- landed b_hd_cray_merge3

[fs/lustre-release.git] / lustre / tests / conf-sanity.sh

diff --git a/lustre/tests/conf-sanity.sh b/lustre/tests/conf-sanity.sh
index acab312..c6eff64 100644 (file)
--- a/lustre/tests/conf-sanity.sh
+++ b/lustre/tests/conf-sanity.sh
@@ -47,9 +47,10 @@ start_mds() {
        start mds1 --reformat $MDSLCONFARGS  || return 94
        start_lsvcgssd || return 501
 }
+
 stop_mds() {
        echo "stop mds1 service on `facet_active_host mds1`"
-       stop mds1 $@  || return 97
+       stop mds1 $@ || return 97
        stop_lsvcgssd
 }
 
@@ -60,7 +61,7 @@ start_ost() {
 
 stop_ost() {
        echo "stop ost service on `facet_active_host ost`"
-       stop ost $@  || return 98
+       stop ost $@ || return 98
 }
 
 mount_client() {
@@ -188,7 +189,17 @@ test_5() {
        # cleanup may return an error from the failed
        # disconnects; for now I'll consider this successful
        # if all the modules have unloaded.
-       umount $MOUNT &
+
+       # as MDS is down, umount without -f may cause blocking
+       # and this test will never finish. Blocking is possible
+       # as umount may want to cancel locks with RPC's and these
+       # RPC's will wait forever, as pinger thread will try to
+       # recover failed import endlessly.
+       #
+       # Thus, main point is: nobody should expect umount finish
+       # quickly and cleanly without -f flag when MDS or OST is 
+       # down for sure. --umka
+       umount -f $MOUNT &
        UMOUNT_PID=$!
        sleep 2
        echo "killing umount"
@@ -198,16 +209,21 @@ test_5() {
        stop_lgssd
 
        # cleanup client modules
-       $LCONF --cleanup --nosetup --node client_facet $XMLCONFIG > /dev/null
+       $LCONF --force --cleanup --nosetup --node client_facet $XMLCONFIG > /dev/null
        
        # stop_mds is a no-op here, and should not fail
-       stop_mds  || return 4
-       stop_ost || return 5
+       stop_mds || return 4
+       
+       # this should have --force flag specified, as umount -f
+       # will skip disconnect phase and thus OST will have one
+       # extra refcount what will cause class_cleanup() failure
+       # if --force is not specified. --umka
+       stop_ost --force || return 5
 
        lsmod | grep -q portals && return 6
        return 0
 }
-run_test 5 "force cleanup mds, then cleanup"
+run_test 5 "force cleanup mds, then cleanup --force"
 
 test_5b() {
        start_ost
@@ -217,17 +233,17 @@ test_5b() {
        [ -d $MOUNT ] || mkdir -p $MOUNT
        $LCONF --nosetup --node client_facet $XMLCONFIG > /dev/null
        start_lgssd || return 1
-       llmount $mds_HOST://mds1_svc/client_facet $MOUNT  && exit 1
+       llmount -o nettype=$NETTYPE $mds_HOST://mds_svc/client_facet $MOUNT  && exit 2
 
        # cleanup client modules
        $LCONF --cleanup --nosetup --node client_facet $XMLCONFIG > /dev/null
        stop_lgssd
        
        # stop_mds is a no-op here, and should not fail
-       stop_mds || return 2
-       stop_ost || return 3
+       stop_mds || return 3
+       stop_ost || return 4
 
-       lsmod | grep -q portals && return 4 
+       lsmod | grep -q portals && return 5
        return 0
 
 }
@@ -240,7 +256,7 @@ test_5c() {
        [ -d $MOUNT ] || mkdir -p $MOUNT
        $LCONF --nosetup --node client_facet $XMLCONFIG > /dev/null
        start_lgssd || return 1
-        llmount $mds_HOST://wrong_mds1_svc/client_facet $MOUNT  && return 2
+       llmount -o nettype=$NETTYPE $mds_HOST://wrong_mds_svc/client_facet $MOUNT  && return 2
 
        # cleanup client modules
        $LCONF --cleanup --nosetup --node client_facet $XMLCONFIG > /dev/null
@@ -642,7 +658,7 @@ test_16() {
             mount_client $MOUNT
             check_mount || return 41
             cleanup || return $?
-         fi
+        fi
                                                                                                                              
         echo "change the mode of $MDSDEV/OBJECTS,LOGS,PENDING to 555"
         [ -d $TMPMTPT ] || mkdir -p $TMPMTPT