[v3,0/2] GPIO-based hotplug i2c bus

ASUS Transformers require this driver for proper work with their dock.
Dock is controlled by EC and its presence is detected by a GPIO.

The Transformers have a connector that's used for USB, charging or
for attaching a keyboard (called a dock; it also has a battery and
a touchpad). This connector probably (I don't have the means to verify
that) has an I2C bus lines and a "detect" line (pulled low on the dock
side) among the pins. I guess there is either no additional chip or
a transparent bridge/buffer chip, but nothing that could be controlled
by software. For DT this setup could be modelled like an I2C gate or
a 2-port mux with enable joining two I2C buses (one "closer" to the
CPU as a parent).

In this case it's hard to tell the difference if this is real or virtual
hardware.

This patchset is a predecessor of a possible larger patchset which
should bring support for a asus-ec, an i2c mfd device programmed by
Asus for their Transformers tablet line. Similar approach is used in
Microsoft Surface RT for attachable Type Cover.

> What is this actually doing?
Basically it duplicates the parent i2c bus once detection GPIO triggers
and probes all hot-pluggable devices which are connected to it. Once
GPIO triggers a detach signal all hot-pluggable devices are unprobed and
bus removed.

> Is the GPIO an irq line for signalling hoplugging and can be used by
> any driver or just this one?
It can be shared if necessary but usually all hot-pluggable devices
are gathered in one container and are plugged simultaneously.

---
Changes from v2:
- expanded descryption of driver implementation commit
- expanded descryption in patchset cover
- no changes to code or yaml from v2

Changes from v1:
- documentation changes:
  - dropped | from description
  - dropped nodename
  - unified use of quotes
  - used GPIO_ACTIVE_LOW define
  - used phandle instead of path
---

Michał Mirosław (1):
  i2c: Add GPIO-based hotplug gate

Svyatoslav Ryhel (1):
  dt-bindings: i2c: add binding for i2c-hotplug-gpio

 .../bindings/i2c/i2c-hotplug-gpio.yaml        |  65 +++++
 drivers/i2c/Kconfig                           |  11 +
 drivers/i2c/Makefile                          |   1 +
 drivers/i2c/i2c-hotplug-gpio.c                | 266 ++++++++++++++++++
 4 files changed, 343 insertions(+)
 create mode 100644 Documentation/devicetree/bindings/i2c/i2c-hotplug-gpio.yaml
 create mode 100644 drivers/i2c/i2c-hotplug-gpio.c

On 29/07/2023 18:08, Svyatoslav Ryhel wrote:
> Document device tree schema which describes hot-pluggable via GPIO
> i2c bus.
> 

A nit, subject: drop second/last, redundant "binding for". The
"dt-bindings" prefix is already stating that these are bindings.

> Signed-off-by: Svyatoslav Ryhel <clamor95@gmail.com>
> ---
>  .../bindings/i2c/i2c-hotplug-gpio.yaml        | 65 +++++++++++++++++++
>  1 file changed, 65 insertions(+)
>  create mode 100644 Documentation/devicetree/bindings/i2c/i2c-hotplug-gpio.yaml
> 
> diff --git a/Documentation/devicetree/bindings/i2c/i2c-hotplug-gpio.yaml b/Documentation/devicetree/bindings/i2c/i2c-hotplug-gpio.yaml
> new file mode 100644
> index 000000000000..21f2b74ca6c1
> --- /dev/null
> +++ b/Documentation/devicetree/bindings/i2c/i2c-hotplug-gpio.yaml
> @@ -0,0 +1,65 @@
> +# SPDX-License-Identifier: (GPL-2.0 OR BSD-2-Clause)
> +%YAML 1.2
> +---
> +$id: http://devicetree.org/schemas/i2c/i2c-hotplug-gpio.yaml#
> +$schema: http://devicetree.org/meta-schemas/core.yaml#
> +
> +title: GPIO detected hot-plugged I2C bus
> +
> +maintainers:
> +  - Michał Mirosław <mirq-linux@rere.qmqm.pl>
> +
> +description:
> +  Driver for hot-plugged I2C busses, where some devices on a bus
> +  are hot-pluggable and their presence is indicated by GPIO line.
> +
> +properties:
> +  compatible:
> +    items:

Drop items.

> +      - const: i2c-hotplug-gpio
> +
> +  '#address-cells':
> +    const: 1
> +
> +  '#size-cells':
> +    const: 0

Why do you need these two? You do not have any children.
> +
> +  interrupts-extended:

Just interrupts.

> +    minItems: 1

maxItems instead

> +
> +  detect-gpios:
> +    maxItems: 1
> +
> +  i2c-parent:
> +    maxItems: 1

Where is the type defined? Why this has maxItems? Is it an array?

> +
> +required:
> +  - compatible
> +  - '#address-cells'
> +  - '#size-cells'
> +  - interrupts-extended
> +  - detect-gpios
> +  - i2c-parent
> +
> +unevaluatedProperties: false

Without any $ref, this should be additionalProperties: false.

