diff mbox series

[2/2] gpio: rockchip: get pinctrl node from 'gpio-ranges' property

Message ID 20220303062211.1378883-3-jay.xu@rock-chips.com (mailing list archive)
State New, archived
Headers show
Series gpio-rochchip | expand

Commit Message

Jianqun Xu March 3, 2022, 6:22 a.m. UTC
The dt nodes for rockchip soc designes as

	pinctrl: pinctrl {
		gpio {
			gpio-ranges = <&pinctrl xxx>;
		};
	};

Currently, we get the pinctrl dt node from parent of gpio, this patch
try to get pinctrl dt node from 'gpio-ranges' property.

After this patch, the dt nodes possible to be

	gpio {
		gpio-ranges = <&pinctrl xxx>;
	};

	pinctrl: pinctrl {

	};

then the gpio driver could register as platform device itself, but not
populate from pinctrl driver.

Signed-off-by: Jianqun Xu <jay.xu@rock-chips.com>
---
 drivers/gpio/gpio-rockchip.c | 10 +++++++---
 1 file changed, 7 insertions(+), 3 deletions(-)

Comments

Heiko Stübner March 3, 2022, 11:40 a.m. UTC | #1
Hi Jianqun,

Am Donnerstag, 3. März 2022, 07:22:11 CET schrieb Jianqun Xu:
> The dt nodes for rockchip soc designes as
> 
> 	pinctrl: pinctrl {
> 		gpio {
> 			gpio-ranges = <&pinctrl xxx>;
> 		};
> 	};
> 
> Currently, we get the pinctrl dt node from parent of gpio, this patch
> try to get pinctrl dt node from 'gpio-ranges' property.
> 
> After this patch, the dt nodes possible to be
> 
> 	gpio {
> 		gpio-ranges = <&pinctrl xxx>;
> 	};
> 
> 	pinctrl: pinctrl {
> 
> 	};
> 
> then the gpio driver could register as platform device itself, but not
> populate from pinctrl driver.

The change looks interesting, as it would solve that long-standing
design-issue I "created" back in 2013 ;-) .

Though you need to keep some things in mind:

(1) Such a change should be reflected in the devicetree binding
    as it involves a different form of nodes and introduces.

    Looking at the binding description, using gpio-ranges to map
    to specific pinctrl pins really seems to be a valid use for this.


(2) Keep things backwards compatible.
    Old devicetrees should stay working with new kernel versions

    A common pattern is to try the new approach and if that fails
    try the "deprecated" method, which should be nicely doable
    when looking at the code change below.


Heiko

> Signed-off-by: Jianqun Xu <jay.xu@rock-chips.com>
> ---
>  drivers/gpio/gpio-rockchip.c | 10 +++++++---
>  1 file changed, 7 insertions(+), 3 deletions(-)
> 
> diff --git a/drivers/gpio/gpio-rockchip.c b/drivers/gpio/gpio-rockchip.c
> index 1da0324445cc..46c54dff92db 100644
> --- a/drivers/gpio/gpio-rockchip.c
> +++ b/drivers/gpio/gpio-rockchip.c
> @@ -690,6 +690,9 @@ rockchip_gpio_find_bank(struct pinctrl_dev *pctldev, int id)
>  	int i, found = 0;
>  
>  	info = pinctrl_dev_get_drvdata(pctldev);
> +	if (!info)
> +		return NULL;
> +
>  	bank = info->ctrl->pin_banks;
>  	for (i = 0; i < info->ctrl->nr_banks; i++, bank++) {
>  		if (bank->bank_num == id) {
> @@ -705,15 +708,16 @@ static int rockchip_gpio_probe(struct platform_device *pdev)
>  {
>  	struct device *dev = &pdev->dev;
>  	struct device_node *np = dev->of_node;
> -	struct device_node *pctlnp = of_get_parent(np);
> +	struct device_node *pctlnp = NULL;
>  	struct pinctrl_dev *pctldev = NULL;
>  	struct rockchip_pin_bank *bank = NULL;
>  	struct rockchip_pin_output_deferred *cfg;
>  	static int gpio;
>  	int id, ret;
>  
> -	if (!np || !pctlnp)
> -		return -ENODEV;
> +	pctlnp = of_parse_phandle(np, "gpio-ranges", 0);
> +	if (!pctlnp)
> +		pctlnp = of_get_parent(np);
>  
>  	pctldev = of_pinctrl_get(pctlnp);
>  	if (!pctldev)
>
Jianqun Xu March 4, 2022, 3:02 a.m. UTC | #2
Hi Heiko:

