X-Git-Url: https://git.whamcloud.com/?a=blobdiff_plain;f=lustre%2Ftests%2Frecovery-small.sh;h=d1e51f8e65c9339bb15830a4294b2f641d194664;hb=cc980df563ef86847aae1e0a3f0a5b17589e6297;hp=43507fdeb99e2e9d7d12d88c85a3fd3a92ef43b6;hpb=fc2e06325690234c31ccb9a5bfc570bd17641a6f;p=fs%2Flustre-release.git

diff --git a/lustre/tests/recovery-small.sh b/lustre/tests/recovery-small.sh
index 43507fd..d1e51f8 100755
--- a/lustre/tests/recovery-small.sh
+++ b/lustre/tests/recovery-small.sh
@@ -2,9 +2,10 @@
 
 set -e
 
-#         bug  5494 7288 5493
-ALWAYS_EXCEPT="24   27   52 $RECOVERY_SMALL_EXCEPT"
+#         bug  5493  LU2034
+ALWAYS_EXCEPT="52    60      $RECOVERY_SMALL_EXCEPT"
 
+export MULTIOP=${MULTIOP:-multiop}
 PTLDEBUG=${PTLDEBUG:--1}
 LUSTRE=${LUSTRE:-`dirname $0`/..}
 . $LUSTRE/tests/test-framework.sh
@@ -12,13 +13,6 @@ init_test_env $@
 . ${CONFIG:=$LUSTRE/tests/cfg/$NAME.sh}
 init_logging
 
-if [ "$FAILURE_MODE" = "HARD" ] && mixed_ost_devs; then
-    CONFIG_EXCEPTIONS="52"
-    echo -n "Several ost services on one ost node are used with FAILURE_MODE=$FAILURE_MODE. "
-    echo "Except the tests: $CONFIG_EXCEPTIONS"
-    ALWAYS_EXCEPT="$ALWAYS_EXCEPT $CONFIG_EXCEPTIONS"
-fi
-
 require_dsh_mds || exit 0
 
 # also long tests: 19, 21a, 21e, 21f, 23, 27
@@ -35,64 +29,90 @@ CLEANUP=${CLEANUP:-""}
 check_and_setup_lustre
 
 assert_DIR
-rm -rf $DIR/[df][0-9]*
+rm -rf $DIR/d[0-9]* $DIR/f.${TESTSUITE}*
 
 test_1() {
-    drop_request "mcreate $DIR/f1"  || return 1
-    drop_reint_reply "mcreate $DIR/f2"    || return 2
-}
-run_test 1 "mcreate: drop req, drop rep"
+	local f1="$DIR/$tfile"
+	local f2="$DIR/$tfile.2"
 
-test_2() {
-    drop_request "tchmod 111 $DIR/f2"  || return 1
-    drop_reint_reply "tchmod 666 $DIR/f2"    || return 2
-}
-run_test 2 "chmod: drop req, drop rep"
+	drop_request "mcreate $f1" ||
+		error_noexit "create '$f1': drop req"
 
-test_3() {
-    drop_request "statone $DIR/f2" || return 1
-    drop_reply "statone $DIR/f2"   || return 2
-}
-run_test 3 "stat: drop req, drop rep"
+	drop_reint_reply "mcreate $f2" ||
+		error_noexit "create '$f2': drop rep"
+
+	drop_request "tchmod 111 $f2" ||
+		error_noexit "chmod '$f2': drop req"
+
+	drop_reint_reply "tchmod 666 $f2" ||
+		error_noexit "chmod '$f2': drop rep"
 
-SAMPLE_NAME=f0.recovery-small.junk
-SAMPLE_FILE=$TMP/$SAMPLE_NAME
-# make this big, else test 9 doesn't wait for bulk -- bz 5595
-dd if=/dev/urandom of=$SAMPLE_FILE bs=1M count=4
+	drop_request "statone $f2" ||
+		error_noexit "stat '$f2': drop req"
+
+	drop_reply  "statone $f2" ||
+		error_noexit "stat '$f2': drop rep"
+}
+run_test 1 "create, chmod, stat: drop req, drop rep"
 
 test_4() {
-    do_facet client "cp $SAMPLE_FILE $DIR/$SAMPLE_NAME" || return 1
-    drop_request "cat $DIR/$SAMPLE_NAME > /dev/null"   || return 2
-    drop_reply "cat $DIR/$SAMPLE_NAME > /dev/null"     || return 3
+	local t=$DIR/$tfile
+	do_facet_create_file client $t 10K ||
+		error_noexit "Create file $t"
+
+	drop_request "cat $t > /dev/null" ||
+		error_noexit "Open request for $t file"
+
+	drop_reply "cat $t > /dev/null" ||
+		error_noexit "Open replay for $t file"
 }
 run_test 4 "open: drop req, drop rep"
 
-RENAMED_AGAIN=$DIR/f0.renamed-again
-
 test_5() {
-    drop_request "mv $DIR/$SAMPLE_NAME $DIR/$tfile-renamed" || return 1
-    drop_reint_reply "mv $DIR/$tfile-renamed $RENAMED_AGAIN" || return 2
-    do_facet client "checkstat -v $RENAMED_AGAIN"  || return 3
+	local T=$DIR/$tfile
+	local R="$T-renamed"
+	local RR="$T-renamed-again"
+	do_facet_create_file client $T 10K ||
+		error_noexit "Create file $T"
+
+	drop_request "mv $T $R" ||
+		error_noexit "Rename $T"
+
+	drop_reint_reply "mv $R $RR" ||
+		error_noexit "Failed rename replay on $R"
+
+	do_facet client "checkstat -v $RR" ||
+		error_noexit "checkstat error on $RR"
+
+	do_facet client "rm $RR" ||
+		error_noexit "Can't remove file $RR"
 }
 run_test 5 "rename: drop req, drop rep"
 
-[ ! -e $RENAMED_AGAIN ] && cp $SAMPLE_FILE $RENAMED_AGAIN
-LINK1=$DIR/f0.link1
-LINK2=$DIR/f0.link2
-
 test_6() {
-    drop_request "mlink $RENAMED_AGAIN $LINK1" || return 1
-    drop_reint_reply "mlink $RENAMED_AGAIN $LINK2"   || return 2
-}
-run_test 6 "link: drop req, drop rep"
+	local T=$DIR/$tfile
+	local LINK1=$DIR/$tfile.link1
+	local LINK2=$DIR/$tfile.link2
+
+	do_facet_create_file client $T 10K ||
+		error_noexit "Create file $T"
 
