RE: [PATCH v5 1/4] dt-bindings: aspeed: Add eSPI controller

[ChiaWei Wang <chiawei_wang@aspeedtech.com>]

Hi Rob,

> From: Rob Herring <robh@kernel.org>
> Sent: Wednesday, May 18, 2022 2:32 AM
> 
> On Mon, May 16, 2022 at 08:54:09AM +0800, Chia-Wei Wang wrote:
> > Add dt-bindings for Aspeed eSPI controller
> >
> > Signed-off-by: Chia-Wei Wang <chiawei_wang@aspeedtech.com>
> > ---
> >  .../devicetree/bindings/soc/aspeed/espi.yaml  | 162
> > ++++++++++++++++++
> 
> bindings/spi/ includes SPI slaves. Is there a reason this doesn't fit there?

eSPI resues the timing and electrical specification of SPI but runs completely different protocol.
Only the flash channel is related to SPI and the other 3 channels are for EC/BMC/SIO.
Therefore, an eSPI driver does not fit into the SPI model.

> 
> >  1 file changed, 162 insertions(+)
> >  create mode 100644
> > Documentation/devicetree/bindings/soc/aspeed/espi.yaml
> >
> > diff --git a/Documentation/devicetree/bindings/soc/aspeed/espi.yaml
> > b/Documentation/devicetree/bindings/soc/aspeed/espi.yaml
> > new file mode 100644
> > index 000000000000..aa91ec8caf6a
> > --- /dev/null
> > +++ b/Documentation/devicetree/bindings/soc/aspeed/espi.yaml
> > @@ -0,0 +1,162 @@
> > +# SPDX-License-Identifier: (GPL-2.0-only OR BSD-2-Clause) # #
> > +Copyright (c) 2021 Aspeed Technology Inc.
> > +%YAML 1.2
> > +---
> > +$id: "http://devicetree.org/schemas/soc/aspeed/espi.yaml#"
> > +$schema: "http://devicetree.org/meta-schemas/core.yaml#"
> > +
> > +title: Aspeed eSPI Controller
> > +
> > +maintainers:
> > +  - Chia-Wei Wang <chiawei_wang@aspeedtech.com>
> > +  - Ryan Chen <ryan_chen@aspeedtech.com>
> > +
> > +description:
> > +  Aspeed eSPI controller implements a slave side eSPI endpoint device
> > +  supporting the four eSPI channels, namely peripheral, virtual wire,
> > +  out-of-band, and flash.
> > +
> > +properties:
> > +  compatible:
> > +    items:
> > +      - enum:
> > +          - aspeed,ast2500-espi
> > +          - aspeed,ast2600-espi
> > +      - const: simple-mfd
> > +      - const: syscon
> > +
> > +  reg:
> > +    maxItems: 1
> > +
> > +  "#address-cells":
> > +    const: 1
> > +
> > +  "#size-cells":
> > +    const: 1
> > +
> > +  ranges: true
> > +
> > +patternProperties:
> > +  "^espi-ctrl@[0-9a-f]+$":
> > +    type: object
> > +
> > +    description: Control of the four basic eSPI channels
> > +
> > +    properties:
> > +      compatible:
> > +        items:
> > +          - enum:
> > +              - aspeed,ast2500-espi-ctrl
> > +              - aspeed,ast2600-espi-ctrl
> > +
> > +      interrupts:
> > +        maxItems: 1
> > +
> > +      clocks:
> > +        maxItems: 1
> > +
> > +      perif,memcyc-enable:
> 
> What vendor is 'perif'?

It refers to the eSPI peripheral channel.

> 
> > +        type: boolean
> > +        description: Enable memory cycle over eSPI peripheral channel
> > +
> > +      perif,memcyc-src-addr:
> > +        $ref: /schemas/types.yaml#/definitions/uint32
> > +        description: The Host side address to be decoded into the
> > + memory cycle over eSPI peripheral channel
> > +
> > +      perif,memcyc-size:
> > +        $ref: /schemas/types.yaml#/definitions/uint32
> > +        description: The size of the memory region allocated for the
> memory cycle over eSPI peripheral channel
> > +        minimum: 65536
> 
> This region is defined by the h/w or just some carveout of system memory? In
> the former, perhaps this should be part of 'reg'. In the latter case, use a
> /reserved-memory node and memory-region here.

The region is going to be allocated at runtime phase.
It is a kind of shared memory between Host and BMC.

