Re: [PATCH v3 04/21] pinctrl: starfive: Add StarFive JHB100 sys0 controller driver

[Philipp Zabel]

On Di, 2026-06-02 at 22:53 -0700, Changhuang Liang wrote:
> Add pinctrl driver for StarFive JHB100 SoC System-0(sys0) pinctrl
> controller.
> 
> Co-developed-by: Lianfeng Ouyang <lianfeng.ouyang@xxxxxxxxxxxxxxxx>
> Signed-off-by: Lianfeng Ouyang <lianfeng.ouyang@xxxxxxxxxxxxxxxx>
> Signed-off-by: Changhuang Liang <changhuang.liang@xxxxxxxxxxxxxxxx>
> ---
[...]
> diff --git a/drivers/pinctrl/starfive/pinctrl-starfive-jhb100.c b/drivers/pinctrl/starfive/pinctrl-starfive-jhb100.c
> new file mode 100644
> index 000000000000..52b97870b991
> --- /dev/null
> +++ b/drivers/pinctrl/starfive/pinctrl-starfive-jhb100.c
> @@ -0,0 +1,1607 @@
[...]
> +int jhb100_pinctrl_probe(struct platform_device *pdev)
> +{
> +	struct device *dev = &pdev->dev;
> +	struct gpio_irq_chip *girq;
> +	const struct jhb100_pinctrl_domain_info *info;
> +	struct jhb100_pinctrl *sfp;
> +	struct pinctrl_desc *jhb100_pinctrl_desc;
> +	struct starfive_pinctrl_regs *pinctrl_regs;
> +	struct reset_control *rst;
> +	struct clk *clk;
> +	int ret;
> +	int irq;
> +
> +	info = of_device_get_match_data(&pdev->dev);
> +	if (!info)
> +		return -ENODEV;
> +
> +	pinctrl_regs = info->regs;
> +
> +	sfp = devm_kzalloc(dev, sizeof(*sfp), GFP_KERNEL);
> +	if (!sfp)
> +		return -ENOMEM;
> +
> +	sfp->base = devm_platform_ioremap_resource(pdev, 0);
> +	if (IS_ERR(sfp->base))
> +		return PTR_ERR(sfp->base);
> +
> +	clk = devm_clk_get_optional_enabled(dev, NULL);
> +	if (IS_ERR(clk))
> +		return dev_err_probe(dev, PTR_ERR(clk), "could not get & enable clock\n");
> +
> +	rst = devm_reset_control_array_get_optional_shared(dev);

Why is this shared? Are there multiple pinctrl controllers sharing
reset lines?

> +	if (IS_ERR(rst))
> +		return dev_err_probe(dev, PTR_ERR(rst), "could not get reset control\n");
> +
> +	/*
> +	 * we don't want to assert reset and risk undoing pin muxing for the
> +	 * early boot serial console, but let's make sure the reset line is
> +	 * deasserted in case someone runs a really minimal bootloader.
> +	 */
> +	ret = reset_control_deassert(rst);
> +	if (ret)
> +		return dev_err_probe(dev, ret, "could not deassert reset\n");

Missing reset_control_assert() in .remove will unbalance the
deassertion counter if a pin controller with shared reset is
unbound/rebound, causing the reset to never be asserted again.

regards
Philipp