2 # vim:expandtab:shiftwidth=4:softtabstop=4:tabstop=4:
4 trap 'print_summary && 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: $0 [-r] [-f cfgfile]"
38 [ -n "$ONLY" ] && echo "WARNING: ONLY is set to ${ONLY}."
39 local form="%-13s %-17s %s\n"
40 printf "$form" "status" "script" "skipped tests E(xcluded) S(low)"
41 echo "------------------------------------------------------------------------------------"
42 for O in $TESTSUITE_LIST; do
45 local o=$(echo $O | tr "[:upper:]" "[:lower:]")
48 local log=${TMP}/${o}.log
49 [ -f $log ] && skipped=$(grep excluded $log | awk '{ printf " %s", $3 }' | sed 's/test_//g')
50 [ -f $log ] && slow=$(grep SLOW $log | awk '{ printf " %s", $3 }' | sed 's/test_//g')
51 [ "${!O}" = "done" ] && \
52 printf "$form" "Done" "$O" "E=$skipped" && \
53 [ -n "$slow" ] && printf "$form" "-" "-" "S=$slow"
57 for O in $TESTSUITE_LIST; do
58 [ "${!O}" = "no" ] && \
59 printf "$form" "Skipped" "$O" ""
62 for O in $TESTSUITE_LIST; do
63 [ "${!O}" = "done" -o "${!O}" = "no" ] || \
64 printf "$form" "UNFINISHED" "$O" ""
69 export LUSTRE=`absolute_path $LUSTRE`
70 export TESTSUITE=`basename $0 .sh`
71 export LTESTDIR=${LTESTDIR:-$LUSTRE/../ltest}
73 [ -d /r ] && export ROOT=${ROOT:-/r}
74 export TMP=${TMP:-$ROOT/tmp}
75 export TESTSUITELOG=${TMP}/${TESTSUITE}.log
77 export PATH=:$PATH:$LUSTRE/utils:$LUSTRE/utils/gss:$LUSTRE/tests
78 export LCTL=${LCTL:-"$LUSTRE/utils/lctl"}
79 export LFS=${LFS:-"$LUSTRE/utils/lfs"}
80 [ ! -f "$LCTL" ] && export LCTL=$(which lctl)
81 export LFS=${LFS:-"$LUSTRE/utils/lfs"}
82 [ ! -f "$LFS" ] && export LFS=$(which lfs)
83 export MKFS=${MKFS:-"$LUSTRE/utils/mkfs.lustre"}
84 [ ! -f "$MKFS" ] && export MKFS=$(which mkfs.lustre)
85 export TUNEFS=${TUNEFS:-"$LUSTRE/utils/tunefs.lustre"}
86 [ ! -f "$TUNEFS" ] && export TUNEFS=$(which tunefs.lustre)
87 export CHECKSTAT="${CHECKSTAT:-"checkstat -v"} "
88 export FSYTPE=${FSTYPE:-"ldiskfs"}
89 export NAME=${NAME:-local}
90 export LPROC=/proc/fs/lustre
91 export LGSSD=${LGSSD:-"$LUSTRE/utils/gss/lgssd"}
92 export LSVCGSSD=${LSVCGSSD:-"$LUSTRE/utils/gss/lsvcgssd"}
93 export KRB5DIR=${KRB5DIR:-"/usr/kerberos"}
96 if [ "$ACCEPTOR_PORT" ]; then
97 export PORT_OPT="--port $ACCEPTOR_PORT"
102 echo "Using GSS/krb5 ptlrpc security flavor"
108 # Paths on remote nodes, if different
109 export RLUSTRE=${RLUSTRE:-$LUSTRE}
110 export RPWD=${RPWD:-$PWD}
111 export I_MOUNTED=${I_MOUNTED:-"no"}
115 while getopts "rvf:" opt $*; do
118 r) REFORMAT=--reformat;;
124 shift $((OPTIND - 1))
127 [ "$TESTSUITELOG" ] && rm -f $TESTSUITELOG || true
135 BASE=`basename $module $EXT`
136 lsmod | grep -q ${BASE} || \
137 if [ -f ${LUSTRE}/${module}${EXT} ]; then
138 insmod ${LUSTRE}/${module}${EXT} $@
140 # must be testing a "make install" or "rpm" installation
141 # note failed to load ptlrpc_gss is considered not fatal
142 if [ "$BASE" == "ptlrpc_gss" ]; then
143 modprobe $BASE $@ 2>/dev/null || echo "gss/krb5 is not supported"
151 if [ -n "$MODPROBE" ]; then
155 if [ "$HAVE_MODULES" = true ]; then
161 echo Loading modules from $LUSTRE
162 load_module ../lnet/libcfs/libcfs
163 [ -f /etc/modprobe.conf ] && MODPROBECONF=/etc/modprobe.conf
164 [ -f /etc/modprobe.d/Lustre ] && MODPROBECONF=/etc/modprobe.d/Lustre
165 [ -z "$LNETOPTS" -a -n "$MODPROBECONF" ] && \
166 LNETOPTS=$(awk '/^options lnet/ { print $0}' $MODPROBECONF | sed 's/^options lnet //g')
167 echo "lnet options: '$LNETOPTS'"
168 # note that insmod will ignore anything in modprobe.conf
169 load_module ../lnet/lnet/lnet $LNETOPTS
170 LNETLND=${LNETLND:-"socklnd/ksocklnd"}
171 load_module ../lnet/klnds/$LNETLND
172 load_module lvfs/lvfs
173 load_module obdclass/obdclass
174 load_module ptlrpc/ptlrpc
175 load_module ptlrpc/gss/ptlrpc_gss
176 # Now, some modules depend on lquota without USE_QUOTA check,
177 # will fix later. Disable check "$USE_QUOTA" = "yes" temporary.
