Re: [PATCH 08/20] dt-bindings: reset: add Synaptics SL261X SoCs

[Jisheng Zhang <jszhang@kernel.org>]

On Wed, Sep 30, 2026 at 05:49:18PM +0100, Conor Dooley wrote:
> On Wed, Sep 30, 2026 at 11:15:48PM +0800, Jisheng Zhang wrote:
> > On Tue, Sep 29, 2026 at 08:44:13PM +0100, Conor Dooley wrote:
> > > On Tue, Sep 29, 2026 at 02:14:05PM +0800, Jisheng Zhang wrote:
> > > > Add device tree bindings for the resets on Synaptics SL261X SoCs.
> > > > 
> > > > Signed-off-by: Jisheng Zhang <jszhang@kernel.org>
> > > > ---
> > > >  .../bindings/reset/syna,sl261x-reset.yaml     | 40 ++++++++++
> > > >  include/dt-bindings/reset/syna,sl261x-reset.h | 80 +++++++++++++++++++
> > > >  2 files changed, 120 insertions(+)
> > > >  create mode 100644 Documentation/devicetree/bindings/reset/syna,sl261x-reset.yaml
> > > >  create mode 100644 include/dt-bindings/reset/syna,sl261x-reset.h
> > > > 
> > > > diff --git a/Documentation/devicetree/bindings/reset/syna,sl261x-reset.yaml b/Documentation/devicetree/bindings/reset/syna,sl261x-reset.yaml
> > > > new file mode 100644
> > > > index 000000000000..8cfec17df800
> > > > --- /dev/null
> > > > +++ b/Documentation/devicetree/bindings/reset/syna,sl261x-reset.yaml
> > > > @@ -0,0 +1,40 @@
> > > > +# SPDX-License-Identifier: (GPL-2.0 OR BSD-2-Clause)
> > > > +# Copyright (C) 2026 Synaptics Incorporated
> > > > +%YAML 1.2
> > > > +---
> > > > +$id: http://devicetree.org/schemas/reset/syna,sl261x-reset.yaml#
> > > > +$schema: http://devicetree.org/meta-schemas/core.yaml#
> > > > +
> > > > +title: Synaptics SL261X reset controller
> > > > +
> > > > +maintainers:
> > > > +  - Jisheng Zhang <jszhang@kernel.org>
> > > > +
> > > > +description: The reset controller node must be a sub-node of the chip
> > > > +  controller node on SL261X SoCs.
> > > > +
> > > > +properties:
> > > > +  compatible:
> > > > +    enum:
> > > > +      - syna,sl261x-soc-reset
> > > > +      - syna,sl261x-system-reset
> > > > +
> > > > +  "#reset-cells":
> > > > +    const: 1
> > > > +
> > > > +required:
> > > > +  - compatible
> > > > +  - "#reset-cells"
> > > > +
> > > > +additionalProperties: false
> > > > +
> > > > +examples:
> > > > +  - |
> > > > +    chip: chip-control@f7e10000 {
> > > > +        reg = <0xf7e10000 0x1000>;
> > > > +
> > > > +        chip_rst: reset {
> > > > +            compatible = "syna,sl261x-soc-reset";
> > > > +            #reset-cells = <1>;
> > > > +        };
> > > 
> > > Why can't this just be part of the parent node?
> > 
> > Good question! IMHO, the reasons are:
> > 
> > 1. the so called Gbls(Gbl means global) such as mcuGbl and
> > chip control Gbl contain not only reset, pinctrl and clk/plls
> > but also some other registers for different purposes, e.g in the
> > chip control gbl, there are some registers to control eth phy sel(RMII
> > or RGMII), and TXC 90 degree selection etc. These registers will
> > be used by the Ethernet driver;
> > 
> > While there's no pinctrl regs in avioGbl and vppGbl. In the vppgbl,
> > there are some registers for DPHYTX etc.
> > 
> > As can been, in different Gbls, there maybe different purposes registers, no
> > obvious patterns.
> > 
> > So the question here is why not split the gbl into different register
> > spaces, and abstract each space for each function group?
> 
> No, you misunderstood. I'm not asking for the register space to be
> split, I am asking why you cannot do
> | syscon@f7e1000 {
> | 	compatible = "syna,sl2610-soc-syscon", "syscon";
> | 	reg = <0xf7e10000 0x1000>;
> | 	#clock-cells = <1>;
> | 	#reset-cells = <1>;
> | };
> With this kind of thing, you can either use the auxiliary bus or
> MFD_CELL_NAME() to create the other devices. The latter is good if
> you've got a region with clocks that also has resets thrown in, the
> former is probably better if you have a syscon with a genuine
> assortment of different functions that you're going to be looking up via
> phandle.

