Re: [PATCH] PNP: add put_device() in pnpbios_init()

[Greg KH <gregkh@linuxfoundation.org>]

On Thu, Dec 18, 2025 at 04:19:55PM +0800, Haoxiang Li wrote:
> If pnp_register_protocol() fails, call put_device()
> to drop the device reference.
> 
> Fixes: 1da177e4c3f4 ("Linux-2.6.12-rc2")
> Cc: stable@vger.kernel.org
> Signed-off-by: Haoxiang Li <lihaoxiang@isrc.iscas.ac.cn>
> ---
>  drivers/pnp/pnpbios/core.c | 1 +
>  1 file changed, 1 insertion(+)
> 
> diff --git a/drivers/pnp/pnpbios/core.c b/drivers/pnp/pnpbios/core.c
> index f7e86ae9f72f..997e0153d6e3 100644
> --- a/drivers/pnp/pnpbios/core.c
> +++ b/drivers/pnp/pnpbios/core.c
> @@ -538,6 +538,7 @@ static int __init pnpbios_init(void)
>  	/* register with the pnp layer */
>  	ret = pnp_register_protocol(&pnpbios_protocol);
>  	if (ret) {
> +		put_device(&pnpbios_protocol.dev)

This feels wrong, as this is a static structure, why would it have a
reference count?  Something is not working properly here if this is
required.

What testing caused this to trigger?  How was this found?

thanks,

greg k-h