nfsd4: release openowners on free in >=4.1 case
authorJ. Bruce Fields <bfields@redhat.com>
Wed, 6 Jun 2012 16:53:48 +0000 (12:53 -0400)
committerJ. Bruce Fields <bfields@redhat.com>
Tue, 10 Jul 2012 20:41:34 +0000 (16:41 -0400)
We don't need to keep openowners around in the >=4.1 case, because they
aren't needed to handle CLOSE replays any more (that's a problem for
sessions).  And doing so causes unexpected failures on a subsequent
destroy_clientid to fail.

We probably also need something comparable for lock owners on last
unlock.

Signed-off-by: J. Bruce Fields <bfields@redhat.com>
fs/nfsd/nfs4state.c

index 9efa4055b5a801a5d35d5b5706d9392a9e5bb9d1..e404fca08260a111896430782f4ff5dda3fef914 100644 (file)
@@ -3763,12 +3763,19 @@ nfsd4_close(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
        nfsd4_close_open_stateid(stp);
        oo->oo_last_closed_stid = stp;
 
-       /* place unused nfs4_stateowners on so_close_lru list to be
-        * released by the laundromat service after the lease period
-        * to enable us to handle CLOSE replay
-        */
-       if (list_empty(&oo->oo_owner.so_stateids))
-               move_to_close_lru(oo);
+       if (list_empty(&oo->oo_owner.so_stateids)) {
+               if (cstate->minorversion) {
+                       release_openowner(oo);
+                       cstate->replay_owner = NULL;
+               } else {
+                       /*
+                        * In the 4.0 case we need to keep the owners around a
+                        * little while to handle CLOSE replay.
+                        */
+                       if (list_empty(&oo->oo_owner.so_stateids))
+                               move_to_close_lru(oo);
+               }
+       }
 out:
        if (!cstate->replay_owner)
                nfs4_unlock_state();