nfsd: Fix race to FREE_STATEID and cl_revoked

We have some reports of linux NFS clients that cannot satisfy a linux knfsd
server that always sets SEQ4_STATUS_RECALLABLE_STATE_REVOKED even though
those clients repeatedly walk all their known state using TEST_STATEID and
receive NFS4_OK for all.

Its possible for revoke_delegation() to set NFS4_REVOKED_DELEG_STID, then
nfsd4_free_stateid() finds the delegation and returns NFS4_OK to
FREE_STATEID.  Afterward, revoke_delegation() moves the same delegation to
cl_revoked.  This would produce the observed client/server effect.

Fix this by ensuring that the setting of sc_type to NFS4_REVOKED_DELEG_STID
and move to cl_revoked happens within the same cl_lock.  This will allow
nfsd4_free_stateid() to properly remove the delegation from cl_revoked.

Link: https://bugzilla.redhat.com/show_bug.cgi?id=2217103
Link: https://bugzilla.redhat.com/show_bug.cgi?id=2176575
Signed-off-by: Benjamin Coddington <[email protected]>
Cc: [email protected] # v4.17+
Reviewed-by: Jeff Layton <[email protected]>
Signed-off-by: Chuck Lever <[email protected]>

Author: bcodding-rh

fs/nfsd/nfs4state.c

@@ -1354,9 +1354,9 @@ static void revoke_delegation(struct nfs4_delegation *dp)
 	trace_nfsd_stid_revoke(&dp->dl_stid);
 
 	if (clp->cl_minorversion) {
+		spin_lock(&clp->cl_lock);
 		dp->dl_stid.sc_type = NFS4_REVOKED_DELEG_STID;
 		refcount_inc(&dp->dl_stid.sc_count);
-		spin_lock(&clp->cl_lock);
 		list_add(&dp->dl_recall_lru, &clp->cl_revoked);
 		spin_unlock(&clp->cl_lock);
 	}