[PATCH] memory: emif: fix device tree node reference leak

[PATCH] memory: emif: fix device tree node reference leak

[Weigang He]

The EMIF driver acquires a device tree node reference via
of_parse_phandle() but never releases it with of_node_put(),
causing a reference count leak.

This affects:
1. Error paths in of_get_memory_device_details() - if allocation
   or validation fails after of_parse_phandle() succeeds, the
   reference is leaked
2. Normal driver removal - emif_remove() never releases the
   reference stored in emif->np_ddr

Fix by adding an err_put_node label to properly release the node
reference on error paths, and adding of_node_put() in emif_remove()
to release the reference during normal driver unload.

Fixes: e6b42eb6a66c ("memory: emif: add device tree support to emif driver")
Signed-off-by: Weigang He <geoffreyhe2@gmail.com>
---
 drivers/memory/emif.c | 7 +++++--
 1 file changed, 5 insertions(+), 2 deletions(-)

diff --git a/drivers/memory/emif.c b/drivers/memory/emif.c
index 2fadad0666b1f..e4f7547efbff4 100644
--- a/drivers/memory/emif.c
+++ b/drivers/memory/emif.c
@@ -922,7 +922,7 @@ static struct emif_data *of_get_memory_device_details(
 	if (!emif || !pd || !dev_info) {
 		dev_err(dev, "%s: Out of memory!!\n",
 			__func__);
-		goto error;
+		goto err_put_node;
 	}
 
 	emif->plat_data		= pd;
@@ -946,7 +946,7 @@ static struct emif_data *of_get_memory_device_details(
 			pd->device_info->io_width, pd->phy_type, pd->ip_rev,
 			emif->dev)) {
 		dev_err(dev, "%s: invalid device data!!\n", __func__);
-		goto error;
+		goto err_put_node;
 	}
 	/*
 	 * For EMIF instances other than EMIF1 see if the devices connected
@@ -970,6 +970,8 @@ static struct emif_data *of_get_memory_device_details(
 	emif->plat_data->min_tck = of_get_min_tck(np_ddr, emif->dev);
 	goto out;
 
+err_put_node:
+	of_node_put(np_ddr);
 error:
 	return NULL;
 out:
@@ -1139,6 +1141,7 @@ static void emif_remove(struct platform_device *pdev)
 {
 	struct emif_data *emif = platform_get_drvdata(pdev);
 
+	of_node_put(emif->np_ddr);
 	emif_debugfs_exit(emif);
 }
 
-- 
2.34.1

Re: [PATCH] memory: emif: fix device tree node reference leak

[Krzysztof Kozlowski]

On 15/01/2026 07:42, Weigang He wrote:
> The EMIF driver acquires a device tree node reference via
> of_parse_phandle() but never releases it with of_node_put(),
> causing a reference count leak.
> 
> This affects:
> 1. Error paths in of_get_memory_device_details() - if allocation
>    or validation fails after of_parse_phandle() succeeds, the
>    reference is leaked
> 2. Normal driver removal - emif_remove() never releases the
>    reference stored in emif->np_ddr
> 
> Fix by adding an err_put_node label to properly release the node
> reference on error paths, and adding of_node_put() in emif_remove()
> to release the reference during normal driver unload.
> 
> Fixes: e6b42eb6a66c ("memory: emif: add device tree support to emif driver")

Missing Cc stable.

> Signed-off-by: Weigang He <geoffreyhe2@gmail.com>
> ---
>  drivers/memory/emif.c | 7 +++++--
>  1 file changed, 5 insertions(+), 2 deletions(-)
> 
> diff --git a/drivers/memory/emif.c b/drivers/memory/emif.c
> index 2fadad0666b1f..e4f7547efbff4 100644
> --- a/drivers/memory/emif.c
> +++ b/drivers/memory/emif.c
> @@ -922,7 +922,7 @@ static struct emif_data *of_get_memory_device_details(
>  	if (!emif || !pd || !dev_info) {
>  		dev_err(dev, "%s: Out of memory!!\n",
>  			__func__);
> -		goto error;
> +		goto err_put_node;
>  	}
>  
>  	emif->plat_data		= pd;
> @@ -946,7 +946,7 @@ static struct emif_data *of_get_memory_device_details(
>  			pd->device_info->io_width, pd->phy_type, pd->ip_rev,
>  			emif->dev)) {
>  		dev_err(dev, "%s: invalid device data!!\n", __func__);
> -		goto error;
> +		goto err_put_node;
>  	}
>  	/*
>  	 * For EMIF instances other than EMIF1 see if the devices connected
> @@ -970,6 +970,8 @@ static struct emif_data *of_get_memory_device_details(
>  	emif->plat_data->min_tck = of_get_min_tck(np_ddr, emif->dev);
>  	goto out;
>  
> +err_put_node:
> +	of_node_put(np_ddr);
>  error:

This is unused now. You do not need new label, just use existing one.


Best regards,
Krzysztof