2 # vim:expandtab:shiftwidth=4:softtabstop=4:tabstop=4:
4 trap 'echo "test-framework exiting on error"' ERR
9 export REFORMAT=${REFORMAT:-""}
11 export GMNALNID=${GMNALNID:-/usr/sbin/gmlndnid}
12 export CATASTROPHE=${CATASTROPHE:-/proc/sys/lnet/catastrophe}
15 export GSS_PIPEFS=false
16 #export PDSH="pdsh -S -Rssh -w"
18 # eg, assert_env LUSTRE MDSNODES OSTNODES CLIENTS
22 if [ -z "${!name}" ]; then
23 echo "$0: $name must be set"
27 [ $failed ] && exit 1 || true
31 echo "usage: $0 [-r] [-f cfgfile]"
38 export LUSTRE=`absolute_path $LUSTRE`
39 export TESTSUITE=`basename $0 .sh`
40 export LTESTDIR=${LTESTDIR:-$LUSTRE/../ltest}
42 [ -d /r ] && export ROOT=${ROOT:-/r}
43 export TMP=${TMP:-$ROOT/tmp}
44 export TESTSUITELOG=${TMP}/${TESTSUITE}.log
46 export PATH=:$PATH:$LUSTRE/utils:$LUSTRE/utils/gss:$LUSTRE/tests
47 export LCTL=${LCTL:-"$LUSTRE/utils/lctl"}
48 export LFS=${LFS:-"$LUSTRE/utils/lfs"}
49 [ ! -f "$LCTL" ] && export LCTL=$(which lctl)
50 export LFS=${LFS:-"$LUSTRE/utils/lfs"}
51 [ ! -f "$LFS" ] && export LFS=$(which lfs)
52 export MKFS=${MKFS:-"$LUSTRE/utils/mkfs.lustre"}
53 [ ! -f "$MKFS" ] && export MKFS=$(which mkfs.lustre)
54 export TUNEFS=${TUNEFS:-"$LUSTRE/utils/tunefs.lustre"}
55 [ ! -f "$TUNEFS" ] && export TUNEFS=$(which tunefs.lustre)
56 export CHECKSTAT="${CHECKSTAT:-"checkstat -v"} "
57 export FSYTPE=${FSTYPE:-"ldiskfs"}
58 export NAME=${NAME:-local}
59 export LPROC=/proc/fs/lustre
60 export LGSSD=${LGSSD:-"$LUSTRE/utils/gss/lgssd"}
61 export LSVCGSSD=${LSVCGSSD:-"$LUSTRE/utils/gss/lsvcgssd"}
62 export KRB5DIR=${KRB5DIR:-"/usr/kerberos"}
65 if [ "$ACCEPTOR_PORT" ]; then
66 export PORT_OPT="--port $ACCEPTOR_PORT"
71 echo "Using GSS/krb5 ptlrpc security flavor"
77 # Paths on remote nodes, if different
78 export RLUSTRE=${RLUSTRE:-$LUSTRE}
79 export RPWD=${RPWD:-$PWD}
80 export I_MOUNTED=${I_MOUNTED:-"no"}
84 while getopts "rvf:" opt $*; do
87 r) REFORMAT=--reformat;;
96 [ "$TESTSUITELOG" ] && rm -f $TESTSUITELOG || true
104 BASE=`basename $module $EXT`
105 lsmod | grep -q ${BASE} || \
106 if [ -f ${LUSTRE}/${module}${EXT} ]; then
107 insmod ${LUSTRE}/${module}${EXT} $@
109 # must be testing a "make install" or "rpm" installation
110 # note failed to load ptlrpc_gss is considered not fatal
111 if [ "$BASE" == "ptlrpc_gss" ]; then
112 modprobe $BASE $@ 2>/dev/null || echo "gss/krb5 is not supported"
120 if [ -n "$MODPROBE" ]; then
124 if [ "$HAVE_MODULES" = true ]; then
130 echo Loading modules from $LUSTRE
131 load_module ../lnet/libcfs/libcfs
132 [ -z "$LNETOPTS" ] && \
133 LNETOPTS=$(awk '/^options lnet/ { print $0}' /etc/modprobe.conf | sed 's/^options lnet //g')
134 echo "lnet options: '$LNETOPTS'"
135 # note that insmod will ignore anything in modprobe.conf
136 load_module ../lnet/lnet/lnet $LNETOPTS
137 LNETLND=${LNETLND:-"socklnd/ksocklnd"}
138 load_module ../lnet/klnds/$LNETLND
139 [ "$FSTYPE" = "ldiskfs" ] && load_module ../ldiskfs/ldiskfs/ldiskfs
140 load_module lvfs/lvfs
141 load_module obdclass/obdclass
142 load_module lvfs/fsfilt_$FSTYPE
143 load_module ptlrpc/ptlrpc
144 load_module ptlrpc/gss/ptlrpc_gss
145 # Now, some modules depend on lquota without USE_QUOTA check,
146 # will fix later. Disable check "$USE_QUOTA" = "yes" temporary.
