Re: [PATCH] drm/msm/adreno: Fix a reference leak in a6xx_gpu_init()
Hi Dmitry,
Thanks for your review.
linux/cleanup.h is included in linux/of.h.
So it should be included already.
Best regards,
Feix Gu
On Sat, Jan 24, 2026 at 1:58 AM Dmitry Baryshkov
wrote:
>
> On Sat, Jan 24, 2026 at 12:37:38AM +0800, Felix Gu wrote:
> > In a6xx_gpu_init(), node is obtained via of_parse_phandle().
> > While there was a manual of_node_put() at the end of the
> > common path, several early error returns would bypass this call,
> > resulting in a reference leak.
> > Fix this by using the __free(device_node) cleanup handler to
> > release the reference when the variable goes out of scope.
> >
> > Fixes: 5a903a44a984 ("drm/msm/a6xx: Introduce GMU wrapper support")
> > Signed-off-by: Felix Gu
> > ---
> > drivers/gpu/drm/msm/adreno/a6xx_gpu.c | 5 ++---
> > 1 file changed, 2 insertions(+), 3 deletions(-)
> >
> > diff --git a/drivers/gpu/drm/msm/adreno/a6xx_gpu.c
> > b/drivers/gpu/drm/msm/adreno/a6xx_gpu.c
> > index 2129d230a92b..0bc518d9fd65 100644
> > --- a/drivers/gpu/drm/msm/adreno/a6xx_gpu.c
> > +++ b/drivers/gpu/drm/msm/adreno/a6xx_gpu.c
> > @@ -2640,7 +2640,6 @@ static struct msm_gpu *a6xx_gpu_init(struct
> > drm_device *dev)
> > struct msm_drm_private *priv = dev->dev_private;
> > struct platform_device *pdev = priv->gpu_pdev;
> > struct adreno_platform_config *config = pdev->dev.platform_data;
> > - struct device_node *node;
> > struct a6xx_gpu *a6xx_gpu;
> > struct adreno_gpu *adreno_gpu;
> > struct msm_gpu *gpu;
> > @@ -2660,7 +2659,8 @@ static struct msm_gpu *a6xx_gpu_init(struct
> > drm_device *dev)
> > adreno_gpu->registers = NULL;
> >
> > /* Check if there is a GMU phandle and set it up */
> > - node = of_parse_phandle(pdev->dev.of_node, "qcom,gmu", 0);
> > + struct device_node *node __free(device_node) =
> > + of_parse_phandle(pdev->dev.of_node, "qcom,gmu", 0);
>
> Missing include for cleanup.h.
>
> > /* FIXME: How do we gracefully handle this? */
> > BUG_ON(!node);
> >
> > @@ -2702,7 +2702,6 @@ static struct msm_gpu *a6xx_gpu_init(struct
> > drm_device *dev)
> > ret = a6xx_gmu_wrapper_init(a6xx_gpu, node);
> > else
> > ret = a6xx_gmu_init(a6xx_gpu, node);
> > - of_node_put(node);
> > if (ret) {
> > a6xx_destroy(&(a6xx_gpu->base.base));
> > return ERR_PTR(ret);
> >
> > ---
> > base-commit: a0c666c25aeefd16f4b088c6549a6fb6b65a8a1d
> > change-id: 20260123-a6xx_gpu-cbc095dbe423
> >
> > Best regards,
> > --
> > Felix Gu
> >
>
> --
> With best wishes
> Dmitry
[PATCH] drm/msm/adreno: Fix a reference leak in a6xx_gpu_init()
In a6xx_gpu_init(), node is obtained via of_parse_phandle().
While there was a manual of_node_put() at the end of the
common path, several early error returns would bypass this call,
resulting in a reference leak.
Fix this by using the __free(device_node) cleanup handler to
release the reference when the variable goes out of scope.
