+test_6b() {
+ mcreate $MOUNT1/a
+ multiop $MOUNT2/a o_c &
+ pid1=$!
+ multiop $MOUNT1/a o_c &
+ pid2=$!
+ # give multiop a chance to open
+ sleep 1
+ rm -f $MOUNT1/a
+ replay_barrier mds1
+ kill -USR1 $pid2
+ wait $pid2 || return 1
+
+ fail mds1
+ kill -USR1 $pid1
+ wait $pid1 || return 1
+ [ -e $MOUNT2/a ] && return 2
+ return 0
+}
+run_test 6b "open1, open2, unlink |X| close2 [fail mds] close1"
+
+test_8() {
+ replay_barrier mds1
+ drop_reint_reply "mcreate $MOUNT1/$tfile" || return 1
+ fail mds1
+ checkstat $MOUNT2/$tfile || return 2
+ rm $MOUNT1/$tfile || return 3
+
+ return 0
+}
+run_test 8 "replay of resent request"
+
+test_9() {
+ replay_barrier mds1
+ mcreate $MOUNT1/$tfile-1
+ mcreate $MOUNT2/$tfile-2
+ # drop first reint reply
+ sysctl -w lustre.fail_loc=0x80000119
+ fail mds1
+ sysctl -w lustre.fail_loc=0
+
+ rm $MOUNT1/$tfile-[1,2] || return 1
+
+ return 0
+}
+run_test 9 "resending a replayed create"
+
+test_10() {
+ mcreate $MOUNT1/$tfile-1
+ replay_barrier mds1
+ munlink $MOUNT1/$tfile-1
+ mcreate $MOUNT2/$tfile-2
+ # drop first reint reply
+ sysctl -w lustre.fail_loc=0x80000119
+ fail mds1
+ sysctl -w lustre.fail_loc=0
+
+ checkstat $MOUNT1/$tfile-1 && return 1
+ checkstat $MOUNT1/$tfile-2 || return 2
+ rm $MOUNT1/$tfile-2
+
+ return 0
+}
+run_test 10 "resending a replayed unlink"
+
+test_11() {
+ replay_barrier mds1
+ mcreate $MOUNT1/$tfile-1
+ mcreate $MOUNT2/$tfile-2
+ mcreate $MOUNT1/$tfile-3
+ mcreate $MOUNT2/$tfile-4
+ mcreate $MOUNT1/$tfile-5
+ # drop all reint replies for a while
+ sysctl -w lustre.fail_loc=0x0119
+ facet_failover mds1
+ #sleep for while, let both clients reconnect and timeout
+ sleep $((TIMEOUT * 2))
+ sysctl -w lustre.fail_loc=0
+
+ rm $MOUNT1/$tfile-[1-5] || return 1
+
+ return 0
+}
+run_test 11 "both clients timeout during replay"
+
+test_12() {
+ replay_barrier mds1
+
+ multiop $DIR/$tfile mo_c &
+ MULTIPID=$!
+ sleep 5
+
+ # drop first enqueue
+ sysctl -w lustre.fail_loc=0x80000302
+ facet_failover mds1
+ df $MOUNT || return 1
+ sysctl -w lustre.fail_loc=0
+
+ ls $DIR/$tfile
+ $CHECKSTAT -t file $DIR/$tfile || return 2
+ kill -USR1 $MULTIPID || return 3
+ wait $MULTIPID || return 4
+ rm $DIR/$tfile
+
+ return 0
+}
+run_test 12 "open resend timeout"
+
+test_13() {
+ multiop $DIR/$tfile mo_c &
+ MULTIPID=$!
+ sleep 5
+
+ replay_barrier mds1
+
+ kill -USR1 $MULTIPID || return 3
+ wait $MULTIPID || return 4
+
+ # drop close
+ sysctl -w lustre.fail_loc=0x80000115
+ facet_failover mds1
+ df $MOUNT || return 1
+ sysctl -w lustre.fail_loc=0
+
+ ls $DIR/$tfile
+ $CHECKSTAT -t file $DIR/$tfile || return 2
+ rm $DIR/$tfile
+
+ return 0
+}
+run_test 13 "close resend timeout"
+
+
+test_14() {
+ replay_barrier mds1
+ createmany -o $MOUNT1/$tfile- 25
+ createmany -o $MOUNT2/$tfile-2- 1
+ createmany -o $MOUNT1/$tfile-3- 25
+ umount $MOUNT2
+
+ facet_failover mds1
+ # expect failover to fail
+ df $MOUNT && return 1
+ sleep 1
+
+ # first 25 files shouuld have been
+ # replayed
+ sleep 2
+ unlinkmany $MOUNT1/$tfile- 25 || return 2
+
+ zconf_mount `hostname` $MOUNT2
+ return 0
+}
+run_test 14 "timeouts waiting for lost client during replay"
+
+test_15() {
+ replay_barrier mds1
+ createmany -o $MOUNT1/$tfile- 25
+ createmany -o $MOUNT2/$tfile-2- 1
+ umount $MOUNT2
+
+ facet_failover mds1
+ df $MOUNT || return 1
+ sleep 1
+
+ unlinkmany $MOUNT1/$tfile- 25 || return 2
+
+ zconf_mount `hostname` $MOUNT2
+ return 0
+}
+run_test 15 "timeout waiting for lost client during replay, 1 client completes"
+test_16() {
+ replay_barrier mds1
+ createmany -o $MOUNT1/$tfile- 25
+ createmany -o $MOUNT2/$tfile-2- 1
+ umount $MOUNT2
+
+ facet_failover mds1
+ sleep $TIMEOUT
+ facet_failover mds1
+ df $MOUNT || return 1
+ sleep 1
+
+ unlinkmany $MOUNT1/$tfile- 25 || return 2
+
+ zconf_mount `hostname` $MOUNT2
+ return 0
+
+}
+#run_test 16 "fail MDS during recovery (3571)"
+
+test_17() {
+ createmany -o $MOUNT1/$tfile- 25
+ createmany -o $MOUNT2/$tfile-2- 1
+
+ # Make sure the disconnect is lost
+ replay_barrier ost
+ umount $MOUNT2
+
+ echo -1 > /proc/sys/portals/debug
+ facet_failover ost
+ sleep $TIMEOUT
+ facet_failover ost
+ df $MOUNT || return 1
+ sleep 1
+
+ unlinkmany $MOUNT1/$tfile- 25 || return 2
+
+ zconf_mount `hostname` $MOUNT2
+ return 0
+
+}
+#Still not support ost fail over
+#run_test 17 "fail OST during recovery (3571)"
+
+test_18 () {
+ replay_barrier mds1
+ multiop $MOUNT2/$tfile O_c &
+ pid2=$!
