[PATCH] media: atomisp: Fix resource leak in atomisp_pci_probe()

[PATCH] media: atomisp: Fix resource leak in atomisp_pci_probe()

[Dawei Feng]

During atomisp_pci_probe(), the ISP subdev is initialized via
atomisp_initialize_modules() prior to entity registration. If
atomisp_register_entities() fails, the current error path only
uninitializes the CSI2 modules. This leaks the subdev entity and control
handler that were previously set up by atomisp_subdev_init().

Fix this by calling atomisp_subdev_unregister_entities() to properly
release the subdev state on this specific error path. Later error paths
remain unchanged, as they correctly use atomisp_unregister_entities() to
handle broader cleanup after successful registration.

The bug was first flagged by an experimental analysis tool we are
developing for kernel memory-management bugs while analyzing v6.13-rc1.
The tool is still under development and is not yet publicly available.
Manual inspection confirms that the bug is still present in v7.1-rc5.

An x86_64 allyesconfig build showed no new warnings. As we do not have an
Intel Atom ISP platform to test with, no runtime testing was able to be
performed.

Fixes: 9d4fa1a16b28 ("media: atomisp: cleanup directory hierarchy")
Signed-off-by: Zilin Guan <zilin@seu.edu.cn>
Signed-off-by: Dawei Feng <dawei.feng@seu.edu.cn>
---
 drivers/staging/media/atomisp/pci/atomisp_v4l2.c | 1 +
 1 file changed, 1 insertion(+)

diff --git a/drivers/staging/media/atomisp/pci/atomisp_v4l2.c b/drivers/staging/media/atomisp/pci/atomisp_v4l2.c
index 900a67552d6a..d4e4e845f66e 100644
--- a/drivers/staging/media/atomisp/pci/atomisp_v4l2.c
+++ b/drivers/staging/media/atomisp/pci/atomisp_v4l2.c
@@ -1401,6 +1401,7 @@ static int atomisp_pci_probe(struct pci_dev *pdev, const struct pci_device_id *i
 	err = atomisp_register_entities(isp);
 	if (err < 0) {
 		dev_err(&pdev->dev, "atomisp_register_entities failed (%d)\n", err);
+		atomisp_subdev_unregister_entities(&isp->asd);
 		goto error_uninitialize_modules;
 	}
 
-- 
2.34.1

Re: [PATCH] media: atomisp: Fix resource leak in atomisp_pci_probe()

[Dan Carpenter]

On Mon, Jun 08, 2026 at 04:27:06PM +0800, Dawei Feng wrote:
> During atomisp_pci_probe(), the ISP subdev is initialized via
> atomisp_initialize_modules() prior to entity registration. If
> atomisp_register_entities() fails, the current error path only
> uninitializes the CSI2 modules. This leaks the subdev entity and control
> handler that were previously set up by atomisp_subdev_init().
> 
> Fix this by calling atomisp_subdev_unregister_entities() to properly
> release the subdev state on this specific error path. Later error paths
> remain unchanged, as they correctly use atomisp_unregister_entities() to
> handle broader cleanup after successful registration.
> 
> The bug was first flagged by an experimental analysis tool we are
> developing for kernel memory-management bugs while analyzing v6.13-rc1.
> The tool is still under development and is not yet publicly available.
> Manual inspection confirms that the bug is still present in v7.1-rc5.
> 
> An x86_64 allyesconfig build showed no new warnings. As we do not have an
> Intel Atom ISP platform to test with, no runtime testing was able to be
> performed.
> 
> Fixes: 9d4fa1a16b28 ("media: atomisp: cleanup directory hierarchy")
> Signed-off-by: Zilin Guan <zilin@seu.edu.cn>
> Signed-off-by: Dawei Feng <dawei.feng@seu.edu.cn>
> ---

The code is buggy, but this isn't the right fix.