-[ ! -e $LINK1 ] && mlink $RENAMED_AGAIN $LINK1
-[ ! -e $LINK2 ] && mlink $RENAMED_AGAIN $LINK2
-test_7() {
-    drop_request "munlink $LINK1"   || return 1
-    drop_reint_reply "munlink $LINK2"     || return 2
+	drop_request "mlink $T $LINK1" ||
+		error_noexit "mlink request for $T"
+
+	drop_reint_reply "mlink $T $LINK2" ||
+		error_noexit "mlink reply for $T"
+
+	drop_request "munlink $LINK1" ||
+		error_noexit "munlink request for $T"
+
+	drop_reint_reply "munlink $LINK2" ||
+		error_noexit "munlink reply for $T"
+
+	do_facet client "rm $T" ||
+		error_noexit "Can't remove file $T"
 }
-run_test 7 "unlink: drop req, drop rep"
+run_test 6 "link, unlink: drop req, drop rep"
 
 #bug 1423
 test_8() {
@@ -102,67 +122,94 @@ run_test 8 "touch: drop rep (bug 1423)"
 
 #bug 1420
 test_9() {
-    remote_ost_nodsh && skip "remote OST with nodsh" && return 0
+	remote_ost_nodsh && skip "remote OST with nodsh" && return 0
 
-    pause_bulk "cp /etc/profile $DIR/$tfile"       || return 1
-    do_facet client "cp $SAMPLE_FILE $DIR/${tfile}.2"  || return 2
-    do_facet client "sync"
-    do_facet client "rm $DIR/$tfile $DIR/${tfile}.2" || return 3
+	local t1=${tfile}.1
+	local t2=${tfile}.2
+	do_facet_random_file client $TMP/$tfile 1K ||
+		error_noexit "Create random file $TMP/$tfile"
+	# make this big, else test 9 doesn't wait for bulk -- bz 5595
+	do_facet_create_file client $TMP/$t1 4M ||
+		error_noexit "Create file $TMP/$t1"
+	do_facet client "cp $TMP/$t1 $DIR/$t1" ||
+		error_noexit "Can't copy to $DIR/$t1 file"
+	pause_bulk "cp $TMP/$tfile $DIR/$tfile" ||
+		error_noexit "Can't pause_bulk copy"
+	do_facet client "cp $TMP/$t1 $DIR/$t2" ||
+		error_noexit "Can't copy file"
+	do_facet client "sync"
+	do_facet client "rm $DIR/$tfile $DIR/$t2 $DIR/$t1" ||
+		error_noexit "Can't remove files"
+	do_facet client "rm $TMP/$t1 $TMP/$tfile"
 }
 run_test 9 "pause bulk on OST (bug 1420)"
 
 #bug 1521
 test_10() {
-    do_facet client mcreate $DIR/$tfile        || return 1
-    drop_bl_callback "chmod 0777 $DIR/$tfile"  || echo "evicted as expected"
-    # wait for the mds to evict the client
-    #echo "sleep $(($TIMEOUT*2))"
-    #sleep $(($TIMEOUT*2))
-    do_facet client touch $DIR/$tfile || echo "touch failed, evicted"
-    do_facet client checkstat -v -p 0777 $DIR/$tfile  || return 3
-    do_facet client "munlink $DIR/$tfile"
+	do_facet client mcreate $DIR/$tfile ||
+		{ error "mcreate failed: $?"; return 1; }
+	drop_bl_callback "chmod 0777 $DIR/$tfile"  || echo "evicted as expected"
+	# wait for the mds to evict the client
+	#echo "sleep $(($TIMEOUT*2))"
+	#sleep $(($TIMEOUT*2))
+	do_facet client touch $DIR/$tfile || echo "touch failed, evicted"
+	do_facet client checkstat -v -p 0777 $DIR/$tfile ||
+		{ error "client checkstat failed: $?"; return 3; }
+	do_facet client "munlink $DIR/$tfile"
+	# allow recovery to complete
+	client_up || client_up || sleep $TIMEOUT
 }
 run_test 10 "finish request on server after client eviction (bug 1521)"
 
 #bug 2460
 # wake up a thread waiting for completion after eviction
 test_11(){
-    do_facet client multiop $DIR/$tfile Ow  || return 1
-    do_facet client multiop $DIR/$tfile or  || return 2
+	do_facet client $MULTIOP $DIR/$tfile Ow  ||
+		{ error "multiop write failed: $?"; return 1; }
+	do_facet client $MULTIOP $DIR/$tfile or  ||
+		{ error "multiop read failed: $?"; return 2; }
 
-    cancel_lru_locks osc
+	cancel_lru_locks osc
 
-    do_facet client multiop $DIR/$tfile or  || return 3
-    drop_bl_callback multiop $DIR/$tfile Ow || echo "evicted as expected"
+	do_facet client $MULTIOP $DIR/$tfile or  ||
+		{ error "multiop read failed: $?"; return 3; }
+	drop_bl_callback $MULTIOP $DIR/$tfile Ow || echo "evicted as expected"
 
-    do_facet client munlink $DIR/$tfile  || return 4
+	do_facet client munlink $DIR/$tfile ||
+		{ error "munlink failed: $?"; return 4; }
+	# allow recovery to complete
+	client_up || client_up || sleep $TIMEOUT
 }
 run_test 11 "wake up a thread waiting for completion after eviction (b=2460)"
 
