Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]

Groups > linux.kernel > #1316836 > unrolled thread

[PATCH] input: keyboard: cap11xx: Add missing of_node_put

Back to article view | Back to linux.kernel

Contents

  [PATCH] input: keyboard: cap11xx: Add missing of_node_put Amitoj Kaur Chawla <amitoj1606@gmail.com> - 2016-01-25 16:40 +0100
    Re: [PATCH] input: keyboard: cap11xx: Add missing of_node_put Dmitry Torokhov <dmitry.torokhov@gmail.com> - 2016-01-28 00:50 +0100
      Re: [PATCH] input: keyboard: cap11xx: Add missing of_node_put Julia Lawall <julia.lawall@lip6.fr> - 2016-01-28 10:20 +0100
        Re: [PATCH] input: keyboard: cap11xx: Add missing of_node_put Dmitry Torokhov <dmitry.torokhov@gmail.com> - 2016-01-28 20:00 +0100

#1316836 — [PATCH] input: keyboard: cap11xx: Add missing of_node_put

for_each_child_of_node performs an of_node_get on each iteration, so
to break out of the loop an of_node_put is required.

Found using Coccinelle. The semantic patch used for this is as follows:

// <smpl>
@@
expression e;
local idexpression n;
@@

 for_each_child_of_node(..., n) {
   ... when != of_node_put(n)
       when != e = n
(
   return n;
|
+  of_node_put(n);
?  return ...;
)
   ...
 }
// </smpl

Signed-off-by: Amitoj Kaur Chawla <amitoj1606@gmail.com>
---
 drivers/input/keyboard/cap11xx.c | 12 +++++++++---
 1 file changed, 9 insertions(+), 3 deletions(-)

diff --git a/drivers/input/keyboard/cap11xx.c b/drivers/input/keyboard/cap11xx.c
index 378db10..27cd7df 100644
--- a/drivers/input/keyboard/cap11xx.c
+++ b/drivers/input/keyboard/cap11xx.c
@@ -304,8 +304,10 @@ static int cap11xx_init_leds(struct device *dev,
 		led->cdev.brightness = LED_OFF;
 
 		error = of_property_read_u32(child, "reg", &reg);
-		if (error != 0 || reg >= num_leds)
-			return -EINVAL;
+		if (error != 0 || reg >= num_leds) {
+			error = -EINVAL;
+			goto putchild;
+		}
 
 		led->reg = reg;
 		led->priv = priv;
@@ -314,13 +316,17 @@ static int cap11xx_init_leds(struct device *dev,
 
 		error = devm_led_classdev_register(dev, &led->cdev);
 		if (error)
-			return error;
+			goto putchild;
 
 		priv->num_leds++;
 		led++;
 	}
 
 	return 0;
+
+putchild:
+	of_node_put(child);
+	return error;
 }
 #else
 static int cap11xx_init_leds(struct device *dev,
-- 
1.9.1

[toc] | [next] | [standalone]

#1320153

On Mon, Jan 25, 2016 at 09:01:21PM +0530, Amitoj Kaur Chawla wrote:
> for_each_child_of_node performs an of_node_get on each iteration, so
> to break out of the loop an of_node_put is required.
> 
> Found using Coccinelle. The semantic patch used for this is as follows:
> 
> // <smpl>
> @@
> expression e;
> local idexpression n;
> @@
> 
>  for_each_child_of_node(..., n) {
>    ... when != of_node_put(n)
>        when != e = n
> (
>    return n;
> |
> +  of_node_put(n);
> ?  return ...;
> )
>    ...
>  }
> // </smpl
> 
> Signed-off-by: Amitoj Kaur Chawla <amitoj1606@gmail.com>
> ---
>  drivers/input/keyboard/cap11xx.c | 12 +++++++++---
>  1 file changed, 9 insertions(+), 3 deletions(-)
> 
> diff --git a/drivers/input/keyboard/cap11xx.c b/drivers/input/keyboard/cap11xx.c
> index 378db10..27cd7df 100644
> --- a/drivers/input/keyboard/cap11xx.c
> +++ b/drivers/input/keyboard/cap11xx.c
> @@ -304,8 +304,10 @@ static int cap11xx_init_leds(struct device *dev,
>  		led->cdev.brightness = LED_OFF;
>  
>  		error = of_property_read_u32(child, "reg", &reg);
> -		if (error != 0 || reg >= num_leds)
> -			return -EINVAL;
> +		if (error != 0 || reg >= num_leds) {
> +			error = -EINVAL;
> +			goto putchild;

Instead of jumping to a label I added of_node_put here and also below
and applied, thank you.

I believe there is another input driver that returns from
for_each_child_of_node() without dropping reference.

Thanks.

