[PATCH v2 2/2] reset: stm32mp1: Enable stm32mp1 reset driver

[Philipp Zabel]

Hi Gabriel,

this looks mostly good to me, a few questions and comments below:

On Wed, 2018-03-14 at 17:30 +0100, gabriel.fernandez at st.com wrote:
> From: Gabriel Fernandez <gabriel.fernandez at st.com>
> 
> stm32mp1 RCC IP 1 has a reset SET register and a reset CLEAR register.
> 
> Writing '0' on reset SET register has no effect
> Writing '1' on reset SET register
> 	activates the reset of the corresponding peripheral
> 
> Writing '0' on reset CLEAR register	has no effect
> Writing '1' on reset CLEAR register
> 	releases the reset of the corresponding peripheral
> 
> See Documentation/devicetree/bindings/clock/st,stm32mp1-rcc.txt
> 
> Signed-off-by: Gabriel Fernandez <gabriel.fernandez at st.com>
> ---
>  drivers/reset/Kconfig          |   6 ++
>  drivers/reset/Makefile         |   1 +
>  drivers/reset/reset-stm32mp1.c | 122 +++++++++++++++++++++++++++++++++++++++++
>  3 files changed, 129 insertions(+)
>  create mode 100644 drivers/reset/reset-stm32mp1.c
> 
> diff --git a/drivers/reset/Kconfig b/drivers/reset/Kconfig
> index 1efbc6c..c0b292b 100644
> --- a/drivers/reset/Kconfig
> +++ b/drivers/reset/Kconfig
> @@ -97,6 +97,12 @@ config RESET_SIMPLE
>  	   - Allwinner SoCs
>  	   - ZTE's zx2967 family
>  
> +config RESET_STM32MP157
> +	bool "STM32MP157 Reset Driver" if COMPILE_TEST
> +	default MACH_STM32MP157
> +	help
> +	  This enables the RCC reset controller driver for STM32 MPUs.
> +
>  config RESET_SUNXI
>  	bool "Allwinner SoCs Reset Driver" if COMPILE_TEST && !ARCH_SUNXI
>  	default ARCH_SUNXI
> diff --git a/drivers/reset/Makefile b/drivers/reset/Makefile
> index 132c24f..c1261dc 100644
> --- a/drivers/reset/Makefile
> +++ b/drivers/reset/Makefile
> @@ -15,6 +15,7 @@ obj-$(CONFIG_RESET_MESON) += reset-meson.o
>  obj-$(CONFIG_RESET_OXNAS) += reset-oxnas.o
>  obj-$(CONFIG_RESET_PISTACHIO) += reset-pistachio.o
>  obj-$(CONFIG_RESET_SIMPLE) += reset-simple.o
> +obj-$(CONFIG_RESET_STM32MP157) += reset-stm32mp1.o
>  obj-$(CONFIG_RESET_SUNXI) += reset-sunxi.o
>  obj-$(CONFIG_RESET_TI_SCI) += reset-ti-sci.o
>  obj-$(CONFIG_RESET_TI_SYSCON) += reset-ti-syscon.o
> diff --git a/drivers/reset/reset-stm32mp1.c b/drivers/reset/reset-stm32mp1.c
> new file mode 100644
> index 0000000..5e25388
> --- /dev/null
> +++ b/drivers/reset/reset-stm32mp1.c
> @@ -0,0 +1,122 @@
> +// SPDX-License-Identifier: GPL-2.0
> +/*
> + * Copyright (C) STMicroelectronics 2018 - All Rights Reserved
> + * Author: Gabriel Fernandez <gabriel.fernandez at st.com> for STMicroelectronics.
> + */
> +
> +#include <linux/arm-smccc.h>

This does not seem to be necessary.

> +#include <linux/device.h>
> +#include <linux/err.h>
> +#include <linux/of.h>

> +#include <linux/of_device.h>

This does not seem to be necessary either.

