From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1422931AbWJPXai (ORCPT ); Mon, 16 Oct 2006 19:30:38 -0400 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S1750814AbWJPXai (ORCPT ); Mon, 16 Oct 2006 19:30:38 -0400 Received: from ns.suse.de ([195.135.220.2]:45969 "EHLO mx1.suse.de") by vger.kernel.org with ESMTP id S1422931AbWJPXaa (ORCPT ); Mon, 16 Oct 2006 19:30:30 -0400 From: NeilBrown To: Andrew Morton Date: Tue, 17 Oct 2006 09:30:25 +1000 Message-Id: <1061016233025.11354@suse.de> X-face: [Gw_3E*Gng}4rRrKRYotwlE?.2|**#s9D Sender: linux-kernel-owner@vger.kernel.org X-Mailing-List: linux-kernel@vger.kernel.org When an nfs server shuts down, lockd needs to release all the locks even though the client still holds them. It should therefore not 'unmonitor' the clients, so that the files in nfs/sm will still be there when the nfs server restarts, so that those clients will be told to reclaim their locks. However the hosts are fully unmonitored, so statd may well remove the files. lockd has a test for 'sm_sticky' and avoid the unmonitor call if it is set, but it is currently not set. So set it when tearing down lockd. Signed-off-by: Neil Brown ### Diffstat output ./fs/lockd/svcsubs.c | 12 +++++++++++- 1 file changed, 11 insertions(+), 1 deletion(-) diff .prev/fs/lockd/svcsubs.c ./fs/lockd/svcsubs.c --- .prev/fs/lockd/svcsubs.c 2006-10-17 09:10:39.000000000 +1000 +++ ./fs/lockd/svcsubs.c 2006-10-17 09:10:40.000000000 +1000 @@ -324,7 +324,17 @@ nlmsvc_same_host(struct nlm_host *host, static int nlmsvc_is_client(struct nlm_host *host, struct nlm_host *dummy) { - return host->h_server; + if (host->h_server) + { + /* we are destroying locks even though the client + * hasn't asked us too, so don't unmonitor the + * client + */ + if (host->h_nsmhandle) + host->h_nsmhandle->sm_sticky = 1; + return 1; + } else + return 0; } /*