Skip to contentSkip to conversation

gpio: palmas: add dt support

PATCH In review Laxman Dewangan 4 participants 9 messages Mar 21, 2013 – Apr 2, 2013

Subsystems (2)
lore
Share & tools
lore
#0 lore raw

PATCH 1/1 gpio: palmas: add dt support

Add of_device_id table for Palma GPIO to be enable the driver from DT file.

The driver can be registered from DT file as:

	palmas: tps65913@58 {
		:::::::::::
		palmas_gpio: palmas_gpio {
			compatible = "ti,palmas-gpio";
			gpio-controller;
			#gpio-cells = <2>;
		};
	};
---
 drivers/gpio/gpio-palmas.c |   11 ++++++++++-
 1 files changed, 10 insertions(+), 1 deletions(-)
drivers/gpio/gpio-palmas.c +10 -1
index e3a4e56..fddaa0d 100644
@@ -134,7 +134,7 @@ static int palmas_gpio_probe(struct platform_device *pdev)
 	palmas_gpio->gpio_chip.get	= palmas_gpio_get;
 	palmas_gpio->gpio_chip.dev = &pdev->dev;
 #ifdef CONFIG_OF_GPIO
-	palmas_gpio->gpio_chip.of_node = palmas->dev->of_node;
+	palmas_gpio->gpio_chip.of_node = pdev->dev.of_node;
 #endif
 	palmas_pdata = dev_get_platdata(palmas->dev);
 	if (palmas_pdata && palmas_pdata->gpio_base)
@@ -159,9 +159,18 @@ static int palmas_gpio_remove(struct platform_device *pdev)
 	return gpiochip_remove(&palmas_gpio->gpio_chip);
 }
 
+#ifdef CONFIG_OF
+static struct of_device_id of_palmas_gpio_match[] = {
+	{ .compatible = "ti,palmas-gpio"},
+	{ },
+};
+MODULE_DEVICE_TABLE(of, of_palmas_gpio_match);
+#endif
+
 static struct platform_driver palmas_gpio_driver = {
 	.driver.name	= "palmas-gpio",
 	.driver.owner	= THIS_MODULE,
+	.driver.of_match_table = of_match_ptr(of_palmas_gpio_match),
 	.probe		= palmas_gpio_probe,
 	.remove		= palmas_gpio_remove,
 };
Signature · 1 line
1.7.1.1
#1 lore raw
Quoted 2 lines · Laxman Dewangan

Add of_device_id table for Palma GPIO to be enable the driver from DT file.

On this, the RTC patches, and the regulator patches, I think you should be Ccing the Slimlogic guys (CC'd now) and perhaps some people from TI (CC'd now; I hope the right people) , to make sure so that you're all on the same page w.r.t. how the DT bindings work, and so you don't step on each-others toes.

#2 lore raw
In reply to Laxman Dewangan

drivers/gpio/gpio-palmas.c static int palmas_gpio_probe(struct platform_device *pdev) · line 138

Quoted 4 lines · Laxman Dewangan

#ifdef CONFIG_OF_GPIO

drivers/gpio/gpio-palmas.c +1 -1
-       palmas_gpio->gpio_chip.of_node = palmas->dev->of_node;
+       palmas_gpio->gpio_chip.of_node = pdev->dev.of_node;
 #endif

OK I think that #ifdef is necessary...

drivers/gpio/gpio-palmas.c static int palmas_gpio_remove(struct platform_device *pdev) · line 168

Quoted 7 lines
drivers/gpio/gpio-palmas.c +7 -0
+#ifdef CONFIG_OF
+static struct of_device_id of_palmas_gpio_match[] = {
+       { .compatible = "ti,palmas-gpio"},
+       { },
+};
+MODULE_DEVICE_TABLE(of, of_palmas_gpio_match);
+#endif

But please drop the #ifdef here unless it causes compile errors (I don't think it will.)