Fixes: 5a903a44a984 ("drm/msm/a6xx: Introduce GMU wrapper support")
Signed-off-by: Felix Gu
---
drivers/gpu/drm/msm/adreno/a6xx_gpu.c | 5 ++---
1 file changed, 2 insertions(+), 3 deletions(-)
diff --git a/drivers/gpu/drm/msm/adreno/a6xx_gpu.c
b/drivers/gpu/drm/msm/adreno/a6xx_gpu.c
index 2129d230a92b..0bc518d9fd65 100644
--- a/drivers/gpu/drm/msm/adreno/a6xx_gpu.c
+++ b/drivers/gpu/drm/msm/adreno/a6xx_gpu.c
@@ -2640,7 +2640,6 @@ static struct msm_gpu *a6xx_gpu_init(struct drm_device
*dev)
struct msm_drm_private *priv = dev->dev_private;
struct platform_device *pdev = priv->gpu_pdev;
struct adreno_platform_config *config = pdev->dev.platform_data;
- struct device_node *node;
struct a6xx_gpu *a6xx_gpu;
struct adreno_gpu *adreno_gpu;
struct msm_gpu *gpu;
@@ -2660,7 +2659,8 @@ static struct msm_gpu *a6xx_gpu_init(struct drm_device
*dev)
adreno_gpu->registers = NULL;
/* Check if there is a GMU phandle and set it up */
- node = of_parse_phandle(pdev->dev.of_node, "qcom,gmu", 0);
+ struct device_node *node __free(device_node) =
+ of_parse_phandle(pdev->dev.of_node, "qcom,gmu", 0);
/* FIXME: How do we gracefully handle this? */
BUG_ON(!node);
@@ -2702,7 +2702,6 @@ static struct msm_gpu *a6xx_gpu_init(struct drm_device
*dev)
ret = a6xx_gmu_wrapper_init(a6xx_gpu, node);
else
ret = a6xx_gmu_init(a6xx_gpu, node);
- of_node_put(node);
if (ret) {
a6xx_destroy(&(a6xx_gpu->base.base));
return ERR_PTR(ret);
---
base-commit: a0c666c25aeefd16f4b088c6549a6fb6b65a8a1d
change-id: 20260123-a6xx_gpu-cbc095dbe423
Best regards,
--
Felix Gu
Re: [PATCH] drm/msm/adreno: Fix a reference leak in a6xx_gpu_init()
On Sat, Jan 24, 2026 at 2:23 AM Dmitry Baryshkov
wrote:
>
> On 23/01/2026 20:21, Felix Gu wrote:
> > Hi Dmitry,
> > Thanks for your review.
> > linux/cleanup.h is included in linux/of.h.
> > So it should be included already.
>
> Yes. However we should not depend on include files being provided by
> other kernel-wide includes.
Understood, I will send out V2 to fix it.
>
> Also, please don't top-post when replying to the emails on public MLs.
Sorry for that, thanks for coaching.
Best regards,
Felix Gu
>
> >
> > Best regards,
> > Feix Gu
> >
> > On Sat, Jan 24, 2026 at 1:58 AM Dmitry Baryshkov
> > wrote:
> >>
> >> On Sat, Jan 24, 2026 at 12:37:38AM +0800, Felix Gu wrote:
> >>> In a6xx_gpu_init(), node is obtained via of_parse_phandle().
> >>> While there was a manual of_node_put() at the end of the
> >>> common path, several early error returns would bypass this call,
> >>> resulting in a reference leak.
> >>> Fix this by using the __free(device_node) cleanup handler to
> >>> release the reference when the variable goes out of scope.
