From: Wenkai Lin <linwenkai6@hisilicon.com>

In uacce_register(), if cdev_device_add() fails, we should properly
decrease the reference count of the cdev kobject and set uacce->cdev
to NULL to avoid potential use-after-free or double free issues.

This change adds proper error handling after cdev_device_add() fails,
ensuring that kobject_put() is called and uacce->cdev is cleared.

Fixes: 015d239ac014 ("uacce: add uacce driver")
Cc: stable@vger.kernel.org
Signed-off-by: Wenkai Lin <linwenkai6@hisilicon.com>
Signed-off-by: Chenghai Huang <huangchenghai2@huawei.com>
---
 drivers/misc/uacce/uacce.c | 11 ++++++++++-
 1 file changed, 10 insertions(+), 1 deletion(-)

diff --git a/drivers/misc/uacce/uacce.c b/drivers/misc/uacce/uacce.c
index 42e7d2a2a90c..9b82a6731832 100644
--- a/drivers/misc/uacce/uacce.c
+++ b/drivers/misc/uacce/uacce.c
@@ -519,6 +519,8 @@ EXPORT_SYMBOL_GPL(uacce_alloc);
  */
 int uacce_register(struct uacce_device *uacce)
 {
+	int ret;
+
 	if (!uacce)
 		return -ENODEV;
 
@@ -529,7 +531,14 @@ int uacce_register(struct uacce_device *uacce)
 	uacce->cdev->ops = &uacce_fops;
 	uacce->cdev->owner = THIS_MODULE;
 
-	return cdev_device_add(uacce->cdev, &uacce->dev);
+	ret = cdev_device_add(uacce->cdev, &uacce->dev);
+	if (ret) {
+		kobject_put(&uacce->cdev->kobj);
+		uacce->cdev = NULL;
+		return ret;
+	}
+
+	return 0;
 }
 EXPORT_SYMBOL_GPL(uacce_register);
 
-- 
2.33.0

On Wed, 22 Oct 2025 at 10:12, Chenghai Huang <huangchenghai2@huawei.com> wrote:
>
> From: Wenkai Lin <linwenkai6@hisilicon.com>
>
> In uacce_register(), if cdev_device_add() fails, we should properly
> decrease the reference count of the cdev kobject and set uacce->cdev
> to NULL to avoid potential use-after-free or double free issues.
>
> This change adds proper error handling after cdev_device_add() fails,
> ensuring that kobject_put() is called and uacce->cdev is cleared.
>
> Fixes: 015d239ac014 ("uacce: add uacce driver")
> Cc: stable@vger.kernel.org
> Signed-off-by: Wenkai Lin <linwenkai6@hisilicon.com>
> Signed-off-by: Chenghai Huang <huangchenghai2@huawei.com>
> ---
>  drivers/misc/uacce/uacce.c | 11 ++++++++++-
>  1 file changed, 10 insertions(+), 1 deletion(-)
>
> diff --git a/drivers/misc/uacce/uacce.c b/drivers/misc/uacce/uacce.c
> index 42e7d2a2a90c..9b82a6731832 100644
> --- a/drivers/misc/uacce/uacce.c
> +++ b/drivers/misc/uacce/uacce.c
> @@ -519,6 +519,8 @@ EXPORT_SYMBOL_GPL(uacce_alloc);
>   */
>  int uacce_register(struct uacce_device *uacce)
>  {
> +       int ret;
> +
>         if (!uacce)
>                 return -ENODEV;
>
> @@ -529,7 +531,14 @@ int uacce_register(struct uacce_device *uacce)
>         uacce->cdev->ops = &uacce_fops;
>         uacce->cdev->owner = THIS_MODULE;
>
> -       return cdev_device_add(uacce->cdev, &uacce->dev);
> +       ret = cdev_device_add(uacce->cdev, &uacce->dev);
> +       if (ret) {
> +               kobject_put(&uacce->cdev->kobj);

Not understanding this,
What if kobject_put(&p->kobj) is called twice?
cdev_device_add would call cdev_del->kobject_put(&p->kobj) as well.

If you think cdev_device_add has an issue, maybe fixing
cdev_device_add itself is better?