Yours,
Linus Walleij

#3 lore raw
Quoted 8 lines · Linus Walleij

drivers/gpio/gpio-palmas.c static int palmas_gpio_probe(struct platform_device *pdev) · line 138

Quoted 4 lines · Laxman Dewangan

#ifdef CONFIG_OF_GPIO

drivers/gpio/gpio-palmas.c +1 -1
-       palmas_gpio->gpio_chip.of_node = palmas->dev->of_node;
+       palmas_gpio->gpio_chip.of_node = pdev->dev.of_node;
 #endif

OK I think that #ifdef is necessary...

Laxman,

Don't we need to resolve and agree on the final DT bindings before we can start making changes like this? It's not clear yet whether everyone is on the same page re: how the MFD sub-devices are modelled in DT - whether each sub-component really is a standalone device, or whether the MFD itself instantiates all its children based on internal static tables rather than DT.

Presumably, the answer to that question directly determines whether the code change above is correct; the correct of_node might be either the main Palmas node (if DT doesn't represent the MFD components) or it might be the regulator sub-node (if DT is used to represent MFD components).

Given that, I'm not sure why the Slimlogic people aren't CC'd on this patch:-(

#4 lore raw
Quoted 8 lines · Stephen Warren
Quoted 7 lines · Linus Walleij

drivers/gpio/gpio-palmas.c static int palmas_gpio_probe(struct platform_device *pdev) · line 138

Quoted 4 lines · Laxman Dewangan
  #ifdef CONFIG_OF_GPIO
drivers/gpio/gpio-palmas.c +1 -1
-       palmas_gpio->gpio_chip.of_node = palmas->dev->of_node;
+       palmas_gpio->gpio_chip.of_node = pdev->dev.of_node;
  #endif

OK I think that #ifdef is necessary...

Quoted 8 lines

Laxman,

Don't we need to resolve and agree on the final DT bindings before we can start making changes like this? It's not clear yet whether everyone is on the same page re: how the MFD sub-devices are modelled in DT - whether each sub-component really is a standalone device, or whether the MFD itself instantiates all its children based on internal static tables rather than DT.

Yes, we need to agree on DT. I sent the patch based on the patches came from TI/Slimlogic and other discussion about DT patches. Recently, you commented on DT for palmas, it need to be completely IP based or hw based. Mix will not work.
If it is IP based then almost all palmas-* file need to be rewrite, no one is written as IP based. All are using palma structures, palma header, palma macro etc which should not be there.

Quoted 3 lines

Given that, I'm not sure why the Slimlogic people aren't CC'd on this patch:-(

I have not added then on my original patch as ./script/get_maintainers.pl did not give me name. However, you added immediately them (TI and slimlogic) on this patch itself. I have not got any feedback from them yet.

I

#5 lore raw
In reply to Linus Walleij
Quoted 12 lines · Linus Walleij

drivers/gpio/gpio-palmas.c static int palmas_gpio_remove(struct platform_device *pdev) · line 168

Quoted 7 lines · Laxman Dewangan
drivers/gpio/gpio-palmas.c +7 -0
+#ifdef CONFIG_OF
+static struct of_device_id of_palmas_gpio_match[] = {
+       { .compatible = "ti,palmas-gpio"},
+       { },
+};
+MODULE_DEVICE_TABLE(of, of_palmas_gpio_match);
+#endif

But please drop the #ifdef here unless it causes compile errors (I don't think it will.)

I am using this table as

driver.of_match_table = of_match_ptr(of_palmas_gpio_match), of_match_ptr is macro which is NULL in case of CONFIG_OF not defined. So if I remove ifdefs then it may create build warning as unused variable.

Therefore, I think it is require here.