> +#include <linux/platform_device.h>
> +#include <linux/reset-controller.h>
> +
> +#define CLR_OFFSET 0x4
> +
> +struct stm32_reset_data {
> +	struct reset_controller_dev	rcdev;
> +	void __iomem			*membase;
> +};
> +
> +static inline struct stm32_reset_data *
> +to_stm32_reset_data(struct reset_controller_dev *rcdev)
> +{
> +	return container_of(rcdev, struct stm32_reset_data, rcdev);
> +}
> +
> +static int stm32_reset_update(struct reset_controller_dev *rcdev,
> +			      unsigned long id, bool assert)
> +{
> +	struct stm32_reset_data *data = to_stm32_reset_data(rcdev);
> +	int reg_width = sizeof(u32);
> +	int bank = id / (reg_width * BITS_PER_BYTE);
> +	int offset = id % (reg_width * BITS_PER_BYTE);
> +	void __iomem *addr;
> +
> +	addr = data->membase + (bank * reg_width);
> +	if (!assert)
> +		addr += CLR_OFFSET;
> +
> +	writel(BIT(offset), addr);
> +
> +	return 0;
> +}
> +
> +static int stm32_reset_assert(struct reset_controller_dev *rcdev,
> +			      unsigned long id)
> +{
> +	return stm32_reset_update(rcdev, id, true);
> +}
> +
> +static int stm32_reset_deassert(struct reset_controller_dev *rcdev,
> +				unsigned long id)
> +{
> +	return stm32_reset_update(rcdev, id, false);
> +}
> +
> +static int stm32_reset_status(struct reset_controller_dev *rcdev,
> +			      unsigned long id)
> +{
> +	struct stm32_reset_data *data = to_stm32_reset_data(rcdev);
> +	int reg_width = sizeof(u32);
> +	int bank = id / (reg_width * BITS_PER_BYTE);
> +	int offset = id % (reg_width * BITS_PER_BYTE);
> +	u32 reg;
> +
> +	reg = readl(data->membase + (bank * reg_width));
> +
> +	return !(reg & BIT(offset));
> +}

So the SET register can be read back and returns 0 for reset lines that
are currently asserted and returns 1 for reset lines that are currently
deasserted?

> +
> +const struct reset_control_ops stm32_reset_ops = {
> +	.assert		= stm32_reset_assert,
> +	.deassert	= stm32_reset_deassert,
> +	.status		= stm32_reset_status,
> +};
> +
> +static const struct of_device_id stm32_reset_dt_ids[] = {
> +	{ .compatible = "st,stm32mp1-rcc"},
> +	{ /* sentinel */ },
> +};

>From the DT bindings it looks like the clock and reset drivers are
sharing the same node. Is there just no clock platform_driver at all?

> +
> +static int stm32_reset_probe(struct platform_device *pdev)
> +{
> +	struct device *dev = &pdev->dev;
> +	struct stm32_reset_data *data;
> +	void __iomem *membase;
> +	struct resource *res;
> +
> +	data = devm_kzalloc(dev, sizeof(*data), GFP_KERNEL);
> +	if (!data)
> +		return -ENOMEM;
> +
> +	res = platform_get_resource(pdev, IORESOURCE_MEM, 0);
> +	membase = devm_ioremap_resource(dev, res);
> +	if (IS_ERR(membase))
> +		return PTR_ERR(membase);
> +
> +	data->membase = membase;
> +	data->rcdev.owner = THIS_MODULE;
> +	data->rcdev.nr_resets = resource_size(res) * BITS_PER_BYTE;
> +	data->rcdev.ops = &stm32_reset_ops;
> +	data->rcdev.of_node = dev->of_node;
> +
> +	return devm_reset_controller_register(dev, &data->rcdev);
> +}
> +
> +static struct platform_driver stm32_reset_driver = {
> +	.probe	= stm32_reset_probe,
> +	.driver = {
> +		.name		= "stm32mp1-reset",
> +		.of_match_table	= stm32_reset_dt_ids,
> +	},
> +};
> +
> +static int __init stm32_reset_init(void)
> +{
> +	return platform_driver_register(&stm32_reset_driver);
> +}
> +
> +postcore_initcall(stm32_reset_init);

Isn't builtin_platform_driver early enough?

regards
Philipp