[PATCH] fbdev: omapfb: fix reference leak on failed device registration

[PATCH] fbdev: omapfb: fix reference leak on failed device registration

[Guangshuo Li]

When platform_device_register() fails in omapfb_probe(), the embedded
struct device in omapdss_device has already been initialized by
device_initialize(), but the failure path only reports the error and
returns without dropping the device reference for the current platform
device:

  omapfb_probe()
    -> platform_device_register(&omapdss_device)
       -> device_initialize(&omapdss_device.dev)
       -> setup_pdev_dma_masks(&omapdss_device)
       -> platform_device_add(&omapdss_device)

This leads to a reference leak when platform_device_register() fails.
Fix this by calling platform_device_put() before returning the error.

The issue was identified by a static analysis tool I developed and
confirmed by manual review.

Fixes: f778a12dd3320 ("OMAP: OMAPFB: fix clk_get for RFBI")
Cc: stable@vger.kernel.org
Signed-off-by: Guangshuo Li <lgs201920130244@gmail.com>
---
 drivers/video/fbdev/omap/omapfb_main.c | 1 +
 1 file changed, 1 insertion(+)

diff --git a/drivers/video/fbdev/omap/omapfb_main.c b/drivers/video/fbdev/omap/omapfb_main.c
index cafe859d6e5a..0d47a8aec5c5 100644
--- a/drivers/video/fbdev/omap/omapfb_main.c
+++ b/drivers/video/fbdev/omap/omapfb_main.c
@@ -1768,6 +1768,7 @@ static int omapfb_probe(struct platform_device *pdev)
 	r = platform_device_register(&omapdss_device);
 	if (r) {
 		dev_err(&pdev->dev, "can't register omapdss device\n");
+		platform_device_put(&omapdss_device);
 		return r;
 	}
 
-- 
2.43.0

Re: [PATCH] fbdev: omapfb: fix reference leak on failed device registration

[Helge Deller]

Hello Guanghshuo,

On 4/15/26 21:17, Guangshuo Li wrote:
> When platform_device_register() fails in omapfb_probe(), the embedded
> struct device in omapdss_device has already been initialized by
> device_initialize(), but the failure path only reports the error and
> returns without dropping the device reference for the current platform
> device:
> 
>    omapfb_probe()
>      -> platform_device_register(&omapdss_device)
>         -> device_initialize(&omapdss_device.dev)
>         -> setup_pdev_dma_masks(&omapdss_device)
>         -> platform_device_add(&omapdss_device)
> 
> This leads to a reference leak when platform_device_register() fails.
> Fix this by calling platform_device_put() before returning the error.

I see you submitted quite some patches, all about the same issue.
I did not yet fully checked if there is really a reference leak, but even if
it would be, I think it's wrong that all the callers in the whole Linux kernel source
would now need to use platform_device_put(). To me it then seems as if everyone got it wrong.
IMHO if platform_device_register() fails it should put the device itself before
returning.
Just looking at generic code, see platform_add_devices() in drivers/base/platform.c
which seem to have it wrong too then...

Helge


> The issue was identified by a static analysis tool I developed and
> confirmed by manual review.
> 
> Fixes: f778a12dd3320 ("OMAP: OMAPFB: fix clk_get for RFBI")
> Cc: stable@vger.kernel.org
> Signed-off-by: Guangshuo Li <lgs201920130244@gmail.com>
> ---
>   drivers/video/fbdev/omap/omapfb_main.c | 1 +
>   1 file changed, 1 insertion(+)
> 
> diff --git a/drivers/video/fbdev/omap/omapfb_main.c b/drivers/video/fbdev/omap/omapfb_main.c
> index cafe859d6e5a..0d47a8aec5c5 100644
> --- a/drivers/video/fbdev/omap/omapfb_main.c
> +++ b/drivers/video/fbdev/omap/omapfb_main.c
> @@ -1768,6 +1768,7 @@ static int omapfb_probe(struct platform_device *pdev)
>   	r = platform_device_register(&omapdss_device);
>   	if (r) {
>   		dev_err(&pdev->dev, "can't register omapdss device\n");
> +		platform_device_put(&omapdss_device);
>   		return r;
>   	}
>