Re: [PATCH 1/1] firmware_loading_store: fix firmware loading use-after-free oops

[Greg KH <greg@kroah.com>]

On Thu, Dec 01, 2011 at 05:26:10PM +0800, DONG-DONG YANG wrote:
> 
> firmware_loading_store() should check for fw_priv->fw before fw_load_abort.
> Otherwise, the system has a chance to run into use-after-free oops,
> if the user-space handler sends "-1" to firmware controller(loading) in
> exception scenarios before sending "1" to start a load.
> See the attachment for the kernel log and the analysis
> 
> Tested:
>   ARM_ca9 + android init uevent + webtop udev
> 
> Patch:
> diff --git a/drivers/base/firmware_class.c b/drivers/base/firmware_class.c
> index 06ed6b4..b333082 100644
> --- a/drivers/base/firmware_class.c
> +++ b/drivers/base/firmware_class.c
> @@ -271,7 +271,8 @@ static ssize_t firmware_loading_store(struct device *dev,
>                 /* fallthrough */
>         case -1:
>         err:
> -               fw_load_abort(fw_priv);
> +               if (fw_priv->fw)
> +                       fw_load_abort(fw_priv);
>                 break;
>         }
> 

Can you resend this in the format described in
Documentation/SubmittingPatches so that I can accept it?  (hint, you
forgot the signed-off-by: line.)

Also, shouldn't we do the check for ->fw in the fw_load_abort() call?

And why is userspace doing wierd things like this?  What causes that?

thanks,

greg k-h