[PATCH] media: davinci: vpif_display: Mix memory leak on probe error path

[PATCH] media: davinci: vpif_display: Mix memory leak on probe error path

[Anton Vasilyev]

If vpif_probe() fails on v4l2_device_register() then memory allocated
at initialize_vpif() for global vpif_obj.dev[i] become unreleased.

The patch adds deallocation of vpif_obj.dev[i] on the error path and
removes duplicated check on platform_data presence.

Found by Linux Driver Verification project (linuxtesting.org).

Signed-off-by: Anton Vasilyev <vasilyev@ispras.ru>
---
 drivers/media/platform/davinci/vpif_display.c | 24 ++++++++++++-------
 1 file changed, 16 insertions(+), 8 deletions(-)

diff --git a/drivers/media/platform/davinci/vpif_display.c b/drivers/media/platform/davinci/vpif_display.c
index 7be636237acf..0f324055cc9f 100644
--- a/drivers/media/platform/davinci/vpif_display.c
+++ b/drivers/media/platform/davinci/vpif_display.c
@@ -1114,6 +1114,14 @@ static int initialize_vpif(void)
 	return err;
 }
 
+static void free_vpif_objs(void)
+{
+	int i;
+
+	for (i = 0; i < VPIF_DISPLAY_MAX_DEVICES; i++)
+		kfree(vpif_obj.dev[i]);
+}
+
 static int vpif_async_bound(struct v4l2_async_notifier *notifier,
 			    struct v4l2_subdev *subdev,
 			    struct v4l2_async_subdev *asd)
@@ -1255,11 +1263,6 @@ static __init int vpif_probe(struct platform_device *pdev)
 		return -EINVAL;
 	}
 
-	if (!pdev->dev.platform_data) {
-		dev_warn(&pdev->dev, "Missing platform data.  Giving up.\n");
-		return -EINVAL;
-	}
-
 	vpif_dev = &pdev->dev;
 	err = initialize_vpif();
 
@@ -1271,7 +1274,7 @@ static __init int vpif_probe(struct platform_device *pdev)
 	err = v4l2_device_register(vpif_dev, &vpif_obj.v4l2_dev);
 	if (err) {
 		v4l2_err(vpif_dev->driver, "Error registering v4l2 device\n");
-		return err;
+		goto vpif_free;
 	}
 
 	while ((res = platform_get_resource(pdev, IORESOURCE_IRQ, res_idx))) {
@@ -1314,7 +1317,10 @@ static __init int vpif_probe(struct platform_device *pdev)
 			if (vpif_obj.sd[i])
 				vpif_obj.sd[i]->grp_id = 1 << i;
 		}
-		vpif_probe_complete();
+		err = vpif_probe_complete();
+		if (err) {
+			goto probe_subdev_out;
+		}
 	} else {
 		vpif_obj.notifier.subdevs = vpif_obj.config->asd;
 		vpif_obj.notifier.num_subdevs = vpif_obj.config->asd_sizes[0];
@@ -1334,6 +1340,8 @@ static __init int vpif_probe(struct platform_device *pdev)
 	kfree(vpif_obj.sd);
 vpif_unregister:
 	v4l2_device_unregister(&vpif_obj.v4l2_dev);
+vpif_free:
+	free_vpif_objs();
 
 	return err;
 }
@@ -1355,8 +1363,8 @@ static int vpif_remove(struct platform_device *device)
 		ch = vpif_obj.dev[i];
 		/* Unregister video device */
 		video_unregister_device(&ch->video_dev);
-		kfree(vpif_obj.dev[i]);
 	}
+	free_vpif_objs();
 
 	return 0;
 }
-- 
2.18.0

Re: [PATCH] media: davinci: vpif_display: Mix memory leak on probe error path

[Lad, Prabhakar]

Hi,

thank you for the patch.