147 #[ "$USE_QUOTA" = "yes" ] && load_module quota/lquota
148 load_module quota/lquota
161 load_module obdfilter/obdfilter
162 load_module llite/lustre
163 load_module llite/llite_lloop
166 rm -f $TMP/ogdb-`hostname`
167 $LCTL modules > $TMP/ogdb-`hostname`
168 # 'mount' doesn't look in $PATH, just sbin
169 [ -f $LUSTRE/utils/mount.lustre ] && cp $LUSTRE/utils/mount.lustre /sbin/. || true
173 if [ `uname -r | cut -c 3` -eq 4 ]; then
181 MODULES=$($LCTL modules | awk '{ print $2 }')
182 while [ -n "$MODULES" ]; do
184 $RMMOD $MODULES > /dev/null 2>&1 || true
185 MODULES=$($LCTL modules | awk '{ print $2 }')
186 if [ -z "$MODULES" ]; then
190 echo "waiting, $((MAX - WAIT)) secs left"
192 if [ $WAIT -eq $MAX ]; then
193 echo "LNET modules $MODULES will not unload"
201 wait_exit_ST client # bug 12845
203 lsmod | grep lnet > /dev/null && $LCTL dl && $LCTL dk $TMP/debug
204 local MODULES=$($LCTL modules | awk '{ print $2 }')
205 $RMMOD $MODULES > /dev/null 2>&1 || true
206 # do it again, in case we tried to unload ksocklnd too early
207 MODULES=$($LCTL modules | awk '{ print $2 }')
208 [ -n "$MODULES" ] && $RMMOD $MODULES > /dev/null 2>&1 || true
209 MODULES=$($LCTL modules | awk '{ print $2 }')
210 if [ -n "$MODULES" ]; then
211 echo "Modules still loaded: "
213 if [ -e $LPROC ]; then
214 echo "Lustre still loaded"
215 cat $LPROC/devices || true
219 echo "Lustre stopped but LNET is still loaded, waiting..."
220 wait_for_lnet || return 3
225 LEAK_LUSTRE=$(dmesg | tail -n 30 | grep "obd mem.*leaked" || true)
226 LEAK_PORTALS=$(dmesg | tail -n 20 | grep "Portals memory leaked" || true)
227 if [ "$LEAK_LUSTRE" -o "$LEAK_PORTALS" ]; then
228 echo "$LEAK_LUSTRE" 1>&2
229 echo "$LEAK_PORTALS" 1>&2
230 mv $TMP/debug $TMP/debug-leak.`date +%s` || true
231 echo "Memory leaks detected"
232 [ -n "$IGNORE_LEAK" ] && echo "ignoring leaks" && return 0
235 echo "modules unloaded."
239 check_gss_daemon_facet() {
243 num=`do_facet $facet ps -o cmd -C $dname | grep $dname | wc -l`
244 if [ $num -ne 1 ]; then
245 echo "$num instance of $dname on $facet"
254 do_facet $facet "killall -2 $@ 2>/dev/null || true"
257 start_gss_daemons() {
259 for num in `seq $MDSCOUNT`; do
260 do_facet mds$num "$LSVCGSSD -v"
262 do_facet mds$num "$LGSSD -v"
266 for num in `seq $OSTCOUNT`; do
267 do_facet ost$num "$LSVCGSSD -v"
270 # FIXME: is "client" the right facet name?
