Skip to contentSkip to conversation

gpio: pxa: use devres for the clock struct

PATCH In review Bartosz Golaszewski 4 participants 6 messages Aug 15, 2022 – Aug 27, 2022

Subsystems (3)
lore
Share & tools
lore
#0 lore raw

PATCH 1/1 gpio: pxa: use devres for the clock struct

The clock is never released after probe(). Use devres to not leak resources.

---
 drivers/gpio/gpio-pxa.c | 11 ++---------
 1 file changed, 2 insertions(+), 9 deletions(-)
drivers/gpio/gpio-pxa.c +2 -9
index c7fbfa3ae43b..1198ab0305d0 100644
@@ -661,24 +661,17 @@ static int pxa_gpio_probe(struct platform_device *pdev)
 	if (IS_ERR(gpio_reg_base))
 		return PTR_ERR(gpio_reg_base);
 
-	clk = clk_get(&pdev->dev, NULL);
+	clk = devm_clk_get_enabled(&pdev->dev, NULL);
 	if (IS_ERR(clk)) {
 		dev_err(&pdev->dev, "Error %ld to get gpio clock\n",
 			PTR_ERR(clk));
 		return PTR_ERR(clk);
 	}
-	ret = clk_prepare_enable(clk);
-	if (ret) {
-		clk_put(clk);
-		return ret;
-	}
 
 	/* Initialize GPIO chips */
 	ret = pxa_init_gpio_chip(pchip, pxa_last_gpio + 1, gpio_reg_base);
-	if (ret) {
-		clk_put(clk);
+	if (ret)
 		return ret;
-	}
 
 	/* clear all GPIO edge detects */
 	for_each_gpio_bank(gpio, c, pchip) {
Signature · 1 line
2.34.1
#2 lore raw
In reply to Bartosz Golaszewski
Quoted 3 lines · Bartosz Golaszewski

The clock is never released after probe(). Use devres to not leak resources.

...

drivers/gpio/gpio-pxa.c static int pxa_gpio_probe(struct platform_device *pdev) · line 668

Quoted 6 lines
drivers/gpio/gpio-pxa.c +1 -1
-       clk = clk_get(&pdev->dev, NULL);
+       clk = devm_clk_get_enabled(&pdev->dev, NULL);
        if (IS_ERR(clk)) {
                dev_err(&pdev->dev, "Error %ld to get gpio clock\n",
                        PTR_ERR(clk));
                return PTR_ERR(clk);

Shouldn't we fix a potential log saturation issue first (by switching to use dev_err_probe() helper)?

Quoted 1 line
        }
Signature · 2 lines
With Best Regards,
Andy Shevchenko
#3 lore raw
Quoted 14 lines · Andy Shevchenko
Quoted 3 lines · Bartosz Golaszewski

The clock is never released after probe(). Use devres to not leak resources.

...

drivers/gpio/gpio-pxa.c static int pxa_gpio_probe(struct platform_device *pdev) · line 668

Quoted 6 lines
drivers/gpio/gpio-pxa.c +1 -1
-       clk = clk_get(&pdev->dev, NULL);
+       clk = devm_clk_get_enabled(&pdev->dev, NULL);
        if (IS_ERR(clk)) {
                dev_err(&pdev->dev, "Error %ld to get gpio clock\n",
                        PTR_ERR(clk));
                return PTR_ERR(clk);
Quoted 2 lines

Shouldn't we fix a potential log saturation issue first (by switching to use dev_err_probe() helper)?

Can be a separate patch, the clock mem leak is a bigger problem IMO so this should be applied first.

Hm isn't it possible to toss the task of fixing a gazillion dev_err_probe() messages on Cocinelle scripts/coccinelle/? I bet it's something the kernel janitors could fix all over the place.

Yours,
Linus Walleij

#4 lore raw

Le 26/08/2022 à 10:20, Linus Walleij a écrit :

Quoted 22 lines
Quoted 14 lines · Andy Shevchenko
Quoted 3 lines · Bartosz Golaszewski

The clock is never released after probe(). Use devres to not leak resources.

...

drivers/gpio/gpio-pxa.c static int pxa_gpio_probe(struct platform_device *pdev) · line 668

Quoted 6 lines
drivers/gpio/gpio-pxa.c +1 -1
-       clk = clk_get(&pdev->dev, NULL);
+       clk = devm_clk_get_enabled(&pdev->dev, NULL);
         if (IS_ERR(clk)) {
                 dev_err(&pdev->dev, "Error %ld to get gpio clock\n",
                         PTR_ERR(clk));
                 return PTR_ERR(clk);
Quoted 2 lines

Shouldn't we fix a potential log saturation issue first (by switching to use dev_err_probe() helper)?

Can be a separate patch, the clock mem leak is a bigger problem IMO so this should be applied first.

Quoted 7 lines

Hm isn't it possible to toss the task of fixing a gazillion dev_err_probe() messages on Cocinelle scripts/coccinelle/? I bet it's something the kernel janitors could fix all over the place.

Yours,
Linus Walleij

Not perfect and certainly incomplete, but gives an idea:

@@
expression x, dev, e, str;
@@
(

	x = devm_clk_get(dev, e);
|
	x = clk_get(dev, e);

)
- if (IS_ERR(x)) {
- dev_err(dev, str);
- return PTR_ERR(x);
- }
+ if (IS_ERR(x))
+ return dev_err_probe(dev, PTR_ERR(x), str);

// This rule only: 291 files changed, 1233 insertions(+), 1634 deletions(-)

// The output of this rule needs to be adjusted to fix the 'str'. The // error code and related text have to be manually removed @@
expression x, dev, e, str;
@@
(

	x = devm_clk_get(dev, e);
|
	x = clk_get(dev, e);

)
- if (IS_ERR(x)) {
- dev_err(dev, str, PTR_ERR(x));
- return PTR_ERR(x);
- }
+ if (IS_ERR(x))
+ return dev_err_probe(dev, PTR_ERR(x), str);

// Both rules: 316 files changed, 1321 insertions(+), 1774 deletions(-)

#5 lore raw

Le 26/08/2022 à 14:18, Christophe JAILLET a écrit :

Quoted 27 lines

Le 26/08/2022 à 10:20, Linus Walleij a écrit :

Quoted 26 lines
Quoted 14 lines · Andy Shevchenko
Quoted 3 lines · Bartosz Golaszewski

The clock is never released after probe(). Use devres to not leak resources.

...

Quoted 6 lines

-       clk = clk_get(&pdev->dev, NULL);
+       clk = devm_clk_get_enabled(&pdev->dev, NULL);
         if (IS_ERR(clk)) {
                 dev_err(&pdev->dev, "Error %ld to get gpio clock\n", PTR_ERR(clk));
                 return PTR_ERR(clk);

Quoted 2 lines

Shouldn't we fix a potential log saturation issue first (by switching to use dev_err_probe() helper)?

Can be a separate patch, the clock mem leak is a bigger problem IMO so this should be applied first.

Hm isn't it possible to toss the task of fixing a gazillion dev_err_probe() messages on Cocinelle scripts/coccinelle/? I bet it's something the kernel janitors could fix all over the place.

Quoted 6 lines
Quoted 3 lines

Yours,
Linus Walleij

// Both rules: 316 files changed, 1321 insertions(+), 1774 deletions(-)

With an updated script, I spot:

   503 files changed, 1962 insertions(+), 2622 deletions(-)

(and 150-200 still needs some manual check or script adjustment)

Does this really make sense to send SO many patches for it?

If yes, should it be done on a per-system basis, or by driver basis?

CJ