On Fri, Jul 27, 2018 at 12:52 PM, Anton Vasilyev <vasilyev@ispras.ru> wrote:
> If vpif_probe() fails on v4l2_device_register() then memory allocated
> at initialize_vpif() for global vpif_obj.dev[i] become unreleased.
>
> The patch adds deallocation of vpif_obj.dev[i] on the error path and
> removes duplicated check on platform_data presence.
>
> Found by Linux Driver Verification project (linuxtesting.org).
>
> Signed-off-by: Anton Vasilyev <vasilyev@ispras.ru>
> ---
>  drivers/media/platform/davinci/vpif_display.c | 24 ++++++++++++-------
>  1 file changed, 16 insertions(+), 8 deletions(-)
>
> diff --git a/drivers/media/platform/davinci/vpif_display.c b/drivers/media/platform/davinci/vpif_display.c
> index 7be636237acf..0f324055cc9f 100644
> --- a/drivers/media/platform/davinci/vpif_display.c
> +++ b/drivers/media/platform/davinci/vpif_display.c
> @@ -1114,6 +1114,14 @@ static int initialize_vpif(void)
>         return err;
>  }
>
> +static void free_vpif_objs(void)
> +{
> +       int i;
> +
> +       for (i = 0; i < VPIF_DISPLAY_MAX_DEVICES; i++)
> +               kfree(vpif_obj.dev[i]);
> +}
> +
>  static int vpif_async_bound(struct v4l2_async_notifier *notifier,
>                             struct v4l2_subdev *subdev,
>                             struct v4l2_async_subdev *asd)
> @@ -1255,11 +1263,6 @@ static __init int vpif_probe(struct platform_device *pdev)
>                 return -EINVAL;
>         }
>
> -       if (!pdev->dev.platform_data) {
> -               dev_warn(&pdev->dev, "Missing platform data.  Giving up.\n");
> -               return -EINVAL;
> -       }
> -
Could make this as a separate patch.

>         vpif_dev = &pdev->dev;
>         err = initialize_vpif();
>
> @@ -1271,7 +1274,7 @@ static __init int vpif_probe(struct platform_device *pdev)
>         err = v4l2_device_register(vpif_dev, &vpif_obj.v4l2_dev);
>         if (err) {
>                 v4l2_err(vpif_dev->driver, "Error registering v4l2 device\n");
> -               return err;
> +               goto vpif_free;
>         }
>
>         while ((res = platform_get_resource(pdev, IORESOURCE_IRQ, res_idx))) {
> @@ -1314,7 +1317,10 @@ static __init int vpif_probe(struct platform_device *pdev)
>                         if (vpif_obj.sd[i])
>                                 vpif_obj.sd[i]->grp_id = 1 << i;
>                 }
> -               vpif_probe_complete();
> +               err = vpif_probe_complete();
> +               if (err) {
> +                       goto probe_subdev_out;
> +               }

{} braces are not needed

>         } else {
>                 vpif_obj.notifier.subdevs = vpif_obj.config->asd;
>                 vpif_obj.notifier.num_subdevs = vpif_obj.config->asd_sizes[0];
> @@ -1334,6 +1340,8 @@ static __init int vpif_probe(struct platform_device *pdev)
>         kfree(vpif_obj.sd);
>  vpif_unregister:
>         v4l2_device_unregister(&vpif_obj.v4l2_dev);
> +vpif_free:
> +       free_vpif_objs();
>
Just put the for loop here instead.

>         return err;
>  }
> @@ -1355,8 +1363,8 @@ static int vpif_remove(struct platform_device *device)
>                 ch = vpif_obj.dev[i];
>                 /* Unregister video device */
>                 video_unregister_device(&ch->video_dev);
> -               kfree(vpif_obj.dev[i]);
>         }
> +       free_vpif_objs();
>
Just leave this as is, as its already looping and freeing up the objects.

>         return 0;
>  }
> --
> 2.18.0
>

Cheers,
--Prabhakar Lad

[PATCH v2 1/2] media: davinci: vpif_display: Mix memory leak on probe error path

[Anton Vasilyev]

If vpif_probe() fails on v4l2_device_register() then memory allocated
at initialize_vpif() for global vpif_obj.dev[i] become unreleased.

The patch adds deallocation of vpif_obj.dev[i] on the probe error path.

Found by Linux Driver Verification project (linuxtesting.org).

Signed-off-by: Anton Vasilyev <vasilyev@ispras.ru>
---
v2: divided the original patch into two and made stylistic fixes based
on the Prabhakar's rewiev.
---
 drivers/media/platform/davinci/vpif_display.c | 9 +++++++--
 1 file changed, 7 insertions(+), 2 deletions(-)

diff --git a/drivers/media/platform/davinci/vpif_display.c b/drivers/media/platform/davinci/vpif_display.c
index 7be636237acf..d9e578ac79c2 100644
--- a/drivers/media/platform/davinci/vpif_display.c
+++ b/drivers/media/platform/davinci/vpif_display.c
@@ -1271,7 +1271,7 @@ static __init int vpif_probe(struct platform_device *pdev)
 	err = v4l2_device_register(vpif_dev, &vpif_obj.v4l2_dev);
 	if (err) {
 		v4l2_err(vpif_dev->driver, "Error registering v4l2 device\n");
-		return err;
+		goto vpif_free;
 	}
 
