Once the ENOLCK is detected we expect to retry the acquisition.

Anyway, in the edge case the flushing happens and the while is no more
executed, nfs_flush() (and nfs4_flush()) can return with a wrong
err value of ENOLCK.
Bring it back to 0, as we expect to have for that case.

Reported by:	kris
Reviewed by:	kib

1 files changed, 3 insertions, 1 deletions

diff --git a/sys/nfs4client/nfs4_vnops.c b/sys/nfs4client/nfs4_vnops.c
index 276cb1ac3990..bd0cfb1e5148 100644
--- a/sys/nfs4client/nfs4_vnops.c
+++ b/sys/nfs4client/nfs4_vnops.c
@@ -2693,8 +2693,10 @@ loop:
 			    BO_MTX(bo), "nfsfsync", slpflag, slptimeo);
 			if (error == 0)
 				panic("nfs4_fsync: inconsistent lock");
-			if (error == ENOLCK)
+			if (error == ENOLCK) {
+				error = 0;
 				goto loop;
+			}
 			if (nfs4_sigintr(nmp, NULL, td)) {
 				error = EINTR;
 				goto done;