[PATCH] powercap: fix memory leak in intel_rapl_common

[PATCH] powercap: fix memory leak in intel_rapl_common

[Sumeet Pawnikar]

The err_cleanup path in rapl_package_register_powercap() function
doesn't free rp->domains before returning when child domain
registration fails. Also, there is a missing the cleanup for the
parent package power zone which was successfully registered earlier.

When registration of child domains fails, now the previously
registered package domain (parent zone) is being unregistered.
This makes both parent and child powercap zones are properly
unregistered before returning the error.
Also, fix the rp->power_zone dangling pointer issue by setting it
to NULL.

Signed-off-by: Sumeet Pawnikar <sumeet4linux@gmail.com>
---
 drivers/powercap/intel_rapl_common.c | 7 +++++++
 1 file changed, 7 insertions(+)

diff --git a/drivers/powercap/intel_rapl_common.c b/drivers/powercap/intel_rapl_common.c
index b9d87e56cbbc..5ea31433abf2 100644
--- a/drivers/powercap/intel_rapl_common.c
+++ b/drivers/powercap/intel_rapl_common.c
@@ -1406,6 +1406,13 @@ static int rapl_package_register_powercap(struct rapl_package *rp)
 					 &rd->power_zone);
 	}
 
+	/* Also unregister the package domain if it was registered */
+	if (rp->power_zone) {
+		powercap_unregister_zone(rp->priv->control_type,
+					 rp->power_zone);
+		rp->power_zone = NULL;
+	}
+
 	return ret;
 }
 
-- 
2.43.0

Re: [PATCH] powercap: fix memory leak in intel_rapl_common

[Rafael J. Wysocki]

On Sun, Dec 7, 2025 at 8:02 PM Sumeet Pawnikar <sumeet4linux@gmail.com> wrote:
>
> The err_cleanup path in rapl_package_register_powercap() function
> doesn't free rp->domains before returning when child domain
> registration fails. Also, there is a missing the cleanup for the
> parent package power zone which was successfully registered earlier.
>
> When registration of child domains fails, now the previously
> registered package domain (parent zone) is being unregistered.

I'm not sure if this is desirable.

The package domain may be still functional without children AFAICS.

> This makes both parent and child powercap zones are properly
> unregistered before returning the error.
> Also, fix the rp->power_zone dangling pointer issue by setting it
> to NULL.
>
> Signed-off-by: Sumeet Pawnikar <sumeet4linux@gmail.com>
> ---
>  drivers/powercap/intel_rapl_common.c | 7 +++++++
>  1 file changed, 7 insertions(+)
>
> diff --git a/drivers/powercap/intel_rapl_common.c b/drivers/powercap/intel_rapl_common.c
> index b9d87e56cbbc..5ea31433abf2 100644
> --- a/drivers/powercap/intel_rapl_common.c
> +++ b/drivers/powercap/intel_rapl_common.c
> @@ -1406,6 +1406,13 @@ static int rapl_package_register_powercap(struct rapl_package *rp)
>                                          &rd->power_zone);
>         }
>
> +       /* Also unregister the package domain if it was registered */
> +       if (rp->power_zone) {
> +               powercap_unregister_zone(rp->priv->control_type,
> +                                        rp->power_zone);
> +               rp->power_zone = NULL;
> +       }
> +
>         return ret;
>  }
>
> --