 #b=2494
 test_12(){
-    $LCTL mark multiop $DIR/$tfile OS_c 
-    do_facet $SINGLEMDS "lctl set_param fail_loc=0x115"
-    clear_failloc $SINGLEMDS $((TIMEOUT * 2)) &
-    multiop_bg_pause $DIR/$tfile OS_c || return 1
-    PID=$!
+	$LCTL mark $MULTIOP $DIR/$tfile OS_c
+	do_facet $SINGLEMDS "lctl set_param fail_loc=0x115"
+	clear_failloc $SINGLEMDS $((TIMEOUT * 2)) &
+	multiop_bg_pause $DIR/$tfile OS_c ||
+		{ error "multiop failed: $?"; return 1; }
+	PID=$!
 #define OBD_FAIL_MDS_CLOSE_NET           0x115
-    kill -USR1 $PID
-    echo "waiting for multiop $PID"
-    wait $PID || return 2
-    do_facet client munlink $DIR/$tfile  || return 3
+	kill -USR1 $PID
+	echo "waiting for multiop $PID"
+	wait $PID || { error "wait for multiop faile: $?"; return 2; }
+	do_facet client munlink $DIR/$tfile ||
+		{ error "client munlink failed: $?"; return 3; }
+	# allow recovery to complete
+	client_up || client_up || sleep $TIMEOUT
 }
 run_test 12 "recover from timed out resend in ptlrpcd (b=2494)"
 
 # Bug 113, check that readdir lost recv timeout works.
 test_13() {
-    mkdir -p $DIR/$tdir || return 1
-    touch $DIR/$tdir/newentry || return
+	mkdir -p $DIR/$tdir || { error "mkdir failed: $?"; return 1; }
+	touch $DIR/$tdir/newentry || { error "touch failed: $?"; return 2; }
 # OBD_FAIL_MDS_READPAGE_NET|OBD_FAIL_ONCE
-    do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000104"
-    ls $DIR/$tdir || return 3
-    do_facet $SINGLEMDS "lctl set_param fail_loc=0"
-    rm -rf $DIR/$tdir || return 4
+	do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000104"
+	ls $DIR/$tdir || { error "ls failed: $?"; return 3; }
+	do_facet $SINGLEMDS "lctl set_param fail_loc=0"
+	rm -rf $DIR/$tdir || { error "remove test dir failed: $?"; return 4; }
 }
 run_test 13 "mdc_readpage restart test (bug 1138)"
 
@@ -194,22 +241,25 @@ start_read_ahead() {
 }
 
 test_16() {
-    remote_ost_nodsh && skip "remote OST with nodsh" && return 0
+	remote_ost_nodsh && skip "remote OST with nodsh" && return 0
 
-    do_facet client cp $SAMPLE_FILE $DIR
-    sync
-    stop_read_ahead
+	do_facet_random_file client $TMP/$tfile 100K ||
+		{ error_noexit "Create random file $TMP/$T" ; return 0; }
+	do_facet client "cp $TMP/$tfile $DIR/$tfile" ||
+		{ error_noexit "Copy to $DIR/$tfile file" ; return 0; }
+	sync
+	stop_read_ahead
 
 #define OBD_FAIL_PTLRPC_BULK_PUT_NET 0x504 | OBD_FAIL_ONCE
-    do_facet ost1 "lctl set_param fail_loc=0x80000504"
-    cancel_lru_locks osc
-    # OST bulk will time out here, client resends
-    do_facet client "cmp $SAMPLE_FILE $DIR/${SAMPLE_FILE##*/}" || return 1
-    do_facet ost1 lctl set_param fail_loc=0
-    # give recovery a chance to finish (shouldn't take long)
-    sleep $TIMEOUT
-    do_facet client "cmp $SAMPLE_FILE $DIR/${SAMPLE_FILE##*/}" || return 2
-    start_read_ahead
+	do_facet ost1 "lctl set_param fail_loc=0x80000504"
+	cancel_lru_locks osc
+	# OST bulk will time out here, client resends
+	do_facet client "cmp $TMP/$tfile $DIR/$tfile" || return 1
+	do_facet ost1 lctl set_param fail_loc=0
+	# give recovery a chance to finish (shouldn't take long)
+	sleep $TIMEOUT
+	do_facet client "cmp $TMP/$tfile $DIR/$tfile" || return 2
+	start_read_ahead
 }
 run_test 16 "timeout bulk put, don't evict client (2732)"
 
@@ -218,6 +268,10 @@ test_17() {
 
     remote_ost_nodsh && skip "remote OST with nodsh" && return 0
 
+	local SAMPLE_FILE=$TMP/$tfile
+	do_facet_random_file client $SAMPLE_FILE 20K ||
+		{ error_noexit "Create random file $SAMPLE_FILE" ; return 0; }
+
     # With adaptive timeouts, bulk_get won't expire until adaptive_timeout_max
     if at_is_enabled; then
         at_max_saved=$(at_max_get ost1)
@@ -228,7 +282,7 @@ test_17() {
     # OST bulk will time out here, client retries
     do_facet ost1 lctl set_param fail_loc=0x80000503
     # need to ensure we send an RPC
-    do_facet client cp /etc/termcap $DIR/$tfile
+    do_facet client cp $SAMPLE_FILE $DIR/$tfile
     sync
 
     # with AT, client will wait adaptive_max*factor+net_latency before
@@ -238,7 +292,7 @@ test_17() {
     do_facet ost1 lctl set_param fail_loc=0
     do_facet client "df $DIR"
     # expect cmp to succeed, client resent bulk
-    do_facet client "cmp /etc/termcap $DIR/$tfile" || return 3
+    do_facet client "cmp $SAMPLE_FILE $DIR/$tfile" || return 3
     do_facet client "rm $DIR/$tfile" || return 4
     [ $at_max_saved -ne 0 ] && at_max_set $at_max_saved ost1
     return 0
@@ -248,6 +302,9 @@ run_test 17 "timeout bulk get, don't evict client (2732)"
 test_18a() {
     [ -z ${ost2_svc} ] && skip_env "needs 2 osts" && return 0
 
+	do_facet_create_file client $TMP/$tfile 20K ||
+		{ error_noexit "Create file $TMP/$tfile" ; return 0; }
+
     do_facet client mkdir -p $DIR/$tdir
     f=$DIR/$tdir/$tfile
 
@@ -255,14 +312,14 @@ test_18a() {
     pgcache_empty || return 1
 
     # 1 stripe on ost2
-    lfs setstripe $f -s $((128 * 1024)) -i 1 -c 1
-    get_stripe_info client $f
+    $LFS setstripe -i 1 -c 1 $f
+    stripe_index=$($LFS getstripe -i $f)
     if [ $stripe_index -ne 1 ]; then
-        lfs getstripe $f
-        error "$f: different stripe offset ($stripe_index)" && return
+        $LFS getstripe $f
+        error "$f: stripe_index $stripe_index != 1" && return
     fi
 