272 do_facet client "$LGSSD -v"
275 # wait daemons entering "stable" status
279 # check daemons are running
281 for num in `seq $MDSCOUNT`; do
282 check_gss_daemon_facet mds$num lsvcgssd
284 check_gss_daemon_facet mds$num lgssd
287 for num in `seq $OSTCOUNT`; do
288 check_gss_daemon_facet ost$num lsvcgssd
291 check_gss_daemon_facet client lgssd
296 for num in `seq $MDSCOUNT`; do
297 send_sigint mds$num lsvcgssd lgssd
299 for num in `seq $OSTCOUNT`; do
300 send_sigint ost$num lsvcgssd
302 send_sigint client lgssd
306 if [ ! -z $SEC ]; then
307 MDS_MOUNT_OPTS=$MDS_MOUNT_OPTS,sec=$SEC
308 OST_MOUNT_OPTS=$OST_MOUNT_OPTS,sec=$SEC
319 # maybe cleanup credential cache?
325 local device=`mdsdevname $num`
326 local label=`do_facet mds$num "e2label ${device}" | grep -v "CMD: "`
332 local device=`ostdevname $num`
333 local label=`do_facet ost$num "e2label ${device}" | grep -v "CMD: "`
338 # start facet device options
344 echo "Starting ${facet}: $@ ${device} ${MOUNT%/*}/${facet}"
345 do_facet ${facet} mkdir -p ${MOUNT%/*}/${facet}
346 do_facet ${facet} mount -t lustre $@ ${device} ${MOUNT%/*}/${facet}
348 if [ $RC -ne 0 ]; then
349 echo mount -t lustre $@ ${device} ${MOUNT%/*}/${facet}
350 echo Start of ${device} on ${facet} failed ${RC}
352 do_facet ${facet} sync
353 label=$(do_facet ${facet} "e2label ${device}")
354 [ -z "$label" ] && echo no label for ${device} && exit 1
355 eval export ${facet}_svc=${label}
356 eval export ${facet}_dev=${device}
357 eval export ${facet}_opt=\"$@\"
358 echo Started ${label}
367 HOST=`facet_active_host $facet`
368 [ -z $HOST ] && echo stop: no host for $facet && return 0
370 running=$(do_facet ${facet} "grep -c ${MOUNT%/*}/${facet}' ' /proc/mounts") || true
371 if [ ${running} -ne 0 ]; then
372 echo "Stopping ${MOUNT%/*}/${facet} (opts:$@)"
373 do_facet ${facet} umount -d $@ ${MOUNT%/*}/${facet}
376 # umount should block, but we should wait for unrelated obd's
377 # like the MGS or MGC to also stop.
378 wait_exit_ST ${facet}
385 # Only supply -o to mount if we have options
386 if [ -n "$MOUNTOPT" ]; then
387 OPTIONS="-o $MOUNTOPT"
389 local device=$MGSNID:/$FSNAME
390 if [ -z "$mnt" -o -z "$FSNAME" ]; then
391 echo Bad zconf mount command: opt=$OPTIONS dev=$device mnt=$mnt
395 echo "Starting client: $OPTIONS $device $mnt"
396 do_node $client mkdir -p $mnt
397 do_node $client mount -t lustre $OPTIONS $device $mnt || return 1
399 do_node $client "sysctl -w lnet.debug=$PTLDEBUG; sysctl -w lnet.subsystem_debug=${SUBSYSTEM# }"
400 [ -d /r ] && $LCTL modules > /r/tmp/ogdb-`hostname`
408 local running=$(do_node $client "grep -c $mnt' ' /proc/mounts") || true
409 if [ $running -ne 0 ]; then
410 echo "Stopping client $mnt (opts:$force)"
411 do_node $client umount $force $mnt
417 if [ "$FAILURE_MODE" = HARD ]; then
418 $POWER_DOWN `facet_active_host $facet`
420 elif [ "$FAILURE_MODE" = SOFT ]; then
427 if [ "$FAILURE_MODE" = HARD ]; then
428 $POWER_UP `facet_active_host $facet`
434 # verify that lustre actually cleaned up properly
436 [ -f $CATASTROPHE ] && [ `cat $CATASTROPHE` -ne 0 ] && \
437 error "LBUG/LASSERT detected"
438 BUSY=`dmesg | grep -i destruct || true`
441 [ -e $TMP/debug ] && mv $TMP/debug $TMP/debug-busy.`date +%s`
444 LEAK_LUSTRE=`dmesg | tail -n 30 | grep "obd mem.*leaked" || true`
445 LEAK_PORTALS=`dmesg | tail -n 20 | grep "Portals memory leaked" || true`
446 if [ "$LEAK_LUSTRE" -o "$LEAK_PORTALS" ]; then
447 echo "$0: $LEAK_LUSTRE" 1>&2
448 echo "$0: $LEAK_PORTALS" 1>&2
449 echo "$0: Memory leak(s) detected..." 1>&2
450 mv $TMP/debug $TMP/debug-leak.`date +%s`
454 [ "`lctl dl 2> /dev/null | wc -l`" -gt 0 ] && lctl dl && \
455 echo "$0: lustre didn't clean up..." 1>&2 && return 202 || true
457 if [ "`/sbin/lsmod 2>&1 | egrep 'lnet|libcfs'`" ]; then
458 echo "$0: modules still loaded..." 1>&2
465 wait_delete_completed () {
466 local TOTALPREV=`awk 'BEGIN{total=0}; {total+=$1}; END{print total}' \
467 $LPROC/osc/*/kbytesavail`
471 while [ "$WAIT" -ne "$MAX_WAIT" ]; do
473 TOTAL=`awk 'BEGIN{total=0}; {total+=$1}; END{print total}' \
474 $LPROC/osc/*/kbytesavail`
475 [ "$TOTAL" -eq "$TOTALPREV" ] && break
476 echo "Waiting delete completed ... prev: $TOTALPREV current: $TOTAL "
480 echo "Delete completed."