> 
> > +
> > +      perif,dma-mode:
> > +        type: boolean
> > +        description: Enable DMA support for eSPI peripheral channel
> > +
> > +      oob,dma-mode:
> 
> What vendor is 'oob'?

It refers to the eSPI out-of-band channel.

> 
> > +        type: boolean
> > +        description: Enable DMA support for eSPI out-of-band channel
> > +
> > +      oob,dma-tx-desc-num:
> > +        $ref: /schemas/types.yaml#/definitions/uint32
> > +        minimum: 2
> > +        maximum: 1023
> > +        description: The number of TX descriptors available for eSPI
> > + OOB DMA engine
> > +
> > +      oob,dma-rx-desc-num:
> > +        $ref: /schemas/types.yaml#/definitions/uint32
> > +        minimum: 2
> > +        maximum: 1023
> > +        description: The number of RX descriptors available for eSPI
> > + OOB DMA engine
> > +
> > +      flash,dma-mode:
> > +        type: boolean
> > +        description: Enable DMA support for eSPI flash channel
> 
> Why does this need to be in DT. It's configuration.

The property is used to decide the operation mode (i.e. FIFO or DMA) of the eSPI flash channel.
Is it a wrong idea to use the DTS property for?

> 
> > +
> > +      flash,safs-mode:
> > +        $ref: /schemas/types.yaml#/definitions/uint32
> > +        enum: [ 0, 1, 2 ]
> > +        default: 0
> > +        description: Slave-Attached-Sharing-Flash mode, 0->Mix,
> > + 1->SW, 2->HW
> > +
> > +    dependencies:
> > +      perif,memcyc-src-addr: [ "perif,memcyc-enable" ]
> > +      perif,memcyc-size: [ "perif,memcyc-enable" ]
> > +      oob,dma-tx-desc-num: [ "oob,dma-mode" ]
> > +      oob,dma-rx-desc-num: [ "oob,dma-mode" ]
> > +
> > +    required:
> > +      - compatible
> > +      - interrupts
> > +      - clocks
> > +
> > +  "^espi-mmbi@[0-9a-f]+$":
> > +    type: object
> > +
> > +    description: Control of the PCH-BMC data exchange over eSPI
> > + peripheral memory cycle
> > +
> > +    properties:
> > +      compatible:
> > +        const: aspeed,ast2600-espi-mmbi
> > +
> > +      interrupts:
> > +        maxItems: 1
> > +
> > +    required:
> > +      - compatible
> > +      - interrupts
> > +
> > +required:
> > +  - compatible
> > +  - reg
> > +  - "#address-cells"
> > +  - "#size-cells"
> > +  - ranges
> > +
> > +additionalProperties: false
> > +
> > +examples:
> > +  - |
> > +    #include <dt-bindings/interrupt-controller/arm-gic.h>
> > +    #include <dt-bindings/clock/ast2600-clock.h>
> > +
> > +    espi: espi@1e6ee000 {
> > +        compatible = "aspeed,ast2600-espi", "simple-mfd", "syscon";
> > +        reg = <0x1e6ee000 0x1000>;
> > +
> > +        #address-cells = <1>;
> > +        #size-cells = <1>;
> > +        ranges = <0x0 0x1e6ee000 0x1000>;
> > +
> > +        espi_ctrl: espi-ctrl@0 {
> > +            compatible = "aspeed,ast2600-espi-ctrl";
> > +            reg = <0x0 0x800>;
> > +            interrupts = <GIC_SPI 42 IRQ_TYPE_LEVEL_HIGH>;
> > +            clocks = <&syscon ASPEED_CLK_GATE_ESPICLK>;
> > +        };
> > +
> > +        espi_mmbi: espi-mmbi@800 {
> > +            compatible = "aspeed,ast2600-espi-mmbi";
> > +            reg = <0x800 0x50>;
> > +            interrupts = <GIC_SPI 108 IRQ_TYPE_LEVEL_HIGH>;
> > +        };
> 
> Why do you need these child nodes? Are the subblocks somehow useful on
> their own or reuseable in another configuration? If not, looks like this could all
> be 1 node.

espi-mmbi has individual function and control registers.
However, espi-mmbi is also a feature extended based on the memory cycle of eSPI peripheral channel.
Thereby, it has dependency on the eSPI channel initialization conducted by espi-ctrl.
The scenario is similar to the lpc-ctrl and other lpc-xxx drivers of Aspeed SoCs.

Chiawei