks userspace. I don't think it's going to fly but I'm 
interested in feedback from Linus and Bartosz.

If we were to break userspace, a scheme I'd consider with is to pair 
input/output GPIOs. For example, GPIO 0 is input, GPIO 1 is the 
associated output, GPIO 2 is input, GPIO 3 is output etc. That way you 
can increase/decrease the number of GPIOs without affecting userspace 
(after breaking it initially).

Andrew