[PATCH 0/2] gpio: i.MX: make _get_value() able to read output pin state

[PATCH 0/2] gpio: i.MX: make _get_value() able to read output pin state

[Maud Spierings via B4 Relay]

Make _get_value() read back the dr register instead of the psr register
when the direction is set to output. Now _get_value() can actually read
the state of an output gpio instead of being always low.

Also add a new empty migration guide file.

Signed-off-by: Maud Spierings <maud_spierings@hotmail.com>
---
Maud Spierings (2):
      gpio: i.MX: make _get_value() able to read output pin state
      Documentation: migration-2025.11.0: add new empty file

 .../migration-guides/migration-2025.11.0.rst          |  0
 drivers/gpio/gpio-imx.c                               | 19 +++++++++++--------
 2 files changed, 11 insertions(+), 8 deletions(-)
---
base-commit: 8defba1d0ab1aef9dd5d57710e18d0d02e2c48e2
change-id: 20251015-imx_gpio-46efe1e7abd3

Best regards,
-- 
Maud Spierings <maud_spierings@hotmail.com>

[PATCH 1/2] gpio: i.MX: make _get_value() able to read output pin state

[Maud Spierings via B4 Relay]

From: Maud Spierings <maud_spierings@hotmail.com>

Make _get_value() read back the dr register instead of the psr register
when the direction is set to output. Now _get_value() can actually read
the state of an output gpio instead of being always low.

Signed-off-by: Maud Spierings <maud_spierings@hotmail.com>

---
inspired by similar logic being used in u-boot and Linux
---
 drivers/gpio/gpio-imx.c | 19 +++++++++++--------
 1 file changed, 11 insertions(+), 8 deletions(-)

diff --git a/drivers/gpio/gpio-imx.c b/drivers/gpio/gpio-imx.c
index d9499c66cd..bb14c56399 100644
--- a/drivers/gpio/gpio-imx.c
+++ b/drivers/gpio/gpio-imx.c
@@ -89,24 +89,27 @@ static int imx_gpio_direction_output(struct gpio_chip *chip, unsigned gpio, int
 	return 0;
 }
 
-static int imx_gpio_get_value(struct gpio_chip *chip, unsigned gpio)
+static int imx_get_direction(struct gpio_chip *chip, unsigned offset)
 {
 	struct imx_gpio_chip *imxgpio = container_of(chip, struct imx_gpio_chip, chip);
 	void __iomem *base = imxgpio->base;
-	u32 val;
-
-	val = readl(base + imxgpio->regs->psr);
+	u32 val = readl(base + imxgpio->regs->gdir);
 
-	return val & (1 << gpio) ? 1 : 0;
+	return (val & (1 << offset)) ? GPIOF_DIR_OUT : GPIOF_DIR_IN;
 }
 
-static int imx_get_direction(struct gpio_chip *chip, unsigned offset)
+static int imx_gpio_get_value(struct gpio_chip *chip, unsigned gpio)
 {
 	struct imx_gpio_chip *imxgpio = container_of(chip, struct imx_gpio_chip, chip);
 	void __iomem *base = imxgpio->base;
-	u32 val = readl(base + imxgpio->regs->gdir);
+	u32 val;
 
-	return (val & (1 << offset)) ? GPIOF_DIR_OUT : GPIOF_DIR_IN;
+	if (imx_get_direction(chip, gpio) == GPIOF_DIR_IN)
+		val = readl(base + imxgpio->regs->psr);
+	else
+		val = readl(base + imxgpio->regs->dr);
+
+	return val & (1 << gpio) ? 1 : 0;
 }
 
 static struct gpio_ops imx_gpio_ops = {

-- 
2.51.0

[PATCH 2/2] Documentation: migration-2025.11.0: add new empty file

[Maud Spierings via B4 Relay]

From: Maud Spierings <maud_spierings@hotmail.com>

Add an empty file to hold any future migration notes.

Signed-off-by: Maud Spierings <maud_spierings@hotmail.com>
---
 Documentation/migration-guides/migration-2025.11.0.rst | 0
 1 file changed, 0 insertions(+), 0 deletions(-)

diff --git a/Documentation/migration-guides/migration-2025.11.0.rst b/Documentation/migration-guides/migration-2025.11.0.rst
new file mode 100644
index 0000000000..e69de29bb2

-- 
2.51.0

Re: [PATCH 1/2] gpio: i.MX: make _get_value() able to read output pin state

[Ahmad Fatoum]

On 10/15/25 8:03 PM, Maud Spierings via B4 Relay wrote:
> From: Maud Spierings <maud_spierings@hotmail.com>
> 
> Make _get_value() read back the dr register instead of the psr register
> when the direction is set to output. Now _get_value() can actually read
> the state of an output gpio instead of being always low.
> 
> Signed-off-by: Maud Spierings <maud_spierings@hotmail.com>

Reviewed-by: Ahmad Fatoum <a.fatoum@pengutronix.de>

> 
> ---
> inspired by similar logic being used in u-boot and Linux

That's useful info that should go into the commit message IMO, because
it explains why we think regression should be limited.