> >>>
> >>> Fixes: 5a903a44a984 ("drm/msm/a6xx: Introduce GMU wrapper support")
> >>> Signed-off-by: Felix Gu
> >>> ---
> >>> drivers/gpu/drm/msm/adreno/a6xx_gpu.c | 5 ++---
> >>> 1 file changed, 2 insertions(+), 3 deletions(-)
> >>>
> >>> diff --git a/drivers/gpu/drm/msm/adreno/a6xx_gpu.c
> >>> b/drivers/gpu/drm/msm/adreno/a6xx_gpu.c
> >>> index 2129d230a92b..0bc518d9fd65 100644
> >>> --- a/drivers/gpu/drm/msm/adreno/a6xx_gpu.c
> >>> +++ b/drivers/gpu/drm/msm/adreno/a6xx_gpu.c
> >>> @@ -2640,7 +2640,6 @@ static struct msm_gpu *a6xx_gpu_init(struct
> >>> drm_device *dev)
> >>>struct msm_drm_private *priv = dev->dev_private;
> >>>struct platform_device *pdev = priv->gpu_pdev;
> >>>struct adreno_platform_config *config = pdev->dev.platform_data;
> >>> - struct device_node *node;
> >>>struct a6xx_gpu *a6xx_gpu;
> >>>struct adreno_gpu *adreno_gpu;
> >>>struct msm_gpu *gpu;
> >>> @@ -2660,7 +2659,8 @@ static struct msm_gpu *a6xx_gpu_init(struct
> >>> drm_device *dev)
> >>>adreno_gpu->registers = NULL;
> >>>
> >>>/* Check if there is a GMU phandle and set it up */
> >>> - node = of_parse_phandle(pdev->dev.of_node, "qcom,gmu", 0);
> >>> + struct device_node *node __free(device_node) =
> >>> + of_parse_phandle(pdev->dev.of_node, "qcom,gmu", 0);
> >>
> >> Missing include for cleanup.h.
> >>
> >>>/* FIXME: How do we gracefully handle this? */
> >>>BUG_ON(!node);
> >>>
> >>> @@ -2702,7 +2702,6 @@ static struct msm_gpu *a6xx_gpu_init(struct
> >>> drm_device *dev)
> >>>ret = a6xx_gmu_wrapper_init(a6xx_gpu, node);
> >>>else
> >>>ret = a6xx_gmu_init(a6xx_gpu, node);
> >>> - of_node_put(node);
> >>>if (ret) {
> >>>a6xx_destroy(&(a6xx_gpu->base.base));
> >>>return ERR_PTR(ret);
> >>>
> >>> ---
> >>> base-commit: a0c666c25aeefd16f4b088c6549a6fb6b65a8a1d
> >>> change-id: 20260123-a6xx_gpu-cbc095dbe423
> >>>
> >>> Best regards,
> >>> --
> >>> Felix Gu
> >>>
> >>
> >> --
> >> With best wishes
> >> Dmitry
>
>
> --
> With best wishes
> Dmitry
Re: [PATCH] drm/msm/adreno: Fix a reference leak in a6xx_gpu_init()
On 23/01/2026 20:21, Felix Gu wrote:
Hi Dmitry,
Thanks for your review.
linux/cleanup.h is included in linux/of.h.
So it should be included already.
Yes. However we should not depend on include files being provided by
other kernel-wide includes.
Also, please don't top-post when replying to the emails on public MLs.
Best regards,
Feix Gu
On Sat, Jan 24, 2026 at 1:58 AM Dmitry Baryshkov
wrote:
On Sat, Jan 24, 2026 at 12:37:38AM +0800, Felix Gu wrote:
In a6xx_gpu_init(), node is obtained via of_parse_phandle().
While there was a manual of_node_put() at the end of the
common path, several early error returns would bypass this call,
resulting in a reference leak.
Fix this by using the __free(device_node) cleanup handler to
release the reference when the variable goes out of scope.
Fixes: 5a903a44a984 ("drm/msm/a6xx: Introduce GMU wrapper support")
Signed-off-by: Felix Gu
---
drivers/gpu/drm/msm/adreno/a6xx_gpu.c | 5 ++---
1 file changed, 2 insertions(+), 3 deletions(-)
diff --git a/drivers/gpu/drm/msm/adreno/a6xx_gpu.c
b/drivers/gpu/drm/msm/adreno/a6xx_gpu.c
index 2129d230a92b..0bc518d9fd65 100644
--- a/drivers/gpu/drm/msm/adreno/a6xx_gpu.c
+++ b/drivers/gpu/drm/msm/adreno/a6xx_gpu.c
@@ -2640,7 +2640,6 @@ static struct msm_gpu *a6xx_gpu_init(struct drm_device
*dev)
struct msm_drm_private *priv = dev->dev_private;
struct platform_device *pdev = priv->gpu_pdev;
struct adreno_platform_config *config = pdev->dev.platform_data;
- struct device_node *node;
struct a6xx_gpu *a6xx_gpu;
struct adreno_gpu *adreno_gpu;
struct msm_gpu *gpu;
@@ -2660,7 +2659,8 @@ static struct msm_gpu *a6xx_gpu_init(struct drm_device
*dev)
adreno_gpu->registers = NULL;
/* Check if there is a GMU phandle and set it up */
- node = of_parse_phandle(pdev->dev.of_node, "qcom,gmu", 0);
+ struct device_node *node __free(device_node) =
+ of_parse_phandle(pdev->dev.of_node, "qcom,gmu", 0);
Missing include for cleanup.h.
