nfs-utils: nfsserver restart should kill and recreate nfsd kernel threads

From: Qiang Chen <[email protected]>

nfsserver restart without killing kernel threads worked when portmap was the rpc publishing process and portmap was restarted. When rpcbind replaces portmap, nfsserver restart in this way does not work after an rpcbind restart.

Steps to reproduce:
1). Make ext3 filesystem image on local host.
cd /root
dd if=/dev/zero of=test bs=1024K count=50
mkfs.ext3 -F test

2). runqemu qemux86-64
mkdir /mnt/wrtest
mount -t ext3 -o loop test /mnt/wrtest
echo "/mnt/wrtest *(sync,rw,no_subtree_check,no_root_squash)" > /etc/exports /etc/init.d/rpcbind restart
/etc/init.d/nfsserver restart
showmount -e localhost
mkdir wrtest
mount -t nfs localhost:/mnt/wrtest wrtest

mount: mounting localhost:/mnt/wrtest on wrtest failed: Connection refused

Modifying the nfsserver script to kill and restart kernel threads on restart makes the problem go away and is consistent with current RHEL/SUSE and Ubuntu/Debian mechanisms of handling the nfs server.

---
 .../nfs-utils/nfs-utils/nfsserver                  |   11 +++--------
 1 file changed, 3 insertions(+), 8 deletions(-)

index 8ee8d0b..d7cf6e0 100644
@@ -161,14 +161,9 @@ status)
 	[ $RETVAL -eq 0 ] && exit $rval
 	exit $RETVAL;;
 reload)	test -r /etc/exports && exportfs -r;;
-restart)exportfs -ua
-	stop_mountd
-	stop_statd
-	# restart does not restart the kernel threads,
-	# only the user mode processes
-	start_mountd
-	start_statd
-	test -r /etc/exports && exportfs -a;;
+restart)
+	$0 stop
+	$0 start;;
 *)	echo "Usage: $0 {start|stop|status|reload|restart}"
 	exit 1;;
 esac

1.7.9.5