--------------
jay.xu@rock-chips.com
>Hi Jianqun,
>
>Am Donnerstag, 3. März 2022, 07:22:11 CET schrieb Jianqun Xu:
>> The dt nodes for rockchip soc designes as
>>
>> pinctrl: pinctrl {
>> gpio {
>> gpio-ranges = <&pinctrl xxx>;
>> };
>> };
>>
>> Currently, we get the pinctrl dt node from parent of gpio, this patch
>> try to get pinctrl dt node from 'gpio-ranges' property.
>>
>> After this patch, the dt nodes possible to be
>>
>> gpio {
>> gpio-ranges = <&pinctrl xxx>;
>> };
>>
>> pinctrl: pinctrl {
>>
>> };
>>
>> then the gpio driver could register as platform device itself, but not
>> populate from pinctrl driver.
>
>The change looks interesting, as it would solve that long-standing
>design-issue I "created" back in 2013 ;-) .
>
>Though you need to keep some things in mind:
>
>(1) Such a change should be reflected in the devicetree binding
>    as it involves a different form of nodes and introduces.
>
>    Looking at the binding description, using gpio-ranges to map
>    to specific pinctrl pins really seems to be a valid use for this.
> 
thanks, i will add the fix in next patchset by adding another patch.

>
>(2) Keep things backwards compatible.
>    Old devicetrees should stay working with new kernel versions
>
>    A common pattern is to try the new approach and if that fails
>    try the "deprecated" method, which should be nicely doable
>    when looking at the code change below.
> 
okay, i will test this case.
>
>Heiko
>
>> Signed-off-by: Jianqun Xu <jay.xu@rock-chips.com>
>> ---
>>  drivers/gpio/gpio-rockchip.c | 10 +++++++---
>>  1 file changed, 7 insertions(+), 3 deletions(-)
>>
>> diff --git a/drivers/gpio/gpio-rockchip.c b/drivers/gpio/gpio-rockchip.c
>> index 1da0324445cc..46c54dff92db 100644
>> --- a/drivers/gpio/gpio-rockchip.c
>> +++ b/drivers/gpio/gpio-rockchip.c
>> @@ -690,6 +690,9 @@ rockchip_gpio_find_bank(struct pinctrl_dev *pctldev, int id)
>>  int i, found = 0;
>> 
>>  info = pinctrl_dev_get_drvdata(pctldev);
>> +	if (!info)
>> +	return NULL;
>> +
>>  bank = info->ctrl->pin_banks;
>>  for (i = 0; i < info->ctrl->nr_banks; i++, bank++) {
>>  if (bank->bank_num == id) {
>> @@ -705,15 +708,16 @@ static int rockchip_gpio_probe(struct platform_device *pdev)
>>  {
>>  struct device *dev = &pdev->dev;
>>  struct device_node *np = dev->of_node;
>> -	struct device_node *pctlnp = of_get_parent(np);
>> +	struct device_node *pctlnp = NULL;
>>  struct pinctrl_dev *pctldev = NULL;
>>  struct rockchip_pin_bank *bank = NULL;
>>  struct rockchip_pin_output_deferred *cfg;
>>  static int gpio;
>>  int id, ret;
>> 
>> -	if (!np || !pctlnp)
>> -	return -ENODEV;
>> +	pctlnp = of_parse_phandle(np, "gpio-ranges", 0);
>> +	if (!pctlnp)
>> +	pctlnp = of_get_parent(np);
>> 
>>  pctldev = of_pinctrl_get(pctlnp);
>>  if (!pctldev)
>>
>
>
>
>
Jianqun Xu March 7, 2022, 8:08 a.m. UTC | #3
--------------
jay.xu@rock-chips.com
>Hi Jianqun,
>
>Am Donnerstag, 3. März 2022, 07:22:11 CET schrieb Jianqun Xu:
>> The dt nodes for rockchip soc designes as
>>
>> pinctrl: pinctrl {
>> gpio {
>> gpio-ranges = <&pinctrl xxx>;
>> };
>> };
>>
>> Currently, we get the pinctrl dt node from parent of gpio, this patch
>> try to get pinctrl dt node from 'gpio-ranges' property.
>>
>> After this patch, the dt nodes possible to be
>>
>> gpio {
>> gpio-ranges = <&pinctrl xxx>;
>> };
>>
>> pinctrl: pinctrl {
>>
>> };
>>
>> then the gpio driver could register as platform device itself, but not
>> populate from pinctrl driver.
>
>The change looks interesting, as it would solve that long-standing
>design-issue I "created" back in 2013 ;-) .
>
>Though you need to keep some things in mind:
>
>(1) Such a change should be reflected in the devicetree binding
>    as it involves a different form of nodes and introduces.
>
>    Looking at the binding description, using gpio-ranges to map
>    to specific pinctrl pins really seems to be a valid use for this.
>
>
>(2) Keep things backwards compatible.
>    Old devicetrees should stay working with new kernel versions
>
>    A common pattern is to try the new approach and if that fails
>    try the "deprecated" method, which should be nicely doable
>    when looking at the code change below.
> 
Hi heiko

could i get pinctrl device from name ? like this:

pctldev = get_pinctrl_dev_from_devname(ROCKCHIP_PINCTRL_DEV_NAME);

the ROCKCHIP_PINCTRL_DEV_NAME is a const string from pinctrl head file.