485 check_network "$HOST" 900
486 while ! do_node $HOST "ls -d $LUSTRE " > /dev/null; do sleep 5; done
491 HOST=`facet_active_host $facet`
495 wait_mds_recovery_done () {
496 local timeout=`do_facet mds cat /proc/sys/lustre/timeout`
497 #define OBD_RECOVERY_TIMEOUT (obd_timeout * 5 / 2)
498 # as we are in process of changing obd_timeout in different ways
499 # let's set MAX longer than that
500 MAX=$(( timeout * 4 ))
502 while [ $WAIT -lt $MAX ]; do
503 STATUS=`do_facet mds grep status /proc/fs/lustre/mdt/*-MDT*/recovery_status`
504 echo $STATUS | grep COMPLETE && return 0
507 echo "Waiting $(($MAX - $WAIT)) secs for MDS recovery done"
509 echo "MDS recovery not done in $MAX sec"
518 # conf-sanity 31 takes a long time cleanup
519 while [ $WAIT -lt 300 ]; do
520 running=$(do_facet ${facet} "[ -e $LPROC ] && grep ST' ' $LPROC/devices") || true
521 [ -z "${running}" ] && return 0
522 echo "waited $WAIT for${running}"
523 [ $INTERVAL -lt 64 ] && INTERVAL=$((INTERVAL + INTERVAL))
525 WAIT=$((WAIT + INTERVAL))
527 echo "service didn't stop after $WAIT seconds. Still running:"
533 # not every config has many clients
534 if [ ! -z "$CLIENTS" ]; then
535 $PDSH $CLIENTS "df $MOUNT" > /dev/null
540 uname -n >> $MOUNT/recon
541 if [ ! -z "$CLIENTS" ]; then
542 $PDSH $CLIENTS "df $MOUNT; uname -n >> $MOUNT/recon" > /dev/null
544 echo Connected clients:
546 ls -l $MOUNT/recon > /dev/null
552 echo "Failing $facet on node `facet_active_host $facet`"
553 shutdown_facet $facet
557 echo "df pid is $DFPID"
559 TO=`facet_active_host $facet`
560 echo "Failover $facet to $TO"
562 local dev=${facet}_dev
563 local opt=${facet}_opt
564 start $facet ${!dev} ${!opt} || error "Restart of $facet failed"
575 local svc=${facet}_svc
576 do_facet $facet $LCTL --device %${!svc} readonly
577 do_facet $facet $LCTL --device %${!svc} notransno
578 do_facet $facet $LCTL mark "$facet REPLAY BARRIER on ${!svc}"
579 $LCTL mark "local REPLAY BARRIER on ${!svc}"
582 replay_barrier_nodf() {
583 local facet=$1 echo running=${running}
585 local svc=${facet}_svc
586 echo Replay barrier on ${!svc}
587 do_facet $facet $LCTL --device %${!svc} readonly
588 do_facet $facet $LCTL --device %${!svc} notransno
589 do_facet $facet $LCTL mark "$facet REPLAY BARRIER on ${!svc}"
590 $LCTL mark "local REPLAY BARRIER on ${!svc}"
594 UUID=`cat /proc/fs/lustre/mdc/${mds1_svc}-mdc-*/uuid`
595 do_facet mds1 "echo $UUID > /proc/fs/lustre/mdt/${mds1_svc}/evict_client"
599 UUID=`grep ${ost1_svc}-osc- $LPROC/devices | egrep -v 'MDT' | awk '{print $5}'`
600 do_facet ost1 "echo $UUID > /proc/fs/lustre/obdfilter/${ost1_svc}/evict_client"
604 facet_failover $* || error "failover: $?"