-- 
Dmitry

[toc] | [prev] | [next] | [standalone]

#1320459

On Wed, 27 Jan 2016, Dmitry Torokhov wrote:

> On Mon, Jan 25, 2016 at 09:01:21PM +0530, Amitoj Kaur Chawla wrote:
> > for_each_child_of_node performs an of_node_get on each iteration, so
> > to break out of the loop an of_node_put is required.
> >
> > Found using Coccinelle. The semantic patch used for this is as follows:
> >
> > // <smpl>
> > @@
> > expression e;
> > local idexpression n;
> > @@
> >
> >  for_each_child_of_node(..., n) {
> >    ... when != of_node_put(n)
> >        when != e = n
> > (
> >    return n;
> > |
> > +  of_node_put(n);
> > ?  return ...;
> > )
> >    ...
> >  }
> > // </smpl
> >
> > Signed-off-by: Amitoj Kaur Chawla <amitoj1606@gmail.com>
> > ---
> >  drivers/input/keyboard/cap11xx.c | 12 +++++++++---
> >  1 file changed, 9 insertions(+), 3 deletions(-)
> >
> > diff --git a/drivers/input/keyboard/cap11xx.c b/drivers/input/keyboard/cap11xx.c
> > index 378db10..27cd7df 100644
> > --- a/drivers/input/keyboard/cap11xx.c
> > +++ b/drivers/input/keyboard/cap11xx.c
> > @@ -304,8 +304,10 @@ static int cap11xx_init_leds(struct device *dev,
> >  		led->cdev.brightness = LED_OFF;
> >
> >  		error = of_property_read_u32(child, "reg", &reg);
> > -		if (error != 0 || reg >= num_leds)
> > -			return -EINVAL;
> > +		if (error != 0 || reg >= num_leds) {
> > +			error = -EINVAL;
> > +			goto putchild;
>
> Instead of jumping to a label I added of_node_put here and also below
> and applied, thank you.

Do you have a general strategy for this?

I asked Arnd Bergmann, and he said that if things were shared and if all
failures later in the function could use the shared label, then one should
use a label.  But I can see that there could be different opinions about
it.  Maybe two instances is not enough for sharing?  Maybe the fact that
the need for this error handling is limited to the loop means that there
should never be sharing?

thanks,
julia

>
> I believe there is another input driver that returns from
> for_each_child_of_node() without dropping reference.
>
> Thanks.
>
> --
> Dmitry
>

[toc] | [prev] | [next] | [standalone]

#1320947

On Thu, Jan 28, 2016 at 10:16:39AM +0100, Julia Lawall wrote:
> 
> 
> On Wed, 27 Jan 2016, Dmitry Torokhov wrote:
> 
> > On Mon, Jan 25, 2016 at 09:01:21PM +0530, Amitoj Kaur Chawla wrote:
> > > for_each_child_of_node performs an of_node_get on each iteration, so
> > > to break out of the loop an of_node_put is required.
> > >
> > > Found using Coccinelle. The semantic patch used for this is as follows:
> > >
> > > // <smpl>
> > > @@
> > > expression e;
> > > local idexpression n;
> > > @@
> > >
> > >  for_each_child_of_node(..., n) {
> > >    ... when != of_node_put(n)
> > >        when != e = n
> > > (
> > >    return n;
> > > |
> > > +  of_node_put(n);
> > > ?  return ...;
> > > )
> > >    ...
> > >  }
> > > // </smpl
> > >
> > > Signed-off-by: Amitoj Kaur Chawla <amitoj1606@gmail.com>
> > > ---
> > >  drivers/input/keyboard/cap11xx.c | 12 +++++++++---
> > >  1 file changed, 9 insertions(+), 3 deletions(-)
> > >
> > > diff --git a/drivers/input/keyboard/cap11xx.c b/drivers/input/keyboard/cap11xx.c
> > > index 378db10..27cd7df 100644
> > > --- a/drivers/input/keyboard/cap11xx.c
> > > +++ b/drivers/input/keyboard/cap11xx.c
> > > @@ -304,8 +304,10 @@ static int cap11xx_init_leds(struct device *dev,
> > >  		led->cdev.brightness = LED_OFF;
> > >
> > >  		error = of_property_read_u32(child, "reg", &reg);
> > > -		if (error != 0 || reg >= num_leds)
> > > -			return -EINVAL;
> > > +		if (error != 0 || reg >= num_leds) {
> > > +			error = -EINVAL;
> > > +			goto putchild;
> >
> > Instead of jumping to a label I added of_node_put here and also below
> > and applied, thank you.
> 
> Do you have a general strategy for this?
> 
> I asked Arnd Bergmann, and he said that if things were shared and if all
> failures later in the function could use the shared label, then one should
> use a label.  But I can see that there could be different opinions about
> it.  Maybe two instances is not enough for sharing?  Maybe the fact that
> the need for this error handling is limited to the loop means that there
> should never be sharing?

I do not think I can formalize the rule well, it is a bit of everything:

- the function is devm-ised and I do not like mixing "goto err" style of
  cleanups with automatic devm cleanup
- there was no "goto err*" in the function before the change
- as you mentioned the cleanup "belongs" to the loop
- there were only 2 instances where we needed to do cleanup
- amount of cleanup was minimal

As a side not I am unhappy with this API as it is very similar
list_for_each and for_each_set_bit, etc, so needing to drop reference
when breaking/returning is quite unexpected ;(

Thanks.

-- 
Dmitry

[toc] | [prev] | [standalone]

Back to top | Article view | linux.kernel

csiph-web