> ---
>  drivers/gpio/gpio-imx.c | 19 +++++++++++--------
>  1 file changed, 11 insertions(+), 8 deletions(-)
> 
> diff --git a/drivers/gpio/gpio-imx.c b/drivers/gpio/gpio-imx.c
> index d9499c66cd..bb14c56399 100644
> --- a/drivers/gpio/gpio-imx.c
> +++ b/drivers/gpio/gpio-imx.c
> @@ -89,24 +89,27 @@ static int imx_gpio_direction_output(struct gpio_chip *chip, unsigned gpio, int
>  	return 0;
>  }
>  
> -static int imx_gpio_get_value(struct gpio_chip *chip, unsigned gpio)
> +static int imx_get_direction(struct gpio_chip *chip, unsigned offset)
>  {
>  	struct imx_gpio_chip *imxgpio = container_of(chip, struct imx_gpio_chip, chip);
>  	void __iomem *base = imxgpio->base;
> -	u32 val;
> -
> -	val = readl(base + imxgpio->regs->psr);
> +	u32 val = readl(base + imxgpio->regs->gdir);
>  
> -	return val & (1 << gpio) ? 1 : 0;
> +	return (val & (1 << offset)) ? GPIOF_DIR_OUT : GPIOF_DIR_IN;
>  }
>  
> -static int imx_get_direction(struct gpio_chip *chip, unsigned offset)
> +static int imx_gpio_get_value(struct gpio_chip *chip, unsigned gpio)
>  {
>  	struct imx_gpio_chip *imxgpio = container_of(chip, struct imx_gpio_chip, chip);
>  	void __iomem *base = imxgpio->base;
> -	u32 val = readl(base + imxgpio->regs->gdir);
> +	u32 val;
>  
> -	return (val & (1 << offset)) ? GPIOF_DIR_OUT : GPIOF_DIR_IN;
> +	if (imx_get_direction(chip, gpio) == GPIOF_DIR_IN)
> +		val = readl(base + imxgpio->regs->psr);
> +	else
> +		val = readl(base + imxgpio->regs->dr);
> +
> +	return val & (1 << gpio) ? 1 : 0;
>  }
>  
>  static struct gpio_ops imx_gpio_ops = {
> 

-- 
Pengutronix e.K.                  |                             |
Steuerwalder Str. 21              | http://www.pengutronix.de/  |
31137 Hildesheim, Germany         | Phone: +49-5121-206917-0    |
Amtsgericht Hildesheim, HRA 2686  | Fax:   +49-5121-206917-5555 |

Re: [PATCH 2/2] Documentation: migration-2025.11.0: add new empty file

[Ahmad Fatoum]

On 10/15/25 8:03 PM, Maud Spierings via B4 Relay wrote:
> From: Maud Spierings <maud_spierings@hotmail.com>
> 
> Add an empty file to hold any future migration notes.
> 
> Signed-off-by: Maud Spierings <maud_spierings@hotmail.com>
> ---
>  Documentation/migration-guides/migration-2025.11.0.rst | 0
>  1 file changed, 0 insertions(+), 0 deletions(-)
> 
> diff --git a/Documentation/migration-guides/migration-2025.11.0.rst b/Documentation/migration-guides/migration-2025.11.0.rst

Sorry, I should have been clearer; I am a bit wary of breaking scripts
or board code that do SION shenanigans, so my suggestion to add a note.

Something like:

i.MX GPIOs
----------

Reading output GPIOs now returns the configured output level instead
of reading back the input register. This aligns us with what Linux
is doing, but may falsify readings of single-ended GPIOs that have
the SION bit configured.

> new file mode 100644
> index 0000000000..e69de29bb2
> 

-- 
Pengutronix e.K.                  |                             |
Steuerwalder Str. 21              | http://www.pengutronix.de/  |
31137 Hildesheim, Germany         | Phone: +49-5121-206917-0    |
Amtsgericht Hildesheim, HRA 2686  | Fax:   +49-5121-206917-5555 |

Re: [PATCH 2/2] Documentation: migration-2025.11.0: add new empty file

[Maud Spierings]

On 10/16/25 14:58, Ahmad Fatoum wrote:
> 
> 
> On 10/15/25 8:03 PM, Maud Spierings via B4 Relay wrote:
>> From: Maud Spierings <maud_spierings@hotmail.com>
>>
>> Add an empty file to hold any future migration notes.
>>
>> Signed-off-by: Maud Spierings <maud_spierings@hotmail.com>
>> ---
>>   Documentation/migration-guides/migration-2025.11.0.rst | 0
>>   1 file changed, 0 insertions(+), 0 deletions(-)
>>
>> diff --git a/Documentation/migration-guides/migration-2025.11.0.rst b/Documentation/migration-guides/migration-2025.11.0.rst
> 
> Sorry, I should have been clearer; I am a bit wary of breaking scripts
> or board code that do SION shenanigans, so my suggestion to add a note.
> 
> Something like:
> 
> i.MX GPIOs
> ----------
> 
> Reading output GPIOs now returns the configured output level instead
> of reading back the input register. This aligns us with what Linux
> is doing, but may falsify readings of single-ended GPIOs that have
> the SION bit configured.
> 

That makes a lot more sense, I'll send a V2 also fixing the commit on 
[1/2] this evening.

Kind regards,
Maud