605 df $MOUNT || error "post-failover df: $?"
612 local svc=${facet}_svc
613 local dev=${facet}_dev
614 local opt=${facet}_opt
615 start $facet ${!dev} ${!opt}
616 do_facet $facet lctl --device %${!svc} abort_recovery
617 df $MOUNT || echo "first df failed: $?"
619 df $MOUNT || error "post-failover df: $?"
623 echo There is no lmc. This is mountconf, baby.
628 if [ "$1" = "client" -o "$1" = "'*'" ]; then echo \'*\'; else
629 ID=`$PDSH $1 $GMNALNID -l | cut -d\ -f2`
635 if [ "$1" = "client" -o "$1" = "'*'" ]; then echo \'*\'; else
641 if [ "$1" = "client" -o "$1" = "'*'" ]; then echo \'*\'; else
642 ID=`xtprocadmin -n $1 2>/dev/null | egrep -v 'NID' | awk '{print $1}'`
643 if [ -z "$ID" ]; then
644 echo "Could not get a ptl id for $1..."
653 h2name_or_ip "$1" "tcp"
658 if [ "$1" = "client" -o "$1" = "'*'" ]; then echo \'*\'; else
659 if type __h2elan >/dev/null 2>&1; then
662 ID=`echo $1 | sed 's/[^0-9]*//g'`
670 h2name_or_ip "$1" "openib"
675 h2name_or_ip "$1" "o2ib"
681 varname=${facet}_HOST
682 if [ -z "${!varname}" ]; then
683 if [ "${facet:0:3}" == "ost" ]; then
684 eval ${facet}_HOST=${ost_HOST}
692 local activevar=${facet}active
694 if [ -f ./${facet}active ] ; then
695 source ./${facet}active
699 if [ -z "$active" ] ; then
706 facet_active_host() {
708 local active=`facet_active $facet`
709 if [ "$facet" == client ]; then
712 echo `facet_host $active`
718 failover=${facet}failover
719 host=`facet_host $failover`
720 [ -z "$host" ] && return
721 curactive=`facet_active $facet`
722 if [ -z "${curactive}" -o "$curactive" == "$failover" ] ; then
723 eval export ${facet}active=$facet
725 eval export ${facet}active=$failover
727 # save the active host for this facet
728 activevar=${facet}active
729 echo "$activevar=${!activevar}" > ./$activevar
736 if [ "$HOST" = "$(hostname)" ]; then
740 echo "CMD: $HOST $@" >&2
741 $myPDSH $HOST $LCTL mark "$@" > /dev/null 2>&1 || :
743 $myPDSH $HOST "(PATH=\$PATH:$RLUSTRE/utils:$RLUSTRE/tests:/sbin:/usr/sbin; cd $RPWD; sh -c \"$@\")" | sed "s/^${HOST}: //"
744 return ${PIPESTATUS[0]}
750 HOST=`facet_active_host $facet`
751 [ -z $HOST ] && echo No host defined for facet ${facet} && exit 1
758 # make sure its not already running
761 do_facet ${facet} $MKFS $*
767 #if $OSTDEVn isn't defined, default is $OSTDEVBASE + num
768 eval DEVPTR=${!DEVNAME:=${OSTDEVBASE}${num}}
775 #if $MDSDEVn isn't defined, default is $MDSDEVBASE + num
776 eval DEVPTR=${!DEVNAME:=${MDSDEVBASE}${num}}
784 # make sure we are using the primary server, so test-framework will
785 # be able to clean up properly.