-    do_facet client cp $SAMPLE_FILE $f
+    do_facet client cp $TMP/$tfile $f
     sync
     local osc2dev=`lctl get_param -n devices | grep ${ost2_svc}-osc- | egrep -v 'MDT' | awk '{print $1}'`
     $LCTL --device $osc2dev deactivate || return 3
@@ -279,21 +336,23 @@ run_test 18a "manual ost invalidate clears page cache immediately"
 test_18b() {
     remote_ost_nodsh && skip "remote OST with nodsh" && return 0
 
+	do_facet_create_file client $TMP/$tfile 20K ||
+		{ error_noexit "Create file $TMP/$tfile" ; return 0; }
+
     do_facet client mkdir -p $DIR/$tdir
     f=$DIR/$tdir/$tfile
 
     cancel_lru_locks osc
     pgcache_empty || return 1
 
-    # shouldn't have to set stripe size of count==1
-    lfs setstripe $f -s $((128 * 1024)) -i 0 -c 1
-    get_stripe_info client $f
+    $LFS setstripe -i 0 -c 1 $f
+    stripe_index=$($LFS getstripe -i $f)
     if [ $stripe_index -ne 0 ]; then
-        lfs getstripe $f
-        error "$f: different stripe offset ($stripe_index)" && return
+        $LFS getstripe $f
+        error "$f: stripe_index $stripe_index != 0" && return
     fi
 
-    do_facet client cp $SAMPLE_FILE $f
+    do_facet client cp $TMP/$tfile $f
     sync
     ost_evict_client
     # allow recovery to complete
@@ -310,21 +369,23 @@ run_test 18b "eviction and reconnect clears page cache (2766)"
 test_18c() {
     remote_ost_nodsh && skip "remote OST with nodsh" && return 0
 
+	do_facet_create_file client $TMP/$tfile 20K ||
+		{ error_noexit "Create file $TMP/$tfile" ; return 0; }
+
     do_facet client mkdir -p $DIR/$tdir
     f=$DIR/$tdir/$tfile
 
     cancel_lru_locks osc
     pgcache_empty || return 1
 
-    # shouldn't have to set stripe size of count==1
-    lfs setstripe $f -s $((128 * 1024)) -i 0 -c 1
-    get_stripe_info client $f
+    $LFS setstripe -i 0 -c 1 $f
+    stripe_index=$($LFS getstripe -i $f)
     if [ $stripe_index -ne 0 ]; then
-        lfs getstripe $f
-        error "$f: different stripe offset ($stripe_index)" && return
+        $LFS getstripe $f
+        error "$f: stripe_index $stripe_index != 0" && return
     fi
 
-    do_facet client cp $SAMPLE_FILE $f
+    do_facet client cp $TMP/$tfile $f
     sync
     ost_evict_client
 
@@ -332,6 +393,7 @@ test_18c() {
     # lost reply to connect request
     do_facet ost1 lctl set_param fail_loc=0x80000225
     # force reconnect
+    sleep 1
     df $MOUNT > /dev/null 2>&1
     sleep 2
     # my understanding is that there should be nothing in the page
@@ -344,36 +406,81 @@ test_18c() {
 run_test 18c "Dropped connect reply after eviction handing (14755)"
 
 test_19a() {
-    f=$DIR/$tfile
-    do_facet client mcreate $f        || return 1
-    drop_ldlm_cancel "chmod 0777 $f"  || echo "evicted as expected"
+	local BEFORE=`date +%s`
+	local EVICT
 
-    do_facet client checkstat -v -p 0777 $f  || echo evicted
-    # let the client reconnect
-    sleep 5
-    do_facet client "munlink $f"
+	mount_client $DIR2
+
+	do_facet client mcreate $DIR/$tfile        || return 1
+	drop_ldlm_cancel "chmod 0777 $DIR2"
+
+	umount_client $DIR2
+	do_facet client "munlink $DIR/$tfile"
+
+	# let the client reconnect
+	client_reconnect
+	EVICT=$(do_facet client $LCTL get_param mdc.$FSNAME-MDT*.state | \
+	    awk -F"[ [,]" '/EVICTED]$/ { if (mx<$4) {mx=$4;} } END { print mx }')
+
+	[ ! -z "$EVICT" ] && [[ $EVICT -gt $BEFORE ]] || error "no eviction"
 }
 run_test 19a "test expired_lock_main on mds (2867)"
 
 test_19b() {
-    f=$DIR/$tfile
-    do_facet client multiop $f Ow  || return 1
-    do_facet client multiop $f or  || return 2
+	local BEFORE=`date +%s`
+	local EVICT
 
-    cancel_lru_locks osc
+	mount_client $DIR2
+
+	do_facet client $MULTIOP $DIR/$tfile Ow  || return 1
+	drop_ldlm_cancel $MULTIOP $DIR2/$tfile Ow
+	umount_client $DIR2
+	do_facet client munlink $DIR/$tfile
 
-    do_facet client multiop $f or  || return 3
-    drop_ldlm_cancel multiop $f Ow  || echo "client evicted, as expected"
+	# let the client reconnect
+	client_reconnect
+	EVICT=$(do_facet client $LCTL get_param osc.$FSNAME-OST*.state | \
+	    awk -F"[ [,]" '/EVICTED]$/ { if (mx<$4) {mx=$4;} } END { print mx }')
 
-    do_facet client munlink $f  || return 4
+	[ ! -z "$EVICT" ] && [[ $EVICT -gt $BEFORE ]] || error "no eviction"
 }
 run_test 19b "test expired_lock_main on ost (2867)"
 
