Re: [PATCH 1/1] poweroff: change orderly_poweroff() to use schedule_work()

[Andrew Morton <akpm@linux-foundation.org>]

On Wed, 13 Mar 2013 18:47:05 +0100 Oleg Nesterov <oleg@redhat.com> wrote:

> orderly_poweroff() can be used from any context but UMH_WAIT_EXEC
> is sleepable. Move the "force" logic into __orderly_poweroff() and
> change orderly_poweroff() to use the global poweroff_work which
> simply calls __orderly_poweroff().
> 
> While at it, remove the unneeded "int argc" and change argv_split()
> to use GFP_KERNEL.
> 
> We use the global "bool poweroff_force" to pass the argument, this
> can obviously affect the previous request if it is pending/running.
> So we only allow the "false => true" transition assuming that the
> pending "true" should succeed anyway. If schedule_work() fails after
> that we know that work->func() was not called yet, it must see the
> new value.
> 
> This means that orderly_poweroff() becomes async even if we do not
> run the command and always succeeds, schedule_work() can only fail
> if the work is already pending. We can export __orderly_poweroff()
> and change the non-atomic callers which want the old semantics.
> 
> ...
>
> @@ -2218,21 +2237,9 @@ static int __orderly_poweroff(void)
>   */
>  int orderly_poweroff(bool force)
>  {
> -	int ret = __orderly_poweroff();
> -
> -	if (ret && force) {
> -		printk(KERN_WARNING "Failed to start orderly shutdown: "
> -		       "forcing the issue\n");
> -
> -		/*
> -		 * I guess this should try to kick off some daemon to sync and
> -		 * poweroff asap.  Or not even bother syncing if we're doing an
> -		 * emergency shutdown?
> -		 */
> -		emergency_sync();
> -		kernel_power_off();
> -	}
> -
> -	return ret;
> +	if (force) /* do not override the pending "true" */
> +		poweroff_force = true;
> +	schedule_work(&poweroff_work);
> +	return 0;
>  }

afaict the current version of orderly_poweroff() will never return -
either __orderly_poweroff() will block until the machine shuts down or
kernel_power_off() will do so.

However with this patch there is a path via which orderly_poweroff()
can return to its caller, I think?  If so, the caller might be rather
surprised and we're exercising never-before-used code paths.  In fact
if the surprised caller goes oops, the poweroff might not occur at all.