Re: [PATCH v6 2/3] pinctrl: qcom: add support for complementary reserved gpios

From: Bjorn Andersson <andersson@kernel.org>
To: Richard Acayan <mailingradian@gmail.com>
Cc: linux-arm-msm@vger.kernel.org, Andy Gross <agross@kernel.org>,
	Konrad Dybcio <konrad.dybcio@somainline.org>,
	Linus Walleij <linus.walleij@linaro.org>,
	Rob Herring <robh+dt@kernel.org>,
	Krzysztof Kozlowski <krzysztof.kozlowski+dt@linaro.org>,
	linux-gpio@vger.kernel.org, devicetree@vger.kernel.org,
	linux-kernel@vger.kernel.org,
	~postmarketos/upstreaming@lists.sr.ht
Subject: Re: [PATCH v6 2/3] pinctrl: qcom: add support for complementary reserved gpios
Date: Mon, 3 Oct 2022 19:17:48 -0500	[thread overview]
Message-ID: <20221004001748.qohdzdkmdofo35wc@baldur> (raw)
In-Reply-To: <20221001210725.60967-3-mailingradian@gmail.com>

On Sat, Oct 01, 2022 at 05:07:24PM -0400, Richard Acayan wrote:
> The driver-provided list of reserved gpios normally overrides any valid
> ranges provided by the firmware (device tree and ACPI). When the driver
> defines dummy pingroups by itself, it should mark these as invalid but
> should not prevent the firmware from specifying more reserved gpios. Let
> pinctrl drivers indicate that the reserved gpios list complements instead
> of overrides other lists from firmware.
> 
> Signed-off-by: Richard Acayan <mailingradian@gmail.com>
> ---
>  drivers/pinctrl/qcom/pinctrl-msm.c | 5 +++--
>  drivers/pinctrl/qcom/pinctrl-msm.h | 4 ++++
>  2 files changed, 7 insertions(+), 2 deletions(-)
> 
> diff --git a/drivers/pinctrl/qcom/pinctrl-msm.c b/drivers/pinctrl/qcom/pinctrl-msm.c
> index a2abfe987ab1..cea1d2af8c88 100644
> --- a/drivers/pinctrl/qcom/pinctrl-msm.c
> +++ b/drivers/pinctrl/qcom/pinctrl-msm.c
> @@ -687,9 +687,10 @@ static int msm_gpio_init_valid_mask(struct gpio_chip *gc,
>  	const int *reserved = pctrl->soc->reserved_gpios;
>  	u16 *tmp;
>  
> -	/* Driver provided reserved list overrides DT and ACPI */
> +	/* Driver provided reserved list overrides DT and ACPI by default */
>  	if (reserved) {
> -		bitmap_fill(valid_mask, ngpios);
> +		if (!pctrl->soc->complement_fw_gpio_ranges)

reserved_gpios is only defined for ACPI drivers and afaict there's
nothing in the ACPI path that would modify the valid_mask after the
bitmap is being filled in gpiochip_allocate_mask().

If that's the case it seems reasonable that we can just drop the
bitmap_fill() here. But perhaps I'm missing something?

Regards,
Bjorn

> +			bitmap_fill(valid_mask, ngpios);
>  		for (i = 0; reserved[i] >= 0; i++) {
>  			if (i >= ngpios || reserved[i] >= ngpios) {
>  				dev_err(pctrl->dev, "invalid list of reserved GPIOs\n");
> diff --git a/drivers/pinctrl/qcom/pinctrl-msm.h b/drivers/pinctrl/qcom/pinctrl-msm.h
> index dd0d949f7a9e..734fe7b2a472 100644
> --- a/drivers/pinctrl/qcom/pinctrl-msm.h
> +++ b/drivers/pinctrl/qcom/pinctrl-msm.h
> @@ -128,6 +128,9 @@ struct msm_gpio_wakeirq_map {
>   *              function number for eGPIO and any time we see that function
>   *              number used we'll treat it as a request to mux away from
>   *              our TLMM towards another owner.
> + * @complement_fw_gpio_ranges: If true, the reserved gpios list from the
> + *                             driver will not override the reserved gpios
> + *                             list from the firmware.
>   */
>  struct msm_pinctrl_soc_data {
>  	const struct pinctrl_pin_desc *pins;
> @@ -146,6 +149,7 @@ struct msm_pinctrl_soc_data {
>  	bool wakeirq_dual_edge_errata;
>  	unsigned int gpio_func;
>  	unsigned int egpio_func;
> +	bool complement_fw_gpio_ranges;
>  };
>  
>  extern const struct dev_pm_ops msm_pinctrl_dev_pm_ops;
> -- 
> 2.37.3
> 