+test_19c() {
+	local BEFORE=`date +%s`
+
+	mount_client $DIR2
+	$LCTL set_param ldlm.namespaces.*.early_lock_cancel=0
+
+	mkdir -p $DIR1/$tfile
+	stat $DIR1/$tfile
+
+#define OBD_FAIL_PTLRPC_CANCEL_RESEND 0x516
+	do_facet mds $LCTL set_param fail_loc=0x80000516
+
+	touch $DIR2/$tfile/file1 &
+	PID1=$!
+	# let touch to get blocked on the server
+	sleep 2
+
+	wait $PID1
+	$LCTL set_param ldlm.namespaces.*.early_lock_cancel=1
+	umount_client $DIR2
+
+	# let the client reconnect
+	sleep 5
+	EVICT=$(do_facet client $LCTL get_param mdc.$FSNAME-MDT*.state |
+	   awk -F"[ [,]" '/EVICTED]$/ { if (mx<$4) {mx=$4;} } END { print mx }')
+
+	[ -z "$EVICT" ] || [[ $EVICT -le $BEFORE ]] || error "eviction happened"
+}
+run_test 19c "check reconnect and lock resend do not trigger expired_lock_main"
+
 test_20a() {	# bug 2983 - ldlm_handle_enqueue cleanup
 	remote_ost_nodsh && skip "remote OST with nodsh" && return 0
 
 	mkdir -p $DIR/$tdir
-	lfs setstripe $DIR/$tdir/${tfile} -i 0 -c 1
+	$LFS setstripe -i 0 -c 1 $DIR/$tdir/${tfile}
 	multiop_bg_pause $DIR/$tdir/${tfile} O_wc || return 1
 	MULTI_PID=$!
 	cancel_lru_locks osc
@@ -390,7 +497,7 @@ test_20b() {	# bug 2986 - ldlm_handle_enqueue error during open
 	remote_ost_nodsh && skip "remote OST with nodsh" && return 0
 
 	mkdir -p $DIR/$tdir
-	lfs setstripe $DIR/$tdir/${tfile} -i 0 -c 1
+	$LFS setstripe -i 0 -c 1 $DIR/$tdir/${tfile}
 	cancel_lru_locks osc
 #define OBD_FAIL_LDLM_ENQUEUE_EXTENT_ERR 0x308
 	do_facet ost1 lctl set_param fail_loc=0x80000308
@@ -406,7 +513,7 @@ test_21a() {
        close_pid=$!
 
        do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000129"
-       multiop $DIR/$tdir-2/f Oc &
+       $MULTIOP $DIR/$tdir-2/f Oc &
        open_pid=$!
        sleep 1
        do_facet $SINGLEMDS "lctl set_param fail_loc=0"
@@ -481,7 +588,7 @@ test_21d() {
        pid=$!
 
        do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000129"
-       multiop $DIR/$tdir-2/f Oc &
+       $MULTIOP $DIR/$tdir-2/f Oc &
        sleep 1
        do_facet $SINGLEMDS "lctl set_param fail_loc=0"
 
@@ -598,11 +705,11 @@ test_22() {
     f2=$DIR/${tfile}-2
     
     do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000115"
-    multiop $f2 Oc &
+    $MULTIOP $f2 Oc &
     close_pid=$!
 
     sleep 1
-    multiop $f1 msu || return 1
+    $MULTIOP $f1 msu || return 1
 
     cancel_lru_locks mdc
     do_facet $SINGLEMDS "lctl set_param fail_loc=0"
@@ -627,11 +734,11 @@ test_23() { #b=4561
 }
 run_test 23 "client hang when close a file after mds crash"
 
-test_24() { # bug 11710 details correct fsync() behavior
+test_24a() { # bug 11710 details correct fsync() behavior
 	remote_ost_nodsh && skip "remote OST with nodsh" && return 0
 
 	mkdir -p $DIR/$tdir
-	lfs setstripe $DIR/$tdir -s 0 -i 0 -c 1
+	$LFS setstripe -i 0 -c 1 $DIR/$tdir
 	cancel_lru_locks osc
 	multiop_bg_pause $DIR/$tdir/$tfile Owy_wyc || return 1
 	MULTI_PID=$!
@@ -643,16 +750,51 @@ test_24() { # bug 11710 details correct fsync() behavior
 	client_reconnect
 	[ $rc -eq 0 ] && error_ignore 5494 "multiop didn't fail fsync: rc $rc" || true
 }
-run_test 24 "fsync error (should return error)"
+run_test 24a "fsync error (should return error)"
 
 wait_client_evicted () {
 	local facet=$1
 	local exports=$2
 	local varsvc=${facet}_svc
 
-	wait_update $(facet_host $facet) "lctl get_param -n *.${!varsvc}.num_exports | cut -d' ' -f2" $((exports - 1)) $3
+	wait_update $(facet_active_host $facet) \
+		"lctl get_param -n *.${!varsvc}.num_exports | cut -d' ' -f2" \
+		$((exports - 1)) $3
 }
 
+test_24b() {
+	remote_ost_nodsh && skip "remote OST with nodsh" && return 0
+
+	dmesg -c
+	mkdir -p $DIR/$tdir
+	lfs setstripe $DIR/$tdir -s 0 -i 0 -c 1
+	cancel_lru_locks osc
+	multiop_bg_pause $DIR/$tdir/$tfile-1 Ow8192_yc ||
+		error "mulitop Ow8192_yc failed"
+
+	MULTI_PID1=$!
+	multiop_bg_pause $DIR/$tdir/$tfile-2 Ow8192_c ||
+		error "mulitop Ow8192_c failed"
+
+	MULTI_PID2=$!
+	ost_evict_client
+
+	kill -USR1 $MULTI_PID1
+	wait $MULTI_PID1
+	rc1=$?
+	kill -USR1 $MULTI_PID2
+	wait $MULTI_PID2
+	rc2=$?
+	lctl set_param fail_loc=0x0
+	client_reconnect
+	[ $rc1 -eq 0 -o $rc2 -eq 0 ] &&
+	error_ignore "multiop didn't fail fsync: $rc1 or close: $rc2" || true
+
+	dmesg | grep "dirty page discard:" || \
+		error "no discarded dirty page found!"
+}
+run_test 24b "test dirty page discard due to client eviction"
+
 test_26a() {      # was test_26 bug 5921 - evict dead exports by pinger
 # this test can only run from a client on a separate node.
 	remote_ost || { skip "local OST" && return 0; }
@@ -732,12 +874,8 @@ test_27() {
 #define OBD_FAIL_OSC_SHUTDOWN            0x407
 	do_facet $SINGLEMDS lctl set_param fail_loc=0x80000407
 	# need to wait for reconnect
-	echo -n waiting for fail_loc
-	while [ $(do_facet $SINGLEMDS lctl get_param -n fail_loc) -eq -2147482617 ]; do
-	    sleep 1
-	    echo -n .
-	done
-	do_facet $SINGLEMDS lctl get_param -n fail_loc
+	echo waiting for fail_loc
+	wait_update_facet $SINGLEMDS "lctl get_param -n fail_loc" "-2147482617"
 	facet_failover $SINGLEMDS
 	#no crashes allowed!
         kill -USR1 $CLIENT_PID
@@ -805,6 +943,9 @@ test_50() {
 run_test 50 "failover MDS under load"
 
