Re: [PATCH v4 1/4] dt-bindings: gpio: spacemit: add support for K1 SoC

[Rob Herring <robh@kernel.org>]

On Thu, Jan 23, 2025 at 03:19:18PM -0800, Olof Johansson wrote:
> On Thu, Jan 23, 2025 at 11:30:42AM +0000, Yixun Lan wrote:
> > Hi Olof:
> >  thanks for your reivew
> > 
> > On 12:03 Wed 22 Jan     , Olof Johansson wrote:
> > > Hi,
> > > 
> > > On Tue, Jan 21, 2025 at 11:38:11AM +0800, Yixun Lan wrote:
> > > > The GPIO controller of K1 support basic functions as input/output,
> > > > all pins can be used as interrupt which route to one IRQ line,
> > > > trigger type can be select between rising edge, failing edge, or both.
> > > > There are four GPIO ports, each consisting of 32 pins.
> > > > 
> > > > Signed-off-by: Yixun Lan <dlan@gentoo.org>
> > > > ---
> > > >  .../devicetree/bindings/gpio/spacemit,k1-gpio.yaml | 116 +++++++++++++++++++++
> > > >  1 file changed, 116 insertions(+)
> > > > 
> > > > diff --git a/Documentation/devicetree/bindings/gpio/spacemit,k1-gpio.yaml b/Documentation/devicetree/bindings/gpio/spacemit,k1-gpio.yaml
> > > > new file mode 100644
> > > > index 0000000000000000000000000000000000000000..dd9459061aecfcba84e6a3c5052fbcddf6c61150
> > > > --- /dev/null
> > > > +++ b/Documentation/devicetree/bindings/gpio/spacemit,k1-gpio.yaml
> > > > @@ -0,0 +1,116 @@
> > > > +# SPDX-License-Identifier: (GPL-2.0-only OR BSD-2-Clause)
> > > > +%YAML 1.2
> > > > +---
> > > > +$id: http://devicetree.org/schemas/gpio/spacemit,k1-gpio.yaml#
> > > > +$schema: http://devicetree.org/meta-schemas/core.yaml#
> > > > +
> > > > +title: SpacemiT K1 GPIO controller
> > > > +
> > > > +maintainers:
> > > > +  - Yixun Lan <dlan@gentoo.org>
> > > > +
> > > > +description:
> > > > +  The controller's registers are organized as sets of eight 32-bit
> > > > +  registers with each set of port controlling 32 pins.  A single
> > > > +  interrupt line is shared for all of the pins by the controller.
> > > > +  Each port will be represented as child nodes with the generic
> > > > +  GPIO-controller properties in this bindings file.
> > > 
> > > There's only one interrupt line for all ports, but you have a binding that
> > > duplicates them for every set of ports. That seems overly complicated,
> > > doesn't it? They'd all bind the same handler, so there's no benefit in
> > > providing the flexibility,.
> > > 
> > yes, all ports share same interrupt line, but each port has its own
> > irq related handling register, so it make sense to describe as per gpio irqchip
> > 
> > also see comments below
> > 
> > > > +properties:
> > > > +  $nodename:
> > > > +    pattern: "^gpio@[0-9a-f]+$"
> > > > +
> > > > +  compatible:
> > > > +    const: spacemit,k1-gpio
> > > > +
> > > > +  reg:
> > > > +    maxItems: 1
> > > > +
> > > > +  "#address-cells":
> > > > +    const: 1
> > > > +
> > > > +  "#size-cells":
> > > > +    const: 0
> > > > +
> > > > +patternProperties:
> > > > +  "^gpio-port@[0-9a-f]+$":
> > > > +    type: object
> > > > +    properties:
> > > > +      compatible:
> > > > +        const: spacemit,k1-gpio-port
> > > > +
> > > > +      reg:
> > > > +        maxItems: 1
> > > > +
> > > > +      gpio-controller: true
> > > > +
> > > > +      "#gpio-cells":
> > > > +        const: 2
> > > > +
> > > > +      gpio-ranges: true
> > > > +
> > > > +      interrupts:
> > > > +        maxItems: 1
> > > > +
> > > > +      interrupt-controller: true
> > > > +
> > > > +      "#interrupt-cells":
> > > > +        const: 2
> > > > +        description:
> > > > +          The first cell is the GPIO number, the second should specify interrupt
> > > > +          flag. The controller does not support level interrupts, so flags of
> > > > +          IRQ_TYPE_LEVEL_HIGH, IRQ_TYPE_LEVEL_LOW should not be used.
> > > > +          Refer <dt-bindings/interrupt-controller/irq.h> for valid flags.
> > > 
> > > Same here, since there's no real flexibility between the banks, it might
> > > make sense to consider a 3-cell GPIO specifier instead, and having
> > how to handle the fourth gpio port? I would like to have uniform driver for all ports
> > 
> > > the first cell indicate bank. I could see this argument go in either
> > > direction, but I'm not sure I understand why to provide a gpio-controller
> > > per bank.
> > > 
> > 
> > IIUC, your suggestion here was same as the implementation of patch v3 of this driver[1],
> >  while combining all four ports into one irqchip, which NACKed by maintainer[2].
> >  I tend to agree having a gpio-controller per bank provide more flexibility,
> >  easy to leverage generic gpio framework, even each port can be disabled or enabled,
> >  and IMO having shared irq handler isn't really a problem..
> > 
> > [1] https://lore.kernel.org/r/20241225-03-k1-gpio-v3-0-27bb7b441d62@gentoo.org
> > [2] https://lore.kernel.org/r/CACRpkdZPD2C2iPwOX_kW1Ug8jVkdHhhc7iFycHtzj5LQ0XWNgQ@mail.gmail.com
> > https://lore.kernel.org/r/CACRpkdYgGho=VQabonq4HccEiXBH2qM76K45oDaV1Jyi0xZ-YA@mail.gmail.com
> 
> Hmm, I don't understand the reasoning there, but it's not my subsystem.
> 
> It seems worse to me to misdescribe the hardware as separate blocks
> with a device-tree binding that no longer describes the actual hardware,
> but it's not up to me.

I agree. It's clearly 1 block given the first 3 banks are interleaved.

If Linux can't handle 1 node for N gpio_chip's, then that's a Linux 
problem. Maybe it can, IDK. The lookup from a DT node to gpio_chip just 
needs to match on more than just DT node pointer, but look at the node 
ptr and arg cells.

Rob