786 activemds=`facet_active mds1`
787 if [ $activemds != "mds1" ]; then
791 # assume client mount is local
792 grep " $MOUNT " /proc/mounts && zconf_umount `hostname` $MOUNT $*
793 grep " $MOUNT2 " /proc/mounts && zconf_umount `hostname` $MOUNT2 $*
794 for num in `seq $MDSCOUNT`; do
797 for num in `seq $OSTCOUNT`; do
812 test $nr = 1 && echo -n $MDS_MKFS_OPTS || echo -n $MDSn_MKFS_OPTS
816 [ "$FSTYPE" ] && FSTYPE_OPT="--backfstype $FSTYPE"
819 # We need ldiskfs here, may as well load them all
821 [ "$CLIENTONLY" ] && return
822 echo "Formatting mdts, osts"
823 for num in `seq $MDSCOUNT`; do
824 echo "Format mds$num: $(mdsdevname $num)"
826 add mds$num `mdsmkfsopts $num` $FSTYPE_OPT --reformat `mdsdevname $num` || exit 9
828 add mds$num `mdsmkfsopts $num` $FSTYPE_OPT --reformat `mdsdevname $num` > /dev/null || exit 9
832 for num in `seq $OSTCOUNT`; do
833 echo "Format ost$num: $(ostdevname $num)"
835 add ost$num $OST_MKFS_OPTS --reformat `ostdevname $num` || exit 10
837 add ost$num $OST_MKFS_OPTS --reformat `ostdevname $num` > /dev/null || exit 10
843 grep " $1 " /proc/mounts || zconf_mount `hostname` $*
849 if [ -z "$CLIENTONLY" ]; then
850 echo "Setup mdts, osts"
851 for num in `seq $MDSCOUNT`; do
852 DEVNAME=$(mdsdevname $num)
853 echo $REFORMAT | grep -q "reformat" \
854 || do_facet mds$num "$TUNEFS --writeconf $DEVNAME"
855 start mds$num $DEVNAME $MDS_MOUNT_OPTS
857 for num in `seq $OSTCOUNT`; do
858 DEVNAME=$(ostdevname $num)
859 start ost$num $DEVNAME $OST_MOUNT_OPTS
862 [ "$DAEMONFILE" ] && $LCTL debug_daemon start $DAEMONFILE $DAEMONSIZE
864 if [ "$MOUNT_2" ]; then
870 mounted_lustre_filesystems() {
871 awk '($3 ~ "lustre" && $1 ~ ":") { print $2 }' /proc/mounts
874 check_and_setup_lustre() {
875 MOUNTED="`mounted_lustre_filesystems`"
876 if [ -z "$MOUNTED" ]; then
877 [ "$REFORMAT" ] && formatall
879 MOUNTED="`mounted_lustre_filesystems`"
880 [ -z "$MOUNTED" ] && error "NAME=$NAME not mounted"
883 if [ "$ONLY" == "setup" ]; then
888 cleanup_and_setup_lustre() {
889 if [ "$ONLY" == "cleanup" -o "`mount | grep $MOUNT`" ]; then
890 sysctl -w lnet.debug=0 || true
892 if [ "$ONLY" == "cleanup" ]; then
896 check_and_setup_lustre
899 check_and_cleanup_lustre() {
900 if [ "`mount | grep $MOUNT`" ]; then
901 rm -rf $DIR/[Rdfs][1-9]*
902 rm -f $DIR/${TESTSUITE}/[Rdfs][1-9]*
904 if [ "$I_MOUNTED" = "yes" ]; then
905 cleanupall -f || error "cleanup failed"
917 while [ $NETWORK -eq 0 ]; do
918 ping -c 1 -w 3 $1 > /dev/null
919 if [ $? -eq 0 ]; then
923 echo "waiting for $1, $((MAX - WAIT)) secs left"
926 if [ $WAIT -gt $MAX ]; then
927 echo "Network not available"
933 while( !($DSH2 $1 "netstat -tna | grep -q $2") ) ; do
944 # the sed converts spaces to commas, but leaves the last space
945 # alone, so the line doesn't end with a comma.
946 echo "$*" | tr -s " " "\n" | sort -b -u | tr "\n" " " | sed 's/ \([^$]\)/,\1/g'
950 (cd `dirname $1`; echo $PWD/`basename $1`)
953 ##################################
957 # OBD_FAIL_MDS_ALL_REQUEST_NET
959 do_facet mds sysctl -w lustre.fail_loc=0x123
960 do_facet client "$1" || RC=$?
961 do_facet mds sysctl -w lustre.fail_loc=0
966 # OBD_FAIL_MDS_ALL_REPLY_NET
968 do_facet mds sysctl -w lustre.fail_loc=0x122
969 do_facet client "$@" || RC=$?