 test_51() {
+	#define OBD_FAIL_MDS_SYNC_CAPA_SL                    0x1310
+	do_facet ost1 lctl set_param fail_loc=0x00001310
+
 	mkdir -p $DIR/$tdir
 	# put a load of file creates/writes/deletes
 	writemany -q $DIR/$tdir/$tfile 0 5 &
@@ -816,10 +957,11 @@ test_51() {
 	SEQ="1 5 10 $(seq $TIMEOUT 5 $(($TIMEOUT+10)))"
         echo will failover at $SEQ
         for i in $SEQ
-          do
-          echo failover in $i sec
-          sleep $i
-          facet_failover $SINGLEMDS
+        do
+		#echo failover in $i sec
+		log "test_$testnum: failover in $i sec"
+		sleep $i
+		facet_failover $SINGLEMDS
         done
 	# client process should see no problems even though MDS went down
 	# and recovery was interrupted
@@ -898,7 +1040,7 @@ test_55() {
 	mkdir -p $DIR/$tdir
 
 	# first dd should be finished quickly
-	lfs setstripe $DIR/$tdir/$tfile-1 -c 1 -i 0
+	$LFS setstripe -c 1 -i 0 $DIR/$tdir/$tfile-1
 	dd if=/dev/zero of=$DIR/$tdir/$tfile-1 bs=32M count=4  &
 	DDPID=$!
 	count=0
@@ -913,7 +1055,7 @@ test_55() {
 	done	
 	echo "(dd_pid=$DDPID, time=$count)successful"
 
-	lfs setstripe $DIR/$tdir/$tfile-2 -c 1 -i 0
+	$LFS setstripe -c 1 -i 0 $DIR/$tdir/$tfile-2
 	#define OBD_FAIL_OST_DROP_REQ            0x21d
 	do_facet ost1 lctl set_param fail_loc=0x0000021d
 	# second dd will be never finished
@@ -995,7 +1137,10 @@ test_58() { # bug 11546
         lctl set_param fail_loc=0
         drop_bl_callback rm -f $DIR/$tfile
         wait $pid
-        do_facet client "df $DIR"
+        # the first 'df' could tigger the eviction caused by
+        # 'drop_bl_callback', and it's normal case.
+        # but the next 'df' should return successfully.
+        do_facet client "df $DIR" || do_facet client "df $DIR"
 }
 run_test 58 "Eviction in the middle of open RPC reply processing"
 
@@ -1080,13 +1225,15 @@ run_test 60 "Add Changelog entries during MDS failover"
 
 test_61()
 {
-	local cflags='osc.*-OST0000-osc-MDT*.connect_flags'
+	local mdtosc=$(get_mdtosc_proc_path $SINGLEMDS $FSNAME-OST0000)
+	mdtosc=${mdtosc/-MDT*/-MDT\*}
+	local cflags="osc.$mdtosc.connect_flags"
 	do_facet $SINGLEMDS "lctl get_param -n $cflags" |grep -q skip_orphan
 	[ $? -ne 0 ] && skip "don't have skip orphan feature" && return
 
 	mkdir -p $DIR/$tdir || error "mkdir dir $DIR/$tdir failed"
 	# Set the default stripe of $DIR/$tdir to put the files to ost1
-	$LFS setstripe -c 1 --index 0 $DIR/$tdir
+	$LFS setstripe -c 1 -i 0 $DIR/$tdir
 
 	replay_barrier $SINGLEMDS
 	createmany -o $DIR/$tdir/$tfile-%d 10 
@@ -1095,7 +1242,7 @@ test_61()
 	fail_abort $SINGLEMDS
 	
 	touch $DIR/$tdir/$tfile
-	local id=`$LFS getstripe $DIR/$tdir/$tfile |awk '($1 ~ 0 && $2 ~ /^[1-9]+/) {print $2}'`
+	local id=`$LFS getstripe $DIR/$tdir/$tfile | awk '$1 == 0 { print $2 }'`
 	[ $id -le $oid ] && error "the orphan objid was reused, failed"
 