> +
> +examples:
> +  - |
> +    /*
> +     * Asus Transformers use I2C hotplug for attachable dock keyboard
> +     */
> +    #include <dt-bindings/gpio/gpio.h>
> +    #include <dt-bindings/interrupt-controller/irq.h>
> +
> +    i2c-dock {
> +        compatible = "i2c-hotplug-gpio";
> +
> +        #address-cells = <1>;
> +        #size-cells = <0>;
> +
> +        interrupts-extended = <&gpio 164 IRQ_TYPE_EDGE_BOTH>;
> +        detect-gpios = <&gpio 164 GPIO_ACTIVE_LOW>;
> +
> +        i2c-parent = <&gen2_i2c>;

So do you expect here any children or not?

> +    };
> +...

Best regards,
Krzysztof

On Sat, Jul 29, 2023 at 07:08:56PM +0300, Svyatoslav Ryhel wrote:
> Document device tree schema which describes hot-pluggable via GPIO
> i2c bus.

What's that? 'hot-pluggable via GPIO i2c bus' is not coherent.

> 
> Signed-off-by: Svyatoslav Ryhel <clamor95@gmail.com>
> ---
>  .../bindings/i2c/i2c-hotplug-gpio.yaml        | 65 +++++++++++++++++++
>  1 file changed, 65 insertions(+)
>  create mode 100644 Documentation/devicetree/bindings/i2c/i2c-hotplug-gpio.yaml
> 
> diff --git a/Documentation/devicetree/bindings/i2c/i2c-hotplug-gpio.yaml b/Documentation/devicetree/bindings/i2c/i2c-hotplug-gpio.yaml
> new file mode 100644
> index 000000000000..21f2b74ca6c1
> --- /dev/null
> +++ b/Documentation/devicetree/bindings/i2c/i2c-hotplug-gpio.yaml
> @@ -0,0 +1,65 @@
> +# SPDX-License-Identifier: (GPL-2.0 OR BSD-2-Clause)
> +%YAML 1.2
> +---
> +$id: http://devicetree.org/schemas/i2c/i2c-hotplug-gpio.yaml#
> +$schema: http://devicetree.org/meta-schemas/core.yaml#
> +
> +title: GPIO detected hot-plugged I2C bus
> +
> +maintainers:
> +  - Michał Mirosław <mirq-linux@rere.qmqm.pl>
> +
> +description:
> +  Driver for hot-plugged I2C busses, where some devices on a bus

Bindings are for h/w, not a driver.

> +  are hot-pluggable and their presence is indicated by GPIO line.
> +
> +properties:
> +  compatible:
> +    items:
> +      - const: i2c-hotplug-gpio
> +
> +  '#address-cells':
> +    const: 1
> +
> +  '#size-cells':
> +    const: 0

What are these for? You don't have any child nodes.

> +
> +  interrupts-extended:
> +    minItems: 1
> +
> +  detect-gpios:
> +    maxItems: 1
> +
> +  i2c-parent:
> +    maxItems: 1
> +
> +required:
> +  - compatible
> +  - '#address-cells'
> +  - '#size-cells'
> +  - interrupts-extended
> +  - detect-gpios
> +  - i2c-parent
> +
> +unevaluatedProperties: false
> +
> +examples:
> +  - |
> +    /*
> +     * Asus Transformers use I2C hotplug for attachable dock keyboard
> +     */
> +    #include <dt-bindings/gpio/gpio.h>
> +    #include <dt-bindings/interrupt-controller/irq.h>
> +
> +    i2c-dock {
> +        compatible = "i2c-hotplug-gpio";
> +
> +        #address-cells = <1>;
> +        #size-cells = <0>;
> +
> +        interrupts-extended = <&gpio 164 IRQ_TYPE_EDGE_BOTH>;
> +        detect-gpios = <&gpio 164 GPIO_ACTIVE_LOW>;

Looks like the same signal. You should only need one of them. Probably 
'detect-gpios' as I guess you need to read the state of the line.

> +
> +        i2c-parent = <&gen2_i2c>;

Couldn't you just add 'detect-gpios' to the existing bus node? It's 
really part of that bus and there's not a separate bus. That would be a 
lot simpler. I suppose you want to instantiate a driver, but that's not 
DT's problem.

Rob

> Not a driver but a group of devices (possibly discovered dynamically)
> behind a passive gate (like e.g. PCA9517A with EN tied to a connector pin).
> It's not much different to a I2C gate or mux with a single child bus
> (i2c-mux-gpio that has only a single child).

I agree. To prevent the the bus from spikes when connecting /
disconnecting something like PCA9517A should be in place. And this one
can be considered a 1:1 mux. Just with the exception that Linux cannot
control the gate, but can only react to changes to the enable pin.

Can't we have a driver for the PCA9517 which gets interrupts when the
enable pin changes and then adds / removes the child i2c adapter?