/* FIXME: How do we gracefully handle this? */
BUG_ON(!node);
@@ -2702,7 +2702,6 @@ static struct msm_gpu *a6xx_gpu_init(struct drm_device
*dev)
ret = a6xx_gmu_wrapper_init(a6xx_gpu, node);
else
ret = a6xx_gmu_init(a6xx_gpu, node);
- of_node_put(node);
if (ret) {
a6xx_destroy(&(a6xx_gpu->base.base));
return ERR_PTR(ret);
---
base-commit: a0c666c25aeefd16f4b088c6549a6fb6b65a8a1d
change-id: 20260123-a6xx_gpu-cbc095dbe423
Best regards,
--
Felix Gu
--
With best wishes
Dmitry
--
With best wishes
Dmitry
Re: [PATCH] drm/msm/adreno: Fix a reference leak in a6xx_gpu_init()
On Sat, Jan 24, 2026 at 12:37:38AM +0800, Felix Gu wrote:
> In a6xx_gpu_init(), node is obtained via of_parse_phandle().
> While there was a manual of_node_put() at the end of the
> common path, several early error returns would bypass this call,
> resulting in a reference leak.
> Fix this by using the __free(device_node) cleanup handler to
> release the reference when the variable goes out of scope.
>
> Fixes: 5a903a44a984 ("drm/msm/a6xx: Introduce GMU wrapper support")
> Signed-off-by: Felix Gu
> ---
> drivers/gpu/drm/msm/adreno/a6xx_gpu.c | 5 ++---
> 1 file changed, 2 insertions(+), 3 deletions(-)
>
> diff --git a/drivers/gpu/drm/msm/adreno/a6xx_gpu.c
> b/drivers/gpu/drm/msm/adreno/a6xx_gpu.c
> index 2129d230a92b..0bc518d9fd65 100644
> --- a/drivers/gpu/drm/msm/adreno/a6xx_gpu.c
> +++ b/drivers/gpu/drm/msm/adreno/a6xx_gpu.c
> @@ -2640,7 +2640,6 @@ static struct msm_gpu *a6xx_gpu_init(struct drm_device
> *dev)
> struct msm_drm_private *priv = dev->dev_private;
> struct platform_device *pdev = priv->gpu_pdev;
> struct adreno_platform_config *config = pdev->dev.platform_data;
> - struct device_node *node;
> struct a6xx_gpu *a6xx_gpu;
> struct adreno_gpu *adreno_gpu;
> struct msm_gpu *gpu;
> @@ -2660,7 +2659,8 @@ static struct msm_gpu *a6xx_gpu_init(struct drm_device
> *dev)
> adreno_gpu->registers = NULL;
>
> /* Check if there is a GMU phandle and set it up */
> - node = of_parse_phandle(pdev->dev.of_node, "qcom,gmu", 0);
> + struct device_node *node __free(device_node) =
> + of_parse_phandle(pdev->dev.of_node, "qcom,gmu", 0);
Missing include for cleanup.h.
> /* FIXME: How do we gracefully handle this? */
> BUG_ON(!node);
>
> @@ -2702,7 +2702,6 @@ static struct msm_gpu *a6xx_gpu_init(struct drm_device
> *dev)
> ret = a6xx_gmu_wrapper_init(a6xx_gpu, node);
> else
> ret = a6xx_gmu_init(a6xx_gpu, node);
> - of_node_put(node);
> if (ret) {
> a6xx_destroy(&(a6xx_gpu->base.base));
> return ERR_PTR(ret);
>
> ---
> base-commit: a0c666c25aeefd16f4b088c6549a6fb6b65a8a1d
> change-id: 20260123-a6xx_gpu-cbc095dbe423
>
> Best regards,
> --
> Felix Gu
>
--
With best wishes
Dmitry