 	# Cleanup
@@ -1103,6 +1250,399 @@ test_61()
 }
 run_test 61 "Verify to not reuse orphan objects - bug 17025"
 
-equals_msg `basename $0`: test complete, cleaning up
+check_cli_ir_state()
+{
+        local NODE=${1:-$HOSTNAME}
+        local st
+        st=$(do_node $NODE "lctl get_param mgc.*.ir_state |
+                            awk '/imperative_recovery:/ { print \\\$2}'")
+	[ $st != ON -o $st != OFF -o $st != ENABLED -o $st != DISABLED ] ||
+		error "Error state $st, must be ENABLED or DISABLED"
+        echo -n $st
+}
+
+check_target_ir_state()
+{
+        local target=${1}
+        local name=${target}_svc
+        local recovery_proc=obdfilter.${!name}.recovery_status
+        local st
+
+        st=$(do_facet $target "lctl get_param -n $recovery_proc |
+                               awk '/IR:/{ print \\\$2}'")
+	[ $st != ON -o $st != OFF -o $st != ENABLED -o $st != DISABLED ] ||
+		error "Error state $st, must be ENABLED or DISABLED"
+        echo -n $st
+}
+
+set_ir_status()
+{
+        do_facet mgs lctl set_param -n mgs.MGS.live.$FSNAME="state=$1"
+}
+
+get_ir_status()
+{
+        local state=$(do_facet mgs "lctl get_param -n mgs.MGS.live.$FSNAME |
+                                    awk '/state:/{ print \\\$2 }'")
+        echo -n ${state/,/}
+}
+
+nidtbl_version_mgs()
+{
+        local ver=$(do_facet mgs "lctl get_param -n mgs.MGS.live.$FSNAME |
+                                  awk '/nidtbl_version:/{ print \\\$2 }'")
+        echo -n $ver
+}
+
+# nidtbl_version_client <mds1|client> [node]
+nidtbl_version_client()
+{
+        local cli=$1
+        local node=${2:-$HOSTNAME}
+
+        if [ X$cli = Xclient ]; then
+                cli=$FSNAME-client
+        else
+                local obdtype=${cli/%[0-9]*/}
+                [ $obdtype != mds ] && error "wrong parameters $cli"
+
+                node=$(facet_active_host $cli)
+                local t=${cli}_svc
+                cli=${!t}
+        fi
+
+        local vers=$(do_node $node "lctl get_param -n mgc.*.ir_state" |
+                     awk "/$cli/{print \$6}" |sort -u)
+
+        # in case there are multiple mounts on the client node
+        local arr=($vers)
+        [ ${#arr[@]} -ne 1 ] && error "versions on client node mismatch"
+        echo -n $vers
+}
+
+nidtbl_versions_match()
+{
+        [ $(nidtbl_version_mgs) -eq $(nidtbl_version_client ${1:-client}) ]
+}
+
+target_instance_match()
+{
+        local srv=$1
+        local obdtype
+        local cliname
+
+        obdtype=${srv/%[0-9]*/}
+        case $obdtype in
+        mds)
+                obdname="mdt"
+                cliname="mdc"
+                ;;
+        ost)
+                obdname="obdfilter"
+                cliname="osc"
+                ;;
+        *)
+                error "invalid target type" $srv
+                return 1
+                ;;
+        esac
+
+        local target=${srv}_svc
+        local si=$(do_facet $srv lctl get_param -n $obdname.${!target}.instance)
+        local ci=$(lctl get_param -n $cliname.${!target}-${cliname}-*.import | \
+                  awk '/instance/{ print $2 }' |head -1)
+
+        return $([ $si -eq $ci ])
+}
+
+test_100()
+{
+	do_facet mgs $LCTL list_param mgs.*.ir_timeout ||
+		{ skip "MGS without IR support"; return 0; }
+
+	# MDT was just restarted in the previous test, make sure everything
+	# is all set.
+	local cnt=30
+	while [ $cnt -gt 0 ]; do
+		nidtbl_versions_match && break
+		sleep 1
+		cnt=$((cnt - 1))
+	done
+
+	# disable IR
+	set_ir_status disabled
+
+	local prev_ver=$(nidtbl_version_client client)
+
+        local saved_FAILURE_MODE=$FAILURE_MODE
+        [ $(facet_host mgs) = $(facet_host ost1) ] && FAILURE_MODE="SOFT"
+        fail ost1
+
+        # valid check
+	[ $(nidtbl_version_client client) -eq $prev_ver ] ||
+		error "version must not change due to IR disabled"
+        target_instance_match ost1 || error "instance mismatch"
+
+        # restore env
+        set_ir_status full
+        FAILURE_MODE=$saved_FAILURE_MODE
+}
+run_test 100 "IR: Make sure normal recovery still works w/o IR"
+
+test_101()
+{
+        do_facet mgs $LCTL list_param mgs.*.ir_timeout ||
+                { skip "MGS without IR support"; return 0; }
+
+        set_ir_status full
+
+        local OST1_IMP=$(get_osc_import_name client ost1)
+
+        # disable pinger recovery
+        lctl set_param -n osc.$OST1_IMP.pinger_recov=0
+
+        fail ost1
+
+        target_instance_match ost1 || error "instance mismatch"
+        nidtbl_versions_match || error "version must match"
+
+        lctl set_param -n osc.$OST1_IMP.pinger_recov=1
+}
+run_test 101 "IR: Make sure IR works w/o normal recovery"
+
+test_102()
+{
+        do_facet mgs $LCTL list_param mgs.*.ir_timeout ||
+                { skip "MGS without IR support"; return 0; }
+
+        local clients=${CLIENTS:-$HOSTNAME}
+        local old_version
+        local new_version
+        local mgsdev=mgs
+
+        set_ir_status full
+
+        # let's have a new nidtbl version
+        fail ost1
+
+        # sleep for a while so that clients can see the failure of ost
+        # it must be MGC_TIMEOUT_MIN_SECONDS + MGC_TIMEOUT_RAND_CENTISEC.
+        # int mgc_request.c:
+        # define MGC_TIMEOUT_MIN_SECONDS   5
+        # define MGC_TIMEOUT_RAND_CENTISEC 0x1ff /* ~500 *
+        local count=30  # 20 seconds at most
+        while [ $count -gt 0 ]; do
+                nidtbl_versions_match && break
+                sleep 1
+                count=$((count-1))
+        done
+
+        nidtbl_versions_match || error "nidtbl mismatch"
+
+        # get the version #
+        old_version=$(nidtbl_version_client client)
+
+        zconf_umount_clients $clients $MOUNT || error "Cannot umount client"
+
+        # restart mgs
+        combined_mgs_mds && mgsdev=mds1
+        remount_facet $mgsdev
+        fail ost1
+
+        zconf_mount_clients $clients $MOUNT || error "Cannot mount client"
+
+        # check new version
+        new_version=$(nidtbl_version_client client)