I see your points now. But there are two problems with this solution:

1. Besides reset, pinctrl and clk, the so called "Gbl" also contains
registers for other purpose, e.g DPHYTX, when we need the driver for
DPHYTX, I assume I could also go with the auxiliary bus or MFD_CELL_NAME(),

2. where should the driver(s) and dt-bindings doc be put?
I assume drivers/soc/synaptics/ and
Documentation/devicetree/bindings/soc/synaptics?

Kindly correct me if I misunderstood something.

Thanks
> 
> > Two blocking
> > points make this impossible and ugly: different reg space may not be
> > aligned at 4KB boundary, e.g reset: 0x350 ~ 0x37c; eth phy sel:0xa18;
> > some registers which belong to similar functionality, e.g clks, may not be
> > adjacent.
> 
> This sort of thing, with non-adjacent clocks is still possible, you have
> a regmap just like before.
> 
> > So the best way is to abstract these GBLs via. "syscon", "simple-mfd", and
> > put necessary subnodes such as reset, pinctrl and clks under the GBL
> > node. If the registers in GBL is misc control, pass the syscon phandle
> > to the main driver, e.g I planed to pass the syscon phandle to GMAC
> > driver to control the RMII/RGMII selection.
> 
> And you can get the regmap from other drivers by doing phandle or
> compatible based lookups just like before.
> 
> > 
> > 2.follow current bg2 and bg2q style
> > e.g arch/arm/boot/dts/synaptics/berlin2q.dtsi
> > 
> > IMHO, why bg2 and bg2q dtsi files are implemented as current
> > style is due to the above point 1.
> 
> Unfortunately, this is the berlin2q devicetree showing its age, we've
> been pushing people away from nodes that exist just to probe a driver
> and don't provide any additional value. Something like
> | chip-control@f7e10000 {
> |     reg = <0xf7e10000 0x1000>;
> | 
> |     chip_rst: reset {
> |         compatible = "syna,sl261x-soc-reset";
> |         #reset-cells = <1>;
> |     };
> | 
> |     chip_clk: clock {
> |         compatible = "syna,sl261x-soc-clock";
> |         #clock-cells = <1>;
> |     };
> is exactly the pattern we are trying to avoid.
> 
> Cheers,
> Conor.
> 
> > 
> > Kindly let me know if there's a better solution.
> > 
> > > From a quick check of the dts, the node with this subnode didn't also
> > > have a system-reset node too, so there's no conflict or anything of that
> > > nature.
> > > 
> > > > +    };
> > > > diff --git a/include/dt-bindings/reset/syna,sl261x-reset.h b/include/dt-bindings/reset/syna,sl261x-reset.h
> > > > new file mode 100644
> > > > index 000000000000..7034f58aae5b
> > > > --- /dev/null
> > > > +++ b/include/dt-bindings/reset/syna,sl261x-reset.h
> > > > @@ -0,0 +1,80 @@
> > > > +/* SPDX-License-Identifier: (GPL-2.0 OR MIT) */
> > > > +/*