+ multiop $MOUNT1/$tfile O_c &
+ pid1=$!
+ # give multiop a chance to open
+ sleep 1
+ kill -USR1 $pid2
+ kill -USR1 $pid1
+ sleep 1
+ umount $MOUNT2
+ facet_failover mds1
+ df || df || return 1
+ zconf_mount `hostname` $MOUNT2
+}
+run_test 18 "replay open, Abort recovery, don't assert (3892)"
+
+# cleanup with blocked enqueue fails until timer elapses (MDS busy), wait for
+# itexport NOW=0
+
+test_20() { # bug 3822 - evicting client with enqueued lock
+ mkdir -p $MOUNT1/$tdir
+ touch $MOUNT1/$tdir/f0
+#define OBD_FAIL_LDLM_ENQUEUE_BLOCKED 0x30b
+ statmany -s $MOUNT1/$tdir/f 500 &
+ OPENPID=$!
+ NOW=`date +%s`
+ do_facet mds1 sysctl -w lustre.fail_loc=0x8000030b # hold enqueue
+ sleep 1
+#define OBD_FAIL_LDLM_BL_CALLBACK 0x305
+ do_facet client sysctl -w lustre.fail_loc=0x80000305 # drop cb, evict
+ cancel_lru_locks MDC
+ usleep 500 # wait to ensure first client is one that will be evicted
+ openfile -f O_RDONLY $MOUNT2/$tdir/f0
+ wait $OPENPID
+ dmesg | grep "entering recovery in server" && \
+ error "client not evicted" || true
+}
+run_test 20 "ldlm_handle_enqueue succeeds on evicted export (3822)"
+
+# $1 - number of mountpoint
+# $2 - mds
+function find_dev_for_fs_and_mds()
+{
+ local fsuuid=`cat /proc/fs/lustre/llite/fs$1/uuid`
+ $LCTL device_list | awk "/mdc.*$2.*$fsuuid/ {print \$4}"
+}
+
+test_21() {
+ mdc1dev=`find_dev_for_fs_and_mds 0 mds1`
+ mdc2dev=`find_dev_for_fs_and_mds 1 mds1`
+ multiop $MOUNT1/f21 O
+ cancel_lru_locks MDC
+ # generate IT_OPEN to be replayed against existing file
+ multiop $MOUNT1/f21 o_Sc &
+ pid=$!
+
+ # IT_OPEN will be committed by the failover time
+ replay_barrier mds1
+
+ # generate MDS_REINT_UNLINK to be replayed
+ rm -f $MOUNT2/f21 || return 1
+
+ # disable recovery on the both clients
+ $LCTL --device %$mdc1dev disable_recovery
+ $LCTL --device %$mdc2dev disable_recovery
+ facet_failover mds1
+
+ # let unlink to be replayed first
+ $LCTL --device %$mdc2dev enable_recovery
+ sleep $((TIMEOUT/2))
+
+ # now let open to be replaye
+ $LCTL --device %$mdc1dev enable_recovery
+ kill -USR1 $pid
+ wait $pid || return 2
+}
+run_test 21 "open vs. unlink out of order replay"
+
+test_22() { # bug 6063 - AST during recovery
+ mdc1dev=`find_dev_for_fs_and_mds 0 mds1`
+ mdc2dev=`find_dev_for_fs_and_mds 1 mds1`
+ $LCTL --device %$mdc1dev disable_recovery
+ $LCTL --device %$mdc2dev disable_recovery
+
+ replay_barrier mds1
+ mkdir $MOUNT1/${tdir}-1 # client1: request to be replayed
+ ls $MOUNT2 # client2: take lock needed for
+ facet_failover mds1
+
+ # let's recover 2nd connection with granted UPDATE lock
+ $LCTL --device %$mdc2dev enable_recovery
+ sleep $((TIMEOUT / 2))
+
+ $LCTL mark "first recovered?"
+ LOCKS=`grep -v '^0$' /proc/fs/lustre/ldlm/namespaces/mds-*/lock_count`
+ if [ "$LOCKS" != "" ]; then
+ echo "The lock got replayed before mkdir is replayed: $LOCKS"
+ echo 0 >${IMP1}
+ return 1
+ fi
+
+ # let's recover 1st connection with mkdir replay that needs the lock
+ $LCTL --device %$mdc1dev enable_recovery
+ sleep $TIMEOUT
+ $LCTL mark "second recovered?"
+
+ LOCKS=`grep -v '^0$' /proc/fs/lustre/ldlm/namespaces/mds-*/lock_count`
+ if [ "$LOCKS" != "1" ]; then
+ echo "The lock hasn't replayed: $LOCKS"
+ return 2
+ fi
+
+ return 0
+}
+run_test 22 "AST during recovery"
+