Here is generally the standard way to do error handling.
https://staticthinking.wordpress.com/2022/04/28/free-the-last-thing-style/

1. An allocation should clean up it's own partial allocations.  That
should not be handled in the caller.  2.  Every allocation function should
have a mirror cleanup function.

The atomisp_uninitialize_modules() function is just a dummy and was never
actually implemented.  The correct thing is to implement it.

regards,
dan carpenter

Re: [PATCH] media: atomisp: Fix resource leak in atomisp_pci_probe()

[Dawei Feng]

Hi Dan,

On Mon, Jun 08, 2026 at 11:51:45 +0300, Dan Carpenter wrote:
> The code is buggy, but this isn't the right fix.
> 
> Here is generally the standard way to do error handling.
> https://staticthinking.wordpress.com/2022/04/28/free-the-last-thing-style/
> 
> 1. An allocation should clean up it's own partial allocations.  That
> should not be handled in the caller.  2.  Every allocation function should
> have a mirror cleanup function.
> 
> The atomisp_uninitialize_modules() function is just a dummy and was never
> actually implemented.  The correct thing is to implement it.

Thanks for the review and the link to the error handling guidelines. 

Thanks for the pointer. I'll update it in v2.

Thanks,
Dawei

Re: [PATCH] media: atomisp: Fix resource leak in atomisp_pci_probe()

[sashiko-bot]

Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider:

Pre-existing issues:
- [High] The v4l2 async notifier is never unregistered or cleaned up, leading to a memory leak of async subdevice connection objects and dangling notifiers resulting in a Use-After-Free.
--

commit b7444e6b542badbafff9a7d15d5037c35782ba7c
Author: Dawei Feng <dawei.feng@seu.edu.cn>

media: atomisp: Fix resource leak in atomisp_pci_probe()

This commit fixes a resource leak by unregistering subdev entities if
atomisp_register_entities() fails during atomisp_pci_probe().

> diff --git a/drivers/staging/media/atomisp/pci/atomisp_v4l2.c b/drivers/staging/media/atomisp/pci/atomisp_v4l2.c
> index 900a67552d6a0..d4e4e845f66e6 100644
> --- a/drivers/staging/media/atomisp/pci/atomisp_v4l2.c
> +++ b/drivers/staging/media/atomisp/pci/atomisp_v4l2.c
> @@ -1401,6 +1401,7 @@ static int atomisp_pci_probe(struct pci_dev *pdev, const struct pci_device_id *i
>  	err = atomisp_register_entities(isp);
>  	if (err < 0) {
>  		dev_err(&pdev->dev, "atomisp_register_entities failed (%d)\n", err);
> +		atomisp_subdev_unregister_entities(&isp->asd);
>  		goto error_uninitialize_modules;
>  	}

[Severity: High]
This is a pre-existing issue, but while looking at the error paths being
fixed here, is there also a memory leak and use-after-free risk related to
the async notifier?

In atomisp_pci_probe(), the notifier is registered via
v4l2_async_nf_register(), adding it to the global notifier_list. However,
the error paths and the driver teardown function atomisp_pci_remove()
appear to entirely omit v4l2_async_nf_unregister() and
v4l2_async_nf_cleanup():

drivers/staging/media/atomisp/pci/atomisp_v4l2.c:atomisp_pci_probe() {
    ...
	err = v4l2_async_nf_register(&isp->notifier);
	if (err) {
		dev_err(isp->dev, "failed to register async notifier : %d\n", err);
		goto error_unload_firmware;
	}

	return 0;

error_unload_firmware:
	atomisp_pm_uninit(isp);
	ia_css_unload_firmware();
    ...
}

Does this code permanently leak the memory allocated for async subdevices via
v4l2_async_nf_add_fwnode_remote() because cleanup is never called?

Additionally, since the isp struct is allocated via devm_kzalloc(), could
omitting the unregister call leave a dangling pointer in the global
notifier_list when the driver is unloaded or when the probe fails, potentially
causing a use-after-free on subsequent list iterations?

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20260608082706.3287831-1-dawei.feng@seu.edu.cn?part=1