Thanks

在 2025/10/22 14:17, Zhangfei Gao 写道:
> On Wed, 22 Oct 2025 at 10:12, Chenghai Huang <huangchenghai2@huawei.com> wrote:
>> From: Wenkai Lin <linwenkai6@hisilicon.com>
>>
>> In uacce_register(), if cdev_device_add() fails, we should properly
>> decrease the reference count of the cdev kobject and set uacce->cdev
>> to NULL to avoid potential use-after-free or double free issues.
>>
>> This change adds proper error handling after cdev_device_add() fails,
>> ensuring that kobject_put() is called and uacce->cdev is cleared.
>>
>> Fixes: 015d239ac014 ("uacce: add uacce driver")
>> Cc: stable@vger.kernel.org
>> Signed-off-by: Wenkai Lin <linwenkai6@hisilicon.com>
>> Signed-off-by: Chenghai Huang <huangchenghai2@huawei.com>
>> ---
>>   drivers/misc/uacce/uacce.c | 11 ++++++++++-
>>   1 file changed, 10 insertions(+), 1 deletion(-)
>>
>> diff --git a/drivers/misc/uacce/uacce.c b/drivers/misc/uacce/uacce.c
>> index 42e7d2a2a90c..9b82a6731832 100644
>> --- a/drivers/misc/uacce/uacce.c
>> +++ b/drivers/misc/uacce/uacce.c
>> @@ -519,6 +519,8 @@ EXPORT_SYMBOL_GPL(uacce_alloc);
>>    */
>>   int uacce_register(struct uacce_device *uacce)
>>   {
>> +       int ret;
>> +
>>          if (!uacce)
>>                  return -ENODEV;
>>
>> @@ -529,7 +531,14 @@ int uacce_register(struct uacce_device *uacce)
>>          uacce->cdev->ops = &uacce_fops;
>>          uacce->cdev->owner = THIS_MODULE;
>>
>> -       return cdev_device_add(uacce->cdev, &uacce->dev);
>> +       ret = cdev_device_add(uacce->cdev, &uacce->dev);
>> +       if (ret) {
>> +               kobject_put(&uacce->cdev->kobj);
> Not understanding this,
> What if kobject_put(&p->kobj) is called twice?
> cdev_device_add would call cdev_del->kobject_put(&p->kobj) as well.
>
> If you think cdev_device_add has an issue, maybe fixing
> cdev_device_add itself is better?
>
> Thanks
After more testing, a new conclusion was reached.

After the cdev_set_parent function in cdev_device_add is called,
the cdev and the device will share the same lifetime and the memory
of the cdev is managed by the device.
So when put_device is called in the uacce_remove, the memory of the cdev
will also be released, so there is no memory leak issue in the code.

In addition, when cdev_device_add fails, a flag should be set to detect
this situation to prevent the uacce_remove from incorrectly calling
cdev_device_del to release the unallocated resources.
This issue will be fixed in the next patch.

Thanks,
Wenkai

在 2025/10/22 14:17, Zhangfei Gao 写道:
> On Wed, 22 Oct 2025 at 10:12, Chenghai Huang <huangchenghai2@huawei.com> wrote:
>> From: Wenkai Lin <linwenkai6@hisilicon.com>
>>
>> In uacce_register(), if cdev_device_add() fails, we should properly
>> decrease the reference count of the cdev kobject and set uacce->cdev
>> to NULL to avoid potential use-after-free or double free issues.
>>
>> This change adds proper error handling after cdev_device_add() fails,
>> ensuring that kobject_put() is called and uacce->cdev is cleared.
>>
>> Fixes: 015d239ac014 ("uacce: add uacce driver")
>> Cc: stable@vger.kernel.org
>> Signed-off-by: Wenkai Lin <linwenkai6@hisilicon.com>
>> Signed-off-by: Chenghai Huang <huangchenghai2@huawei.com>
>> ---
>>   drivers/misc/uacce/uacce.c | 11 ++++++++++-
>>   1 file changed, 10 insertions(+), 1 deletion(-)
>>
>> diff --git a/drivers/misc/uacce/uacce.c b/drivers/misc/uacce/uacce.c
>> index 42e7d2a2a90c..9b82a6731832 100644
>> --- a/drivers/misc/uacce/uacce.c
>> +++ b/drivers/misc/uacce/uacce.c
>> @@ -519,6 +519,8 @@ EXPORT_SYMBOL_GPL(uacce_alloc);
>>    */
>>   int uacce_register(struct uacce_device *uacce)
>>   {
>> +       int ret;
>> +
>>          if (!uacce)
>>                  return -ENODEV;
>>
>> @@ -529,7 +531,14 @@ int uacce_register(struct uacce_device *uacce)
>>          uacce->cdev->ops = &uacce_fops;
>>          uacce->cdev->owner = THIS_MODULE;
>>
>> -       return cdev_device_add(uacce->cdev, &uacce->dev);
>> +       ret = cdev_device_add(uacce->cdev, &uacce->dev);
>> +       if (ret) {
>> +               kobject_put(&uacce->cdev->kobj);
> Not understanding this,
> What if kobject_put(&p->kobj) is called twice?
> cdev_device_add would call cdev_del->kobject_put(&p->kobj) as well.
>
> If you think cdev_device_add has an issue, maybe fixing
> cdev_device_add itself is better?
>
> Thanks
Regarding your concern, cdev_device_add() use cdev_del rolls back
its cdev_add operations (add the refcount to 2) upon failure when
dev->devt is not null, the refcount of cdev is still 1 after the
rollback, so it does not free the memory of the cdev structure itself.

Therefore, in the error path of uacce_register, we still need to
manually call kobject_put to decrement the refcount of the kobject to
zero within cdev and free it.

Thanks,
Wenkai