在 2024/10/21 22:25, Chuck Lever III 写道:

quoted

On Oct 21, 2024, at 10:18 AM, yangerkun [off-list ref] wrote:



在 2024/10/21 22:00, Chuck Lever 写道:

quoted

On Mon, Oct 21, 2024 at 04:25:40PM +0800, Yang Erkun wrote:

quoted

Fixes: 7746b32f467b ("NFSD: add shrinker to reap courtesy clients on low memory condition")

I think like:
Fixes: 7c24fa225081 ("NFSD: replace delayed_work with work_struct for nfsd_client_shrinker")

Hi,

Thanks a lot for your review! Before this, will this problem exist too
since the concurrently run between upper two threads can happen too?

Yes, the race can happen before 7c24fa, but your patch
won't apply to kernels that don't have 7c24fa applied.

The automation should pull both of these into LTS correctly.
If it doesn't happen as we expect, we can fix it up by hand.

Thanks for explaining. Agree with you.

I plan to apply this to nfsd-fixes (for v6.12-rc).

Thanks!

quoted

quoted

a little better.
I'm very curious how you stumbled across this one!

Our excellent test team has recently performed a lot of fault injection
tests on nfs/nfsd, which helps us find many nfs/nfsd problems. This
problem is only one of them. There will be some bugfixes for other
problems later.

Excellent, looking forward to seeing those.


--
Chuck Lever