On Fri, Aug 14, 2026 at 09:00:56PM +0000, Tarun Sahu wrote:
> As per koject_init_and_add() function kernel document, even if this
> function returns error kobject_put must be used instead of kfree.

Thanks for adding this specific pointer.  I wouldn't repost just for
these nits, and wait at least a few days before reposting for any
reason.  But if you do repost:

s/koject_init_and_add/kobject_init_and_add/

Might also add "()" after function names consistently (kobject_put and
kfree above, edd_release below).  Also applies to the other patches.

The current subject lines basically restate the C code; you might
consider more of a focus on the problem.  I ran this through gemini
and I think it did a decent job:

  firmware: edd: Fix kobject reference leak on registration failure

  Per kobject_init_and_add() kernel-doc, calling kfree() directly on
  error bypasses reference counting and skips the kobject's release
  callback, leaking the reference.

  Use kobject_put() instead of kfree() on registration failure to fix
  this.

> When edd_device_register() fails after initializing the kobject with
> kobject_init_and_add(), calling kfree(edev) directly bypasses the
> kobject release callback (edd_release) and leaks the allocated kobject
> resources.
> 
> Fix this by replacing direct kfree(edev) with kobject_put(&edev->kobj) on
> registration failure.
> 
> Signed-off-by: Tarun Sahu <[email protected]>
> Reviewed-by: Sourabh Jain <[email protected]>
> ---
>  drivers/firmware/edd.c | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/drivers/firmware/edd.c b/drivers/firmware/edd.c
> index f980c5b56858..763e7b16d517 100644
> --- a/drivers/firmware/edd.c
> +++ b/drivers/firmware/edd.c
> @@ -748,7 +748,7 @@ edd_init(void)
>  
>               rc = edd_device_register(edev, i);
>               if (rc) {
> -                     kfree(edev);
> +                     kobject_put(&edev->kobj);
>                       goto out;
>               }
>               edd_devices[i] = edev;
> -- 
> 2.55.0.691.gc56d675ccc-goog
> 

Reply via email to