Re: [PATCH v2 0/4] NFSD: CB_RECALL_ANY fixes and a meaningful keep count

On Wed, 2026-08-12 at 10:59 -0400, Chuck Lever wrote:
> NFSD sends every CB_RECALL_ANY with craa_objects_to_keep set to
> zero, as RFC 8881 Section 20.6.3 does not mandate any particular
> way for a client to choose which delegations to choose, if any.
> 
> However, a zero value can result in some clients giving back more
> delegations than is necessary to relieve temporary memory pressure
> on the server, which needlessly impacts performance.
> 
> NFSD intends the callback as a signal to return unused delegations.
> Only the Linux client has been tested against it, and that client
> ignores craa_objects_to_keep and returns only one unused delegation
> of the named types, so the fixed zero never produced visible
> misbehavior during our testing.
> 
> So, change NFSD so that each CB_RECALL_ANY asks a client to give up
> one delegation. Both reaper callers (the shrinker to relieve memory
> pressure, and the laundromat to cap the total number of delegations
> the server tracks) re-arm while their condition lasts.
> 
> NFSD sets no recall target and remembers nothing across CB_RECALL_ANY
> callbacks. RFC 8881 Section 20.6.4 prescribes the use of CB_RECALL
> to target specific delegations if a client fails to return any. NFSD
> does not take that step yet. CB_RECALL_ANY is asynchronous and
> reports no completion, so NFSD treats it as advisory.
> 
> ---
> Changes in v2:
> - Drop the gate that skipped clients holding a single delegation.
> - Cover letter rewritten to give performance rationale.
> - Link to v1: https://patch.msgid.link/[email protected]
> 
> ---
> Chuck Lever (4):
>       NFSD: Do not send CB_RECALL_ANY to NFSv4.0 clients
>       NFSD: Count the delegations held by each client
>       NFSD: Name directory delegations in the CB_RECALL_ANY type mask
>       NFSD: Send a meaningful CB_RECALL_ANY keep count
> 
>  fs/nfsd/nfs4state.c | 27 +++++++++++++++++++++++----
>  fs/nfsd/state.h     |  2 ++
>  2 files changed, 25 insertions(+), 4 deletions(-)
> ---
> base-commit: 1d479c6b53f684b27da84ec352b7efb97f7f115f
> change-id: 20260810-recall-any-keep-count-f50c2ae1b792
> 
> Best regards,
> --  
> Chuck Lever <[email protected]>

Looks reasonable. Given that the Linux client ignores
craa_objects_to_keep and just sends back a single delegation, this
should keep things working the same even when it's brought into
compliance.

Reviewed-by: Jeff Layton <[email protected]>