Re: [PATCH 2/2] nvme: keyring: fix conditional compilation

[Hannes Reinecke <hare@suse.de>]

On 10/27/23 10:30, Christoph Hellwig wrote:
> On Fri, Oct 27, 2023 at 10:12:11AM +0200, Arnd Bergmann wrote:
>>> In principle. Unfortunately I have to initialize the keyring, and that
>>> can be done only once.
>>> I see if I can come up with a different solution.
>>
>> If the keyring has to be initialized first, I think the safest
>> way is to move it to a different initcall level, to avoid
>> relying on link order for the everything-built-in case.
> 
> I don't really mind the link order, for these cases I usually add a
> comment to the Makefile so that people don't accidentally change it.

Point is not that it has to be initialized first, point is it has to be 
initialized only _once_.
So when moving it into a separate module we cannot use the Makefile 
trick from Christoph, and when initialized it in nvme-core we have a
dependency on nvme-core from nvmet.
Which would be the easiest way (ie have nvmet dependent on nvme-core),
but no sure if that's the way we want to go.

Cheers,

Hannes