#6 lore raw
Quoted 16 lines · Laxman Dewangan
Quoted 14 lines · Linus Walleij

drivers/gpio/gpio-palmas.c static int palmas_gpio_remove(struct platform_device *pdev) · line 168

Quoted 7 lines · Laxman Dewangan
drivers/gpio/gpio-palmas.c +7 -0
+#ifdef CONFIG_OF
+static struct of_device_id of_palmas_gpio_match[] = {
+       { .compatible = "ti,palmas-gpio"},
+       { },
+};
+MODULE_DEVICE_TABLE(of, of_palmas_gpio_match);
+#endif

But please drop the #ifdef here unless it causes compile errors (I don't think it will.)

Quoted 5 lines

I am using this table as

driver.of_match_table = of_match_ptr(of_palmas_gpio_match), of_match_ptr is macro which is NULL in case of CONFIG_OF not defined. So if I remove ifdefs then it may create build warning as unused variable.

I think Linus's point is that you can simply remove the use of of_match_ptr(). The only disadvantage of doing so is that the table will always be included in the object file, but it's so small that it's probably not worth worrying about.

#7 lore raw
Quoted 19 lines · Stephen Warren
Quoted 18 lines · Laxman Dewangan
Quoted 14 lines · Linus Walleij

drivers/gpio/gpio-palmas.c static int palmas_gpio_remove(struct platform_device *pdev) · line 168

Quoted 7 lines · Laxman Dewangan
drivers/gpio/gpio-palmas.c +7 -0
+#ifdef CONFIG_OF
+static struct of_device_id of_palmas_gpio_match[] = {
+       { .compatible = "ti,palmas-gpio"},
+       { },
+};
+MODULE_DEVICE_TABLE(of, of_palmas_gpio_match);
+#endif

But please drop the #ifdef here unless it causes compile errors (I don't think it will.)

I am using this table as

Quoted 8 lines
Quoted 3 lines

driver.of_match_table = of_match_ptr(of_palmas_gpio_match), of_match_ptr is macro which is NULL in case of CONFIG_OF not defined. So if I remove ifdefs then it may create build warning as unused variable.

I think Linus's point is that you can simply remove the use of of_match_ptr(). The only disadvantage of doing so is that the table will always be included in the object file, but it's so small that it's probably not worth worrying about.

Oh I wasn't that smart :-)

But what you're saying seems true.

The of_match_ptr() is something
I haven't quite seen before and don't quite understand the semantics of, why would we use that?

Yours,
Linus Walleij

#8 lore raw
Quoted 29 lines · Linus Walleij
Quoted 19 lines · Stephen Warren
Quoted 18 lines · Laxman Dewangan
Quoted 14 lines · Linus Walleij

drivers/gpio/gpio-palmas.c static int palmas_gpio_remove(struct platform_device *pdev) · line 168

Quoted 7 lines · Laxman Dewangan
drivers/gpio/gpio-palmas.c +7 -0
+#ifdef CONFIG_OF
+static struct of_device_id of_palmas_gpio_match[] = {
+       { .compatible = "ti,palmas-gpio"},
+       { },
+};
+MODULE_DEVICE_TABLE(of, of_palmas_gpio_match);
+#endif

But please drop the #ifdef here unless it causes compile errors (I don't think it will.)

I am using this table as

Quoted 8 lines
Quoted 3 lines

driver.of_match_table = of_match_ptr(of_palmas_gpio_match), of_match_ptr is macro which is NULL in case of CONFIG_OF not defined. So if I remove ifdefs then it may create build warning as unused variable.

I think Linus's point is that you can simply remove the use of of_match_ptr(). The only disadvantage of doing so is that the table will always be included in the object file, but it's so small that it's probably not worth worrying about.

Quoted 7 lines

Oh I wasn't that smart :-)

But what you're saying seems true.

The of_match_ptr() is something
I haven't quite seen before and don't quite understand the semantics of, why would we use that?

Very roughly,

of_match_ptr(x) == CONFIG_OF ? x : NULL

(although with ifdefs rather than the ternary operator)

So, if you have ifdef'd the match table with CONFIG_OF, and reference it from code outside that ifdef, then you need to use of_match_ptr() to reference the table, so it isn't referenced when CONFIG_OF is not defined, and hence never causes link problems.