When enabling a counter, the runtime PM reference taken by
pm_runtime_get_sync() is never dropped if rz_mtu3_initialize_counter()
fails and returns an error, e.g. -EBUSY when the channel is already
requested. Because priv->count_is_enabled[] is left false, a subsequent
disable write also takes the early exit path, so the reference is never
recovered. Additionally, if pm_runtime_get_sync() itself fails the
device usage count is left incremented while the hardware is still
accessed with runtime PM suspended.

Use pm_runtime_resume_and_get() to resume the device and bail out if it
fails, and drop the runtime PM reference when
rz_mtu3_initialize_counter() fails.

Fixes: 0be8907359df ("counter: Add Renesas RZ/G2L MTU3a counter driver")
Cc: stable@vger.kernel.org
Signed-off-by: Wentao Liang <vulab@iscas.ac.cn>
---
 drivers/counter/rz-mtu3-cnt.c | 7 ++++++-
 1 file changed, 6 insertions(+), 1 deletion(-)

diff --git a/drivers/counter/rz-mtu3-cnt.c b/drivers/counter/rz-mtu3-cnt.c
index 7bfb6979193c..2bfc92bea757 100644
--- a/drivers/counter/rz-mtu3-cnt.c
+++ b/drivers/counter/rz-mtu3-cnt.c
@@ -504,10 +504,15 @@ static int rz_mtu3_count_enable_write(struct counter_device *counter,
 		goto exit;
 
 	if (enable) {
-		pm_runtime_get_sync(counter->parent);
+		ret = pm_runtime_resume_and_get(counter->parent);
+		if (ret)
+			goto exit;
+
 		ret = rz_mtu3_initialize_counter(counter, count->id);
 		if (ret == 0)
 			priv->count_is_enabled[count->id] = true;
+		else
+			pm_runtime_put(counter->parent);
 	} else {
 		rz_mtu3_terminate_counter(counter, count->id);
 		priv->count_is_enabled[count->id] = false;
-- 
2.34.1

On Wed, Sep 16, 2026 at 09:22:42AM +0000, Wentao Liang wrote:
> When enabling a counter, the runtime PM reference taken by
> pm_runtime_get_sync() is never dropped if rz_mtu3_initialize_counter()
> fails and returns an error, e.g. -EBUSY when the channel is already
> requested. Because priv->count_is_enabled[] is left false, a subsequent
> disable write also takes the early exit path, so the reference is never
> recovered. Additionally, if pm_runtime_get_sync() itself fails the
> device usage count is left incremented while the hardware is still
> accessed with runtime PM suspended.
> 
> Use pm_runtime_resume_and_get() to resume the device and bail out if it
> fails, and drop the runtime PM reference when
> rz_mtu3_initialize_counter() fails.
> 
> Fixes: 0be8907359df ("counter: Add Renesas RZ/G2L MTU3a counter driver")
> Cc: stable@vger.kernel.org
> Signed-off-by: Wentao Liang <vulab@iscas.ac.cn>
> ---
>  drivers/counter/rz-mtu3-cnt.c | 7 ++++++-
>  1 file changed, 6 insertions(+), 1 deletion(-)
> 
> diff --git a/drivers/counter/rz-mtu3-cnt.c b/drivers/counter/rz-mtu3-cnt.c
> index 7bfb6979193c..2bfc92bea757 100644
> --- a/drivers/counter/rz-mtu3-cnt.c
> +++ b/drivers/counter/rz-mtu3-cnt.c
> @@ -504,10 +504,15 @@ static int rz_mtu3_count_enable_write(struct counter_device *counter,
>  		goto exit;
> 
>  	if (enable) {
> -		pm_runtime_get_sync(counter->parent);
> +		ret = pm_runtime_resume_and_get(counter->parent);
> +		if (ret)
> +			goto exit;
> +
>  		ret = rz_mtu3_initialize_counter(counter, count->id);
>  		if (ret == 0)
>  			priv->count_is_enabled[count->id] = true;
> +		else
> +			pm_runtime_put(counter->parent);
>  	} else {
>  		rz_mtu3_terminate_counter(counter, count->id);
>  		priv->count_is_enabled[count->id] = false;
> --
> 2.34.1
> 

Hello Wentao,

This looks like an earlier patch submission we've received.[^1] In fact,
I wonder whether these patches will be necessary anymore if the
"counter: rz-mtu3-cnt: check clock and runtime PM errors" series is
merged.[^2]

Sincerely,

William Breathitt Gray

[^1]: https://lore.kernel.org/all/20260814134138.1388045-1-ruoyuw560@gmail.com/
[^2]: https://lore.kernel.org/all/20260903090719.3498955-1-dayou5941@163.com/