+        [ $new_version -lt $old_version ] &&
+                error "nidtbl version wrong after mgs restarts"
+        return 0
+}
+run_test 102 "IR: New client gets updated nidtbl after MGS restart"
+
+test_103()
+{
+        do_facet mgs $LCTL list_param mgs.*.ir_timeout ||
+                { skip "MGS without IR support"; return 0; }
+
+        combined_mgs_mds && skip "mgs and mds on the same target" && return 0
+
+        # workaround solution to generate config log on the mds
+        remount_facet mds1
+
+        stop mgs
+        stop mds1
+
+        # We need this test because mds is like a client in IR context.
+        start mds1 $MDSDEV1 || error "MDS should start w/o mgs"
+
+        # start mgs and remount mds w/ ir
+        start mgs $MGSDEV
+        clients_up
+
+        # remount client so that fsdb will be created on the MGS
+        umount_client $MOUNT || error "umount failed"
+        mount_client $MOUNT || error "mount failed"
+
+        # sleep 30 seconds so the MDS has a chance to detect MGS restarting
+        local count=30
+        while [ $count -gt 0 ]; do
+                [ $(nidtbl_version_client mds1) -ne 0 ] && break
+                sleep 1
+                count=$((count-1))
+        done
+
+        # after a while, mds should be able to reconnect to mgs and fetch
+        # up-to-date nidtbl version
+        nidtbl_versions_match mds1 || error "mds nidtbl mismatch"
+
+        # reset everything
+        set_ir_status full
+}
+run_test 103 "IR: MDS can start w/o MGS and get updated nidtbl later"
+
+test_104()
+{
+        do_facet mgs $LCTL list_param mgs.*.ir_timeout ||
+                { skip "MGS without IR support"; return 0; }
+
+        set_ir_status full
+
+        stop ost1
+        start ost1 $(ostdevname 1) "$OST_MOUNT_OPTS -onoir" ||
+                error "OST1 cannot start"
+        clients_up
+
+        local ir_state=$(check_target_ir_state ost1)
+	[ $ir_state = "DISABLED" -o $ir_state = "OFF" ] ||
+		error "ir status on ost1 should be DISABLED"
+}
+run_test 104 "IR: ost can disable IR voluntarily"
+
+test_105()
+{
+        [ -z "$RCLIENTS" ] && skip "Needs multiple clients" && return 0
+        do_facet mgs $LCTL list_param mgs.*.ir_timeout ||
+                { skip "MGS without IR support"; return 0; }
+
+        set_ir_status full
+
+        # get one of the clients from client list
+        local rcli=$(echo $RCLIENTS |cut -d' ' -f 1)
+
+        local old_MOUNTOPT=$MOUNTOPT
+        MOUNTOPT=${MOUNTOPT},noir
+        zconf_umount $rcli $MOUNT || error "umount failed"
+        zconf_mount $rcli $MOUNT || error "mount failed"
+
+        # make sure lustre mount at $rcli disabling IR
+        local ir_state=$(check_cli_ir_state $rcli)
+	[ $ir_state = "DISABLED" -o $ir_state = "OFF" ] ||
+		error "IR state must be DISABLED at $rcli"
+
+	# Since the client just mounted, its last_rcvd entry is not on disk.
+	# Send an RPC so exp_need_sync forces last_rcvd to commit this export
+	# so the client can reconnect during OST recovery (LU-924, LU-1582)
+	$SETSTRIPE -i 0 $DIR/$tfile
+	dd if=/dev/zero of=$DIR/$tfile bs=1M count=1 conv=sync
+
+        # make sure MGS's state is Partial
+        [ $(get_ir_status) = "partial" ] || error "MGS IR state must be partial"
+
+        fail ost1
+	# make sure IR on ost1 is DISABLED
+        local ir_state=$(check_target_ir_state ost1)
+	[ $ir_state = "DISABLED" -o $ir_state = "OFF" ] ||
+		error "IR status on ost1 should be DISABLED"
+
+        # restore it
+        MOUNTOPT=$old_MOUNTOPT
+        zconf_umount $rcli $MOUNT || error "umount failed"
+        zconf_mount $rcli $MOUNT || error "mount failed"
+
+        # make sure MGS's state is full
+        [ $(get_ir_status) = "full" ] || error "MGS IR status must be full"
+
+        fail ost1
+	# make sure IR on ost1 is ENABLED
+        local ir_state=$(check_target_ir_state ost1)
+	[ $ir_state = "ENABLED" -o $ir_state = "ON" ] ||
+		error "IR status on ost1 should be ENABLED"
+
+        return 0
+}
+run_test 105 "IR: NON IR clients support"
+
+cleanup_106() {
+	trap 0
+	umount_client $DIR2
+	debugrestore
+}
+
+test_106() { # LU-1789
+	[[ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.3.50) ]] ||
+		{ skip "Need MDS version at least 2.3.50"; return 0; }
+
+#define OBD_FAIL_MDC_LIGHTWEIGHT         0x805
+	$LCTL set_param fail_loc=0x805
+
+	debugsave
+	trap cleanup_106 EXIT
+
+	# enable lightweight flag on mdc connection
+	mount_client $DIR2
+
+	local MDS_NEXP=$(do_facet $SINGLEMDS \
+			 lctl get_param -n mdt.${mds1_svc}.num_exports |
+			 cut -d' ' -f2)
+	$LCTL set_param fail_loc=0
+
+	touch $DIR2/$tfile || error "failed to create empty file"
+	replay_barrier $SINGLEMDS
+
+	$LCTL set_param debug=console
+	$LCTL clear
+	facet_failover $SINGLEMDS
+
+	# lightweight connection must be evicted
+	touch -c $DIR2/$tfile || true
+	$LCTL dk $TMP/lustre-log-$TESTNAME.log
+	evicted=`awk '/This client was evicted by .*MDT0000/ {
+				      print;
+		      }' $TMP/lustre-log-$TESTNAME.log`
+	[ -z "$evicted" ] && error "lightweight client not evicted by mds"
+
+	# and all operations performed by lightweight client should be
+	# synchronous, so the file created before mds restart should be there
+	$CHECKSTAT -t file $DIR/$tfile || error "file not present"
+	rm -f $DIR/$tfile
+
+	cleanup_106
+}
+run_test 106 "lightweight connection support"
+
+test_107 () {
+	local CLIENT_PID
+	local close_pid
+
+	mkdir -p $DIR/$tdir
+	# OBD_FAIL_MDS_REINT_NET_REP   0x119
+	do_facet $SINGLEMDS lctl set_param fail_loc=0x119
+	multiop $DIR/$tdir D_c &
+	close_pid=$!
+	mkdir $DIR/$tdir/dir_106 &
+	CLIENT_PID=$!
+	do_facet $SINGLEMDS lctl set_param fail_loc=0
+	fail $SINGLEMDS
+
+	wait $CLIENT_PID || rc=$?
+	checkstat -t dir $DIR/$tdir/dir_106 || return 1
+
+	kill -USR1 $close_pid
+	wait $close_pid || return 2
+
+	return $rc
+}
+run_test 107 "drop reint reply, then restart MDT"
+complete $SECONDS
 check_and_cleanup_lustre
-[ -f "$TESTSUITELOG" ] && cat $TESTSUITELOG && grep -q FAIL $TESTSUITELOG && exit 1 || true
+exit_status