970 do_facet mds sysctl -w lustre.fail_loc=0
975 # OBD_FAIL_MDS_REINT_NET_REP
977 do_facet mds sysctl -w lustre.fail_loc=0x119
978 do_facet client "$@" || RC=$?
979 do_facet mds sysctl -w lustre.fail_loc=0
984 #define OBD_FAIL_OST_BRW_PAUSE_BULK 0x214
986 do_facet ost1 sysctl -w lustre.fail_loc=0x214
987 do_facet client "$1" || RC=$?
988 do_facet client "sync"
989 do_facet ost1 sysctl -w lustre.fail_loc=0
994 #define OBD_FAIL_LDLM_CANCEL 0x304
996 do_facet client sysctl -w lustre.fail_loc=0x304
997 do_facet client "$@" || RC=$?
998 do_facet client sysctl -w lustre.fail_loc=0
1002 drop_bl_callback() {
1003 #define OBD_FAIL_LDLM_BL_CALLBACK 0x305
1005 do_facet client sysctl -w lustre.fail_loc=0x305
1006 do_facet client "$@" || RC=$?
1007 do_facet client sysctl -w lustre.fail_loc=0
1012 #define OBD_FAIL_LDLM_REPLY 0x30c
1014 do_facet mds sysctl -w lustre.fail_loc=0x30c
1015 do_facet client "$@" || RC=$?
1016 do_facet mds sysctl -w lustre.fail_loc=0
1024 echo "clearing fail_loc on $facet"
1025 do_facet $facet "sysctl -w lustre.fail_loc=0"
1028 cancel_lru_locks() {
1029 $LCTL mark "cancel_lru_locks $1 start"
1030 for d in `find $LPROC/ldlm/namespaces | egrep -i $1`; do
1031 [ -f $d/lru_size ] && echo clear > $d/lru_size
1032 [ -f $d/lock_unused_count ] && grep [1-9] $d/lock_unused_count /dev/null
1034 $LCTL mark "cancel_lru_locks $1 stop"
1039 for a in /proc/fs/lustre/llite/*/dump_page_cache; do
1040 if [ `wc -l $a | awk '{print $1}'` -gt 1 ]; then
1041 echo there is still data in page cache $a ?
1050 DEBUGSAVE="$(sysctl -n lnet.debug)"
1054 [ -n "$DEBUGSAVE" ] && sysctl -w lnet.debug="${DEBUGSAVE}"
1058 ##################################
1060 ##################################
1063 local FAIL_ON_ERROR=${FAIL_ON_ERROR:-true}
1064 local TYPE=${TYPE:-"FAIL"}
1066 sysctl -w lustre.fail_loc=0 2> /dev/null || true
1067 log "${TESTSUITE} ${TESTNAME}: **** ${TYPE}:" $@
1068 ERRLOG=$TMP/lustre_${TESTSUITE}_${TESTNAME}.$(date +%s)
1069 echo "Dumping lctl log to $ERRLOG"
1070 # We need to dump the logs on all nodes
1072 [ ! "$mds_HOST" = "$(hostname)" ] && do_node $mds_HOST $LCTL dk $ERRLOG
1073 [ ! "$ost_HOST" = "$(hostname)" -a ! "$ost_HOST" = "$mds_HOST" ] && do_node $ost_HOST $LCTL dk $ERRLOG
1075 [ "$TESTSUITELOG" ] && echo "$0: ${TYPE}: $TESTNAME $@" >> $TESTSUITELOG
1076 if $FAIL_ON_ERROR; then
1081 # use only if we are ignoring failures for this test, bugno required.
1082 # (like ALWAYS_EXCEPT, but run the test and ignore the results.)