>
>Heiko
>
>> Signed-off-by: Jianqun Xu <jay.xu@rock-chips.com>
>> ---
>>  drivers/gpio/gpio-rockchip.c | 10 +++++++---
>>  1 file changed, 7 insertions(+), 3 deletions(-)
>>
>> diff --git a/drivers/gpio/gpio-rockchip.c b/drivers/gpio/gpio-rockchip.c
>> index 1da0324445cc..46c54dff92db 100644
>> --- a/drivers/gpio/gpio-rockchip.c
>> +++ b/drivers/gpio/gpio-rockchip.c
>> @@ -690,6 +690,9 @@ rockchip_gpio_find_bank(struct pinctrl_dev *pctldev, int id)
>>  int i, found = 0;
>> 
>>  info = pinctrl_dev_get_drvdata(pctldev);
>> +	if (!info)
>> +	return NULL;
>> +
>>  bank = info->ctrl->pin_banks;
>>  for (i = 0; i < info->ctrl->nr_banks; i++, bank++) {
>>  if (bank->bank_num == id) {
>> @@ -705,15 +708,16 @@ static int rockchip_gpio_probe(struct platform_device *pdev)
>>  {
>>  struct device *dev = &pdev->dev;
>>  struct device_node *np = dev->of_node;
>> -	struct device_node *pctlnp = of_get_parent(np);
>> +	struct device_node *pctlnp = NULL;
>>  struct pinctrl_dev *pctldev = NULL;
>>  struct rockchip_pin_bank *bank = NULL;
>>  struct rockchip_pin_output_deferred *cfg;
>>  static int gpio;
>>  int id, ret;
>> 
>> -	if (!np || !pctlnp)
>> -	return -ENODEV;
>> +	pctlnp = of_parse_phandle(np, "gpio-ranges", 0);
>> +	if (!pctlnp)
>> +	pctlnp = of_get_parent(np);
>> 
>>  pctldev = of_pinctrl_get(pctlnp);
>>  if (!pctldev)
>>
>
>
>
>
Linus Walleij March 15, 2022, 1:15 a.m. UTC | #4
On Thu, Mar 3, 2022 at 7:22 AM Jianqun Xu <jay.xu@rock-chips.com> wrote:

> The dt nodes for rockchip soc designes as
>
>         pinctrl: pinctrl {
>                 gpio {
>                         gpio-ranges = <&pinctrl xxx>;
>                 };
>         };
>
> Currently, we get the pinctrl dt node from parent of gpio, this patch
> try to get pinctrl dt node from 'gpio-ranges' property.
>
> After this patch, the dt nodes possible to be
>
>         gpio {
>                 gpio-ranges = <&pinctrl xxx>;
>         };
>
>         pinctrl: pinctrl {
>
>         };
>
> then the gpio driver could register as platform device itself, but not
> populate from pinctrl driver.
>
> Signed-off-by: Jianqun Xu <jay.xu@rock-chips.com>
(...)
> -       if (!np || !pctlnp)
> -               return -ENODEV;
> +       pctlnp = of_parse_phandle(np, "gpio-ranges", 0);
> +       if (!pctlnp)
> +               pctlnp = of_get_parent(np);

GPIO ranges can be several and point to several different pin controllers.

I understand it does not in your case, so I guess it is fine if Heiko
is fine with it.

Yours,
Linus Walleij
diff mbox series

Patch

diff --git a/drivers/gpio/gpio-rockchip.c b/drivers/gpio/gpio-rockchip.c
index 1da0324445cc..46c54dff92db 100644
--- a/drivers/gpio/gpio-rockchip.c
+++ b/drivers/gpio/gpio-rockchip.c
@@ -690,6 +690,9 @@  rockchip_gpio_find_bank(struct pinctrl_dev *pctldev, int id)
 	int i, found = 0;
 
 	info = pinctrl_dev_get_drvdata(pctldev);
+	if (!info)
+		return NULL;
+
 	bank = info->ctrl->pin_banks;
 	for (i = 0; i < info->ctrl->nr_banks; i++, bank++) {
 		if (bank->bank_num == id) {
@@ -705,15 +708,16 @@  static int rockchip_gpio_probe(struct platform_device *pdev)
 {
 	struct device *dev = &pdev->dev;
 	struct device_node *np = dev->of_node;
-	struct device_node *pctlnp = of_get_parent(np);
+	struct device_node *pctlnp = NULL;
 	struct pinctrl_dev *pctldev = NULL;
 	struct rockchip_pin_bank *bank = NULL;
 	struct rockchip_pin_output_deferred *cfg;
 	static int gpio;
 	int id, ret;
 
-	if (!np || !pctlnp)
-		return -ENODEV;
+	pctlnp = of_parse_phandle(np, "gpio-ranges", 0);
+	if (!pctlnp)
+		pctlnp = of_get_parent(np);
 
 	pctldev = of_pinctrl_get(pctlnp);
 	if (!pctldev)