Re: [PATCH v2 4/8] SUNRPC: undo partial rpcbind registrations when svc_register() fails

On Fri, Aug 14, 2026, at 7:01 AM, Jeff Layton wrote:
> I still think the best solution of all would be to just say
> "henceforth, rpcbind registration is non-fatal".

Fair enough, let's go with that.


> Right now, I'm really unclear on what sort of changes you want to see
> here as a final patchset. What would make this mergeable for you?

Sorry for the ambiguity. For the re-roll:

- 1, 2, 3 and 5 don't need change.

- Let's drop patch 4/8.

- Make registration failure non-fatal for NFSv2 and NFSv3, as
  proposed.
  
- Extend the extack in 6/8 to report a listener that came up without
  registering. dmesg is the only channel write_ports() has. Perhaps
  the netlink path can do better?

- Fix the netns leak in 8/8.

- Ensure that any test that asserts the unwind or the abort is moved
  over to the new failure policy.

HTH.

-- 
Chuck Lever