On Wed, 2025-09-03 at 19:59 +0800, Li Lingfeng wrote:
> When file access conflicts occur between clients, the server recalls
> delegations. If the client holding delegation fails to return it after
> a recall, nfs4_laundromat adds the delegation to cl_revoked list.
> This causes subsequent SEQUENCE operations to set the
> SEQ4_STATUS_RECALLABLE_STATE_REVOKED flag, forcing the client to
> validate all delegations and return the revoked one.
>
> However, if the client fails to return the delegation like this:
> nfs4_laundromat nfsd4_delegreturn
> unhash_delegation_locked
> list_add // add dp to reaplist
> // by dl_recall_lru
> list_del_init // delete dp from
> // reaplist
> destroy_delegation
> unhash_delegation_locked
> // do nothing but return false
> revoke_delegation
> list_add // add dp to cl_revoked
> // by dl_recall_lru
>
> The delegation will remain in the server's cl_revoked list while the
> client marks it revoked and won't find it upon detecting
> SEQ4_STATUS_RECALLABLE_STATE_REVOKED.
> This leads to a loop:
> the server persistently sets SEQ4_STATUS_RECALLABLE_STATE_REVOKED, and the
> client repeatedly tests all delegations, severely impacting performance
> when numerous delegations exist.
>
> Since abnormal delegations are removed from flc_lease via nfs4_laundromat
> --> revoke_delegation --> destroy_unhashed_deleg -->
> nfs4_unlock_deleg_lease --> kernel_setlease, and do not block new open
> requests indefinitely, retaining such a delegation on the server is
> unnecessary.
>
> Reported-by: Zhang Jian <zhangjian496@xxxxxxxxxx>
> Fixes: 3bd64a5ba171 ("nfsd4: implement SEQ4_STATUS_RECALLABLE_STATE_REVOKED")
> Closes: https://lore.kernel.org/all/ff8debe9-6877-4cf7-ba29-fc98eae0ffa0@xxxxxxxxxx/
> Signed-off-by: Li Lingfeng <lilingfeng3@xxxxxxxxxx>
> ---
> Changes in v2:
> 1) Set SC_STATUS_CLOSED unconditionally in destroy_delegation();
> 2) Determine whether to remove the delegation based on SC_STATUS_CLOSED,
> rather than by timeout;
> 3) Modify the commit message.
> fs/nfsd/nfs4state.c | 20 ++++++++++++++++++++
> 1 file changed, 20 insertions(+)
>
> diff --git a/fs/nfsd/nfs4state.c b/fs/nfsd/nfs4state.c
> index 88c347957da5..bb9e1df4e41f 100644
> --- a/fs/nfsd/nfs4state.c
> +++ b/fs/nfsd/nfs4state.c
> @@ -1336,6 +1336,11 @@ static void destroy_delegation(struct nfs4_delegation *dp)
>
> spin_lock(&state_lock);
> unhashed = unhash_delegation_locked(dp, SC_STATUS_CLOSED);
> + /*
> + * Unconditionally set SC_STATUS_CLOSED, regardless of whether the
> + * delegation is hashed, to mark the current delegation as invalid.
> + */
> + dp->dl_stid.sc_status |= SC_STATUS_CLOSED;
> spin_unlock(&state_lock);
> if (unhashed)
> destroy_unhashed_deleg(dp);
> @@ -4326,6 +4331,8 @@ nfsd4_sequence(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
> int buflen;
> struct net *net = SVC_NET(rqstp);
> struct nfsd_net *nn = net_generic(net, nfsd_net_id);
> + struct list_head *pos, *next;
> + struct nfs4_delegation *dp;
>
nit: These could be moved inside the if statement below.
> if (resp->opcnt != 1)
> return nfserr_sequence_pos;
> @@ -4470,6 +4477,19 @@ nfsd4_sequence(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
> default:
> seq->status_flags = 0;
> }
I wouldn't mind a comment here that explains why we have to do this.
This is the sort of thing that will have us all scratching our heads in
a few years.
> + if (!list_empty(&clp->cl_revoked)) {
> + spin_lock(&clp->cl_lock);
> + list_for_each_safe(pos, next, &clp->cl_revoked) {
> + dp = list_entry(pos, struct nfs4_delegation, dl_recall_lru);
> + if (dp->dl_stid.sc_status & SC_STATUS_CLOSED) {
> + list_del_init(&dp->dl_recall_lru);
> + spin_unlock(&clp->cl_lock);
> + nfs4_put_stid(&dp->dl_stid);
> + spin_lock(&clp->cl_lock);
> + }
> + }
> + spin_unlock(&clp->cl_lock);
> + }
nit: I'd move the if statement below inside the above if statement. No
need to check list_empty() twice if it was empty the first time. Maybe
the compiler papers over this and only does it once?
> if (!list_empty(&clp->cl_revoked))
> seq->status_flags |= SEQ4_STATUS_RECALLABLE_STATE_REVOKED;
> if (atomic_read(&clp->cl_admin_revoked))
Otherwise, this looks great. Thanks for the patch!
Reviewed-by: Jeff Layton <jlayton@xxxxxxxxxx>