 	while ((res = platform_get_resource(pdev, IORESOURCE_IRQ, res_idx))) {
@@ -1314,7 +1314,9 @@ static __init int vpif_probe(struct platform_device *pdev)
 			if (vpif_obj.sd[i])
 				vpif_obj.sd[i]->grp_id = 1 << i;
 		}
-		vpif_probe_complete();
+		err = vpif_probe_complete();
+		if (err)
+			goto probe_subdev_out;
 	} else {
 		vpif_obj.notifier.subdevs = vpif_obj.config->asd;
 		vpif_obj.notifier.num_subdevs = vpif_obj.config->asd_sizes[0];
@@ -1334,6 +1336,9 @@ static __init int vpif_probe(struct platform_device *pdev)
 	kfree(vpif_obj.sd);
 vpif_unregister:
 	v4l2_device_unregister(&vpif_obj.v4l2_dev);
+vpif_free:
+	for (i = 0; i < VPIF_DISPLAY_MAX_DEVICES; i++)
+		kfree(vpif_obj.dev[i]);
 
 	return err;
 }
-- 
2.18.0

[PATCH v2 2/2] media: davinci: vpif_display: remove duplicate check

[Anton Vasilyev]

The patch removes the duplicate platform_data check from vpif_probe().

Fixes: 4a5f8ae50b66 ("[media] davinci: vpif_capture: get subdevs from DT when available")

Found by Linux Driver Verification project (linuxtesting.org).

Signed-off-by: Anton Vasilyev <vasilyev@ispras.ru>
---
v2: divided the original patch into two and made stylistic fixes based
on the Prabhakar's rewiev.
---
 drivers/media/platform/davinci/vpif_display.c | 5 -----
 1 file changed, 5 deletions(-)

diff --git a/drivers/media/platform/davinci/vpif_display.c b/drivers/media/platform/davinci/vpif_display.c
index d9e578ac79c2..f000fc492ef9 100644
--- a/drivers/media/platform/davinci/vpif_display.c
+++ b/drivers/media/platform/davinci/vpif_display.c
@@ -1255,11 +1255,6 @@ static __init int vpif_probe(struct platform_device *pdev)
 		return -EINVAL;
 	}
 
-	if (!pdev->dev.platform_data) {
-		dev_warn(&pdev->dev, "Missing platform data.  Giving up.\n");
-		return -EINVAL;
-	}
-
 	vpif_dev = &pdev->dev;
 	err = initialize_vpif();
 
-- 
2.18.0

Re: [PATCH v2 2/2] media: davinci: vpif_display: remove duplicate check

[Lad, Prabhakar]

Hi,

Thanks for the patch.

On Mon, Aug 6, 2018 at 4:51 PM Anton Vasilyev <vasilyev@ispras.ru> wrote:
>
> The patch removes the duplicate platform_data check from vpif_probe().
>
> Fixes: 4a5f8ae50b66 ("[media] davinci: vpif_capture: get subdevs from DT when available")
>
> Found by Linux Driver Verification project (linuxtesting.org).
>
> Signed-off-by: Anton Vasilyev <vasilyev@ispras.ru>
> ---
> v2: divided the original patch into two and made stylistic fixes based
> on the Prabhakar's rewiev.
> ---
>  drivers/media/platform/davinci/vpif_display.c | 5 -----
>  1 file changed, 5 deletions(-)
>
Acked-by: Lad, Prabhakar <prabhakar.csengg@gmail.com>

Cheers,
--Prabhakar Lad

Re: [PATCH v2 1/2] media: davinci: vpif_display: Mix memory leak on probe error path

[Lad, Prabhakar]

Hi,

Thanks for the patch.

On Mon, Aug 6, 2018 at 4:50 PM Anton Vasilyev <vasilyev@ispras.ru> wrote:
>
> If vpif_probe() fails on v4l2_device_register() then memory allocated
> at initialize_vpif() for global vpif_obj.dev[i] become unreleased.
>
> The patch adds deallocation of vpif_obj.dev[i] on the probe error path.
>
> Found by Linux Driver Verification project (linuxtesting.org).
>
> Signed-off-by: Anton Vasilyev <vasilyev@ispras.ru>
> ---
> v2: divided the original patch into two and made stylistic fixes based
> on the Prabhakar's rewiev.
> ---
>  drivers/media/platform/davinci/vpif_display.c | 9 +++++++--
>  1 file changed, 7 insertions(+), 2 deletions(-)
>
Acked-by: Lad, Prabhakar <prabhakar.csengg@gmail.com>

Cheers,
--Prabhakar Lad