NFSD: Add nfsd4_copy time-to-live

Keep async copy state alive for a few lease cycles after the copy
completes so that OFFLOAD_STATUS returns something meaningful.

This means that NFSD's client shutdown processing needs to purge
any of this state that happens to be waiting to die.

Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>

3 files changed, 21 insertions, 2 deletions

diff --git a/fs/nfsd/nfs4proc.c b/fs/nfsd/nfs4proc.c
index dfc118bf9aa5..f8a10f90bc7a 100644
--- a/fs/nfsd/nfs4proc.c
+++ b/fs/nfsd/nfs4proc.c
@@ -1326,8 +1326,10 @@ void nfsd4_async_copy_reaper(struct nfsd_net *nn)
 		list_for_each_safe(pos, next, &clp->async_copies) {
 			copy = list_entry(pos, struct nfsd4_copy, copies);
 			if (test_bit(NFSD4_COPY_F_OFFLOAD_DONE, &copy->cp_flags)) {
-				list_del_init(&copy->copies);
-				list_add(&copy->copies, &reaplist);
+				if (--copy->cp_ttl) {
+					list_del_init(&copy->copies);
+					list_add(&copy->copies, &reaplist);
+				}
 			}
 		}
 		spin_unlock(&clp->async_lock);
@@ -1921,6 +1923,7 @@ nfsd4_copy(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
 		async_copy->cp_nn = nn;
 		INIT_LIST_HEAD(&async_copy->copies);
 		refcount_set(&async_copy->refcount, 1);
+		async_copy->cp_ttl = NFSD_COPY_INITIAL_TTL;
 		/* Arbitrary cap on number of pending async copy operations */
 		if (atomic_inc_return(&nn->pending_async_copies) >
 				(int)rqstp->rq_pool->sp_nrthreads)
diff --git a/fs/nfsd/state.h b/fs/nfsd/state.h
index dcbebd53e5f4..8092894e8f3c 100644
--- a/fs/nfsd/state.h
+++ b/fs/nfsd/state.h
@@ -137,6 +137,21 @@ struct nfs4_cpntf_state {
 	time64_t		cpntf_time;	/* last time stateid used */
 };
 
+/*
+ * RFC 7862 Section 4.8 states:
+ *
+ * | A copy offload stateid will be valid until either (A) the client
+ * | or server restarts or (B) the client returns the resource by
+ * | issuing an OFFLOAD_CANCEL operation or the client replies to a
+ * | CB_OFFLOAD operation.
+ *
+ * Because a client might not reply to a CB_OFFLOAD, or a reply
+ * might get lost due to connection loss, NFSD purges async copy
+ * state after a short period to prevent it from accumulating
+ * over time.
+ */
+#define NFSD_COPY_INITIAL_TTL 10
+
 struct nfs4_cb_fattr {
 	struct nfsd4_callback ncf_getattr;
 	u32 ncf_cb_status;
diff --git a/fs/nfsd/xdr4.h b/fs/nfsd/xdr4.h
index 5951360f6d03..382cc1389396 100644
--- a/fs/nfsd/xdr4.h
+++ b/fs/nfsd/xdr4.h
@@ -715,6 +715,7 @@ struct nfsd4_copy {
 	struct list_head	copies;
 	struct task_struct	*copy_task;
 	refcount_t		refcount;
+	unsigned int		cp_ttl;
 
 	struct nfsd4_ssc_umount_item *ss_nsui;
 	struct nfs_fh		c_fh;