[PATCH] iio: temperature: ltc2983: fix leak of device node iterator

[PATCH] iio: temperature: ltc2983: fix leak of device node iterator

[Tobias Jordan]

Add missing of_node_put calls for the error paths of the
for_each_available_child_of_node loop in ltc2983_parse_dt.

Thought about adding an "goto err_of_node_put" instead, but as the error
paths are quite divergent, I'm not sure if that wouldn't complicate
things.

Fixes: f110f3188e56 ("iio: temperature: Add support for LTC2983")
Signed-off-by: Tobias Jordan <kernel@cdqe.de>
---
 drivers/iio/temperature/ltc2983.c | 6 ++++++
 1 file changed, 6 insertions(+)

diff --git a/drivers/iio/temperature/ltc2983.c b/drivers/iio/temperature/ltc2983.c
index 55ff28a0f1c7..438e0ee29025 100644
--- a/drivers/iio/temperature/ltc2983.c
+++ b/drivers/iio/temperature/ltc2983.c
@@ -1285,6 +1285,7 @@ static int ltc2983_parse_dt(struct ltc2983_data *st)
 		ret = of_property_read_u32(child, "reg", &sensor.chan);
 		if (ret) {
 			dev_err(dev, "reg property must given for child nodes\n");
+			of_node_put(child);
 			return ret;
 		}
 
@@ -1293,9 +1294,11 @@ static int ltc2983_parse_dt(struct ltc2983_data *st)
 		    sensor.chan > LTC2983_MAX_CHANNELS_NR) {
 			dev_err(dev,
 				"chan:%d must be from 1 to 20\n", sensor.chan);
+			of_node_put(child);
 			return -EINVAL;
 		} else if (channel_avail_mask & BIT(sensor.chan)) {
 			dev_err(dev, "chan:%d already in use\n", sensor.chan);
+			of_node_put(child);
 			return -EINVAL;
 		}
 
@@ -1304,6 +1307,7 @@ static int ltc2983_parse_dt(struct ltc2983_data *st)
 		if (ret) {
 			dev_err(dev,
 				"adi,sensor-type property must given for child nodes\n");
+			of_node_put(child);
 			return ret;
 		}
 
@@ -1334,12 +1338,14 @@ static int ltc2983_parse_dt(struct ltc2983_data *st)
 			st->sensors[chan] = ltc2983_adc_new(child, st, &sensor);
 		} else {
 			dev_err(dev, "Unknown sensor type %d\n", sensor.type);
+			of_node_put(child);
 			return -EINVAL;
 		}
 
 		if (IS_ERR(st->sensors[chan])) {
 			dev_err(dev, "Failed to create sensor %ld",
 				PTR_ERR(st->sensors[chan]));
+			of_node_put(child);
 			return PTR_ERR(st->sensors[chan]);
 		}
 		/* set generic sensor parameters */
-- 
2.20.1

Re: [PATCH] iio: temperature: ltc2983: fix leak of device node iterator

[Jonathan Cameron]

On Sat, 26 Sep 2020 12:45:54 +0200
Tobias Jordan <kernel@cdqe.de> wrote:

> Add missing of_node_put calls for the error paths of the
> for_each_available_child_of_node loop in ltc2983_parse_dt.
> 
> Thought about adding an "goto err_of_node_put" instead, but as the error
> paths are quite divergent, I'm not sure if that wouldn't complicate
> things.
> 
> Fixes: f110f3188e56 ("iio: temperature: Add support for LTC2983")
> Signed-off-by: Tobias Jordan <kernel@cdqe.de>
Nuno Sá sent a fix for this yesterday which I've just applied.

Unlucky timing on this one!

Thanks,

Jonathan

> ---
>  drivers/iio/temperature/ltc2983.c | 6 ++++++
>  1 file changed, 6 insertions(+)
> 
> diff --git a/drivers/iio/temperature/ltc2983.c b/drivers/iio/temperature/ltc2983.c
> index 55ff28a0f1c7..438e0ee29025 100644
> --- a/drivers/iio/temperature/ltc2983.c
> +++ b/drivers/iio/temperature/ltc2983.c
> @@ -1285,6 +1285,7 @@ static int ltc2983_parse_dt(struct ltc2983_data *st)
>  		ret = of_property_read_u32(child, "reg", &sensor.chan);
>  		if (ret) {
>  			dev_err(dev, "reg property must given for child nodes\n");
> +			of_node_put(child);
>  			return ret;
>  		}
>  
> @@ -1293,9 +1294,11 @@ static int ltc2983_parse_dt(struct ltc2983_data *st)
>  		    sensor.chan > LTC2983_MAX_CHANNELS_NR) {
>  			dev_err(dev,
>  				"chan:%d must be from 1 to 20\n", sensor.chan);
> +			of_node_put(child);
>  			return -EINVAL;
>  		} else if (channel_avail_mask & BIT(sensor.chan)) {
>  			dev_err(dev, "chan:%d already in use\n", sensor.chan);
> +			of_node_put(child);
>  			return -EINVAL;
>  		}
>  
> @@ -1304,6 +1307,7 @@ static int ltc2983_parse_dt(struct ltc2983_data *st)
>  		if (ret) {
>  			dev_err(dev,
>  				"adi,sensor-type property must given for child nodes\n");
> +			of_node_put(child);
>  			return ret;
>  		}
>  
> @@ -1334,12 +1338,14 @@ static int ltc2983_parse_dt(struct ltc2983_data *st)
>  			st->sensors[chan] = ltc2983_adc_new(child, st, &sensor);
>  		} else {
>  			dev_err(dev, "Unknown sensor type %d\n", sensor.type);
> +			of_node_put(child);
>  			return -EINVAL;
>  		}
>  
>  		if (IS_ERR(st->sensors[chan])) {
>  			dev_err(dev, "Failed to create sensor %ld",
>  				PTR_ERR(st->sensors[chan]));
> +			of_node_put(child);
>  			return PTR_ERR(st->sensors[chan]);
>  		}
>  		/* set generic sensor parameters */

Re: [PATCH] iio: temperature: ltc2983: fix leak of device node iterator

[Jonathan Cameron]

On Sat, 26 Sep 2020 16:45:56 +0200
Markus Elfring <Markus.Elfring@web.de> wrote:

> > Thought about adding an "goto err_of_node_put" instead, but as the error
> > paths are quite divergent, I'm not sure if that wouldn't complicate
> > things.  
> 
> Please add jump targets like “e_inval” and “put_node” so that a bit of
> common exception handling code can be better reused for this function implementation.
> https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/Documentation/process/coding-style.rst?id=7c7ec3226f5f33f9c050d85ec20f18419c622ad6#n475

On this one I think readability would perhaps be hurt a little by
doing so, particular as we need to do the of_put_node in some but
not all non error paths.  

It is a close run thing between the two options however.

I considered another option of suggesting factoring out this whole
per node block, but to do that we would have to do something a bit
odd with the return value as we have 3 options.
* error
* do not parse any more children.
* continue to parse children.

So I think in this case Tobias' solution is the best one available.

Thanks,

Jonathan

> 
> Regards,
> Markus