Re: [PATCH 3/5] mtd: Don't unconditionally unregister reboot notifier

[Brian Norris <computersforpeace@gmail.com>]

On Tue, Jul 05, 2016 at 12:06:21AM +0200, Richard Weinberger wrote:
> del_mtd_device() is allowed to fail.
> i.e. when the MTD is busy.
> Unregister the reboot notifier only when we're really
> about to delete the MTD.
> 
> Signed-off-by: Richard Weinberger <richard@nod.at>
> ---
>  drivers/mtd/mtdcore.c | 15 ++++++++++-----
>  1 file changed, 10 insertions(+), 5 deletions(-)
> 
> diff --git a/drivers/mtd/mtdcore.c b/drivers/mtd/mtdcore.c
> index e3936b8..36e5fb0 100644
> --- a/drivers/mtd/mtdcore.c
> +++ b/drivers/mtd/mtdcore.c
> @@ -654,17 +654,22 @@ int mtd_device_unregister(struct mtd_info *master)
>  {
>  	int err;
>  
> -	if (master->_reboot)
> -		unregister_reboot_notifier(&master->reboot_notifier);
> -
>  	err = del_mtd_partitions(master);
>  	if (err)
>  		return err;
>  
>  	if (!device_is_registered(&master->dev))
> -		return 0;
> +		goto unregister;
>  
> -	return del_mtd_device(master);
> +	err = del_mtd_device(master);
> +	if (err)
> +		return err;

Is there any kind of race issue with unregistering the notifier *after*
we've deleted the device? I had intentionally unregistered first,
because I didn't want any chance of the driver/module and/or data
structures being freed before we call the notifier.

I can't think of any particular issue yet, but I wanted to ask.

Brian

> +
> +unregister:
> +	if (master->_reboot)
> +		unregister_reboot_notifier(&master->reboot_notifier);
> +
> +	return 0;
>  }
>  EXPORT_SYMBOL_GPL(mtd_device_unregister);
>  
> -- 
> 2.7.3
>