> > > > + * Copyright (C) 2026 Synaptics Incorporated
> > > > + *
> > > > + * Author: Jisheng Zhang <jszhang@kernel.org>
> > > > + */
> > > > +
> > > > +#ifndef _DT_BINDINGS_SL261X_RESET_H
> > > > +#define _DT_BINDINGS_SL261X_RESET_H
> > > > +
> > > > +/* ACPU subsystem */
> > > > +#define RST_SOC_SDIO0		0
> > > > +#define RST_SOC_USB0		1
> > > > +#define RST_SOC_EMMC		2
> > > > +#define RST_SOC_GETH0		3
> > > > +#define RST_SOC_SDIO1		4
> > > > +#define RST_SOC_USB1		5
> > > > +#define RST_SOC_GETH1		6
> > > > +#define RST_SOC_USB0PHY		7
> > > > +#define RST_SOC_USB0CORE	8
> > > > +#define RST_SOC_USB0MAHB	9
> > > > +#define RST_SOC_USB1PHY		10
> > > > +#define RST_SOC_USB1CORE	11
> > > > +#define RST_SOC_USB1MAHB	12
> > > > +#define RST_SOC_UART0		13
> > > > +#define RST_SOC_UART1		14
> > > > +#define RST_SOC_UART2		15
> > > > +#define RST_SOC_UART3		16
> > > > +#define RST_SOC_I2C0		17
> > > > +#define RST_SOC_I2C1		18
> > > > +#define RST_SOC_SPI0		19
> > > > +#define RST_SOC_SPI1		20
> > > > +#define RST_SOC_SPI2		21
> > > > +#define RST_SOC_SPI3		22
> > > > +#define RST_SOC_APBTIMERS	23
> > > > +#define RST_SOC_APBSYSCNT	24
> > > > +#define RST_SOC_APBWDT		25
> > > > +#define RST_SOC_APBGPIO		26
> > > > +#define RST_SOC_APBDMA		27
> > > > +#define RST_SOC_GPUCORE		28
> > > > +#define RST_SOC_NPUCORE		29
> > > > +#define RST_SOC_AVIOAIOG	30
> > > > +#define RST_SOC_AVIOVPPG	31
> > > > +#define RST_SOC_AVIOVIPG	32
> > > > +
> > > > +/* system subsystem */
> > > > +#define RST_SM_ADCCORE		0
> > > > +#define RST_SM_ADCPRST		1
> > > > +#define RST_SM_CAN0PRST		2
> > > > +#define RST_SM_CAN0SRST		3
> > > > +#define RST_SM_CAN1PRST		4
> > > > +#define RST_SM_CAN1SRST		5
> > > > +#define RST_SM_GPIOPRST		6
> > > > +#define RST_SM_GPIOSRST		7
> > > > +#define RST_SM_I2CM0PRST	8
> > > > +#define RST_SM_I2CM0SRST	9
> > > > +#define RST_SM_I2CM1PRST	10
> > > > +#define RST_SM_I2CM1SRST	11
> > > > +#define RST_SM_I3CPRST		12
> > > > +#define RST_SM_I3CSRST		13
> > > > +#define RST_SM_PDMPRST		14
> > > > +#define RST_SM_PDMSRST		15
> > > > +#define RST_SM_PVTHSRST		16
> > > > +#define RST_SM_PVTPRST		17
> > > > +#define RST_SM_PWMPERIRST	18
> > > > +#define RST_SM_PWMPRST		19
> > > > +#define RST_SM_SPIMPRST		20
> > > > +#define RST_SM_SPIMSRST		21
> > > > +#define RST_SM_SPISPRST		22
> > > > +#define RST_SM_SPISSRST		23
> > > > +#define RST_SM_UART0PRST	24
> > > > +#define RST_SM_UART0SRST	25
> > > > +#define RST_SM_UART1PRST	26
> > > > +#define RST_SM_UART1SRST	27
> > > > +#define RST_SM_UART2PRST	28
> > > > +#define RST_SM_UART2SRST	29
> > > > +#define RST_SM_UART3PRST	30
> > > > +#define RST_SM_UART3SRST	31
> > > > +
> > > > +#endif /* _DT_BINDINGS_SL261X_RESET_H */
> > > > -- 
> > > > 2.53.0
> > > > 
> > 
> >