Re: [PATCH net-next 03/10] devlink: Serialize access to rate domains

Wed, Feb 19, 2025 at 03:21:30AM +0100, kuba@kernel.org wrote:

On Fri, 14 Feb 2025 13:54:43 +0100 Jiri Pirko wrote:

quoted

For the record, I'm still not convinced that introducing this kind of
shared inter-devlink lock is good idea. We spent quite a bit of painful
times getting rid of global devlink_mutex and making devlink locking
scheme nice and simple as it currently is.

But at the same time I admit I can't think of any other nicer solution
to the problem this patchset is trying to solve.

Jakub, any thoughts?

The problem comes from having a devlink instance per function /
port rather than for the ASIC. Spawn a single instance and the
problem will go away 🤷️

Yeah, we currently have VF devlink ports created under PF devlink instance.
That is aligned with PCI geometry. If we have a single per-ASIC parent
devlink, this does not change and we still need to configure cross
PF devlink instances.

The only benefit I see is that we don't need rate domain, but
we can use parent devlink instance lock instead. The locking ordering
might be a bit tricky to fix though.

I think we talked about this multiple times, I think at least
once with Jake, too. Not that I remember all the details now..
-- 
pw-bot: cr