1083 # e.g. error_ignore 5494 "your message"
1085 FAIL_ON_ERROR=false TYPE="IGNORE (bz$1)" error $2
1089 log " SKIP: ${TESTSUITE} ${TESTNAME} $@"
1090 [ "$TESTSUITELOG" ] && echo "${TESTSUITE}: SKIP: $TESTNAME $@" >> $TESTSUITELOG
1093 build_test_filter() {
1094 [ "$ONLY" ] && log "only running test `echo $ONLY`"
1098 [ "$EXCEPT$ALWAYS_EXCEPT" ] && \
1099 log "skipping tests: `echo $EXCEPT $ALWAYS_EXCEPT`"
1100 for E in $EXCEPT $ALWAYS_EXCEPT; do
1101 eval EXCEPT_${E}=true
1103 for G in $GRANT_CHECK_LIST; do
1104 eval GCHECK_ONLY_${G}=true
1113 IFS=abcdefghijklmnopqrstuvwxyz _basetest $1
1117 export base=`basetest $1`
1118 if [ ! -z "$ONLY" ]; then
1120 if [ ${!testname}x != x ]; then
1125 if [ ${!testname}x != x ]; then
1133 if [ ${!testname}x != x ]; then
1134 TESTNAME=test_$1 skip "skipping excluded test $1"
1137 testname=EXCEPT_$base
1138 if [ ${!testname}x != x ]; then
1139 TESTNAME=test_$1 skip "skipping excluded test $1 (base $base)"
1147 EQUALS="======================================================================"
1151 local suffixlen=$((${#EQUALS} - ${#msg}))
1152 [ $suffixlen -lt 5 ] && suffixlen=5
1153 log `echo $(printf '===== %s %.*s\n' "$msg" $suffixlen $EQUALS)`
1158 lsmod | grep lnet > /dev/null || load_modules
1159 $LCTL mark "$*" 2> /dev/null || true
1164 strace -o $TMP/$1.strace -ttt $*
1166 log "FINISHED: $*: rc $RC"
1175 FFREE=`cat /proc/fs/lustre/mds/*/filesfree`
1176 FTOTAL=`cat /proc/fs/lustre/mds/*/filestotal`
1177 [ $FFREE -ge $FTOTAL ] && error "files free $FFREE > total $FTOTAL" || true
1184 export tdir=d${TESTSUITE}/d${base}
1185 local SAVE_UMASK=`umask`
1190 log "== test $testnum: $message ============ `date +%H:%M:%S` ($BEFORE)"
1192 export TESTNAME=test_$testnum
1193 test_${testnum} || error "test_$testnum failed with $?"
1195 check_grant ${testnum} || error "check_grant $testnum failed with $?"
1196 [ -f $CATASTROPHE ] && [ `cat $CATASTROPHE` -ne 0 ] && \
1197 error "LBUG/LASSERT detected"
1198 pass "($((`date +%s` - $BEFORE))s)"
1199 rmdir ${DIR}/$tdir >/dev/null 2>&1 || true
1208 (cd `dirname $1`; echo $PWD/`basename $1`)
1212 [ -d $DIR1 ] && cd $DIR1 && sync; sleep 1; sync
1213 [ -d $DIR2 ] && cd $DIR2 && sync; sleep 1; sync
1218 export base=`basetest $1`
1219 [ "$CHECK_GRANT" == "no" ] && return 0
1221 testname=GCHECK_ONLY_${base}
1222 [ ${!testname}x == x ] && return 0
1224 echo -n "checking grant......"
1226 # write some data to sync client lost_grant
1227 rm -f $DIR1/${tfile}_check_grant_* 2>&1
1228 for i in `seq $OSTCOUNT`; do
1229 $LFS setstripe $DIR1/${tfile}_check_grant_$i 0 $(($i -1)) 1
1230 dd if=/dev/zero of=$DIR1/${tfile}_check_grant_$i bs=4k \
1231 count=1 > /dev/null 2>&1
1233 # sync all the data and make sure no pending data on server
1236 #get client grant and server grant
1238 for d in ${LPROC}/osc/*/cur_grant_bytes; do
1239 client_grant=$((client_grant + `cat $d`))
1242 for d in ${LPROC}/obdfilter/*/tot_granted; do
1243 server_grant=$((server_grant + `cat $d`))
1246 # cleanup the check_grant file
1247 for i in `seq $OSTCOUNT`; do
1248 rm $DIR1/${tfile}_check_grant_$i
1251 #check whether client grant == server grant
1252 if [ $client_grant != $server_grant ]; then
1253 echo "failed: client:${client_grant} server: ${server_grant}"
1260 ########################
1266 ost=`echo $1 | awk -F_ '{print $3}'`
1267 if [ -z $ost ]; then
1268 ost=`echo $1 | sed 's/-osc.*//'`
1275 [ ! -e /proc/fs/lustre/mdt/*MDT* ]
1280 [ $(grep -c obdfilter $LPROC/devices) -eq 0 ]
1285 grep -q patchless $LPROC/version