Re: [RFC net-next 05/10] devlink: remove the registration guarantee of references

[Jakub Kicinski <kuba@kernel.org>]

On Mon, 19 Dec 2022 09:56:26 -0800 Jacob Keller wrote:
> > -void devlink_register(struct devlink *devlink)
> > +int devl_register(struct devlink *devlink)
> >  {
> >  	ASSERT_DEVLINK_NOT_REGISTERED(devlink);
> > -	/* Make sure that we are in .probe() routine */
> > +	devl_assert_locked(devlink);
> >  
> >  	xa_set_mark(&devlinks, devlink->index, DEVLINK_REGISTERED);
> >  	devlink_notify_register(devlink);
> > +
> > +	return 0;  
> 
> Any particular reason to change this to int when it doesn't have a
> failure case yet? Future patches I assume? You don't check the
> devl_register return value.

I was wondering if anyone would notice :)

Returning errors from the registration helper seems natural,
and if we don't have this ability it may impact our ability
to extend the core in the long run.
I was against making core functions void in the first place.
It's a good opportunity to change back.