178 #[ "$USE_QUOTA" = "yes" ] && load_module quota/lquota
179 load_module quota/lquota
187 if [ -z "$CLIENTONLY" ]; then
188 [ "$FSTYPE" = "ldiskfs" ] && load_module ../ldiskfs/ldiskfs/ldiskfs
193 load_module lvfs/fsfilt_$FSTYPE
197 load_module obdfilter/obdfilter
200 load_module llite/lustre
201 load_module llite/llite_lloop
202 rm -f $TMP/ogdb-`hostname`
203 $LCTL modules > $TMP/ogdb-`hostname`
204 # 'mount' doesn't look in $PATH, just sbin
205 [ -f $LUSTRE/utils/mount.lustre ] && cp $LUSTRE/utils/mount.lustre /sbin/. || true
209 if [ `uname -r | cut -c 3` -eq 4 ]; then
217 MODULES=$($LCTL modules | awk '{ print $2 }')
218 while [ -n "$MODULES" ]; do
220 $RMMOD $MODULES > /dev/null 2>&1 || true
221 MODULES=$($LCTL modules | awk '{ print $2 }')
222 if [ -z "$MODULES" ]; then
226 echo "waiting, $((MAX - WAIT)) secs left"
228 if [ $WAIT -eq $MAX ]; then
229 echo "LNET modules $MODULES will not unload"
237 wait_exit_ST client # bug 12845
239 lsmod | grep lnet > /dev/null && $LCTL dl && $LCTL dk $TMP/debug
240 local MODULES=$($LCTL modules | awk '{ print $2 }')
241 $RMMOD $MODULES > /dev/null 2>&1 || true
242 # do it again, in case we tried to unload ksocklnd too early
243 MODULES=$($LCTL modules | awk '{ print $2 }')
244 [ -n "$MODULES" ] && $RMMOD $MODULES > /dev/null 2>&1 || true
245 MODULES=$($LCTL modules | awk '{ print $2 }')
246 if [ -n "$MODULES" ]; then
247 echo "Modules still loaded: "
249 if [ -e $LPROC ]; then
250 echo "Lustre still loaded"
251 cat $LPROC/devices || true
255 echo "Lustre stopped but LNET is still loaded, waiting..."
256 wait_for_lnet || return 3
261 LEAK_LUSTRE=$(dmesg | tail -n 30 | grep "obd mem.*leaked" || true)
262 LEAK_PORTALS=$(dmesg | tail -n 20 | grep "Portals memory leaked" || true)
263 if [ "$LEAK_LUSTRE" -o "$LEAK_PORTALS" ]; then
264 echo "$LEAK_LUSTRE" 1>&2
265 echo "$LEAK_PORTALS" 1>&2
266 mv $TMP/debug $TMP/debug-leak.`date +%s` || true
267 echo "Memory leaks detected"
268 [ -n "$IGNORE_LEAK" ] && echo "ignoring leaks" && return 0
271 echo "modules unloaded."
275 check_gss_daemon_facet() {
279 num=`do_facet $facet ps -o cmd -C $dname | grep $dname | wc -l`
280 if [ $num -ne 1 ]; then
281 echo "$num instance of $dname on $facet"
290 do_facet $facet "killall -2 $@ 2>/dev/null || true"
293 start_gss_daemons() {
295 for num in `seq $MDSCOUNT`; do
296 do_facet mds$num "$LSVCGSSD -v"
298 do_facet mds$num "$LGSSD -v"
302 for num in `seq $OSTCOUNT`; do
303 do_facet ost$num "$LSVCGSSD -v"
306 # FIXME: is "client" the right facet name?
308 do_facet client "$LGSSD -v"
311 # wait daemons entering "stable" status
315 # check daemons are running
317 for num in `seq $MDSCOUNT`; do
318 check_gss_daemon_facet mds$num lsvcgssd
320 check_gss_daemon_facet mds$num lgssd
323 for num in `seq $OSTCOUNT`; do
324 check_gss_daemon_facet ost$num lsvcgssd
327 check_gss_daemon_facet client lgssd
332 for num in `seq $MDSCOUNT`; do
333 send_sigint mds$num lsvcgssd lgssd
335 for num in `seq $OSTCOUNT`; do
336 send_sigint ost$num lsvcgssd
338 send_sigint client lgssd
342 if [ ! -z $SEC ]; then
343 MDS_MOUNT_OPTS=$MDS_MOUNT_OPTS,sec=$SEC
344 OST_MOUNT_OPTS=$OST_MOUNT_OPTS,sec=$SEC
355 # maybe cleanup credential cache?
361 local device=`mdsdevname $num`
362 local label=`do_facet mds$num "e2label ${device}" | grep -v "CMD: "`
368 local device=`ostdevname $num`
369 local label=`do_facet ost$num "e2label ${device}" | grep -v "CMD: "`
374 # start facet device options
380 echo "Starting ${facet}: $@ ${device} ${MOUNT%/*}/${facet}"
381 do_facet ${facet} mkdir -p ${MOUNT%/*}/${facet}
382 do_facet ${facet} mount -t lustre $@ ${device} ${MOUNT%/*}/${facet}
384 if [ $RC -ne 0 ]; then
385 echo mount -t lustre $@ ${device} ${MOUNT%/*}/${facet}
386 echo Start of ${device} on ${facet} failed ${RC}
388 do_facet ${facet} "sysctl -w lnet.debug=$PTLDEBUG; \
389 sysctl -w lnet.subsystem_debug=${SUBSYSTEM# }; \
390 sysctl -w lnet.debug_mb=${DEBUG_SIZE}"
392 do_facet ${facet} sync
393 label=$(do_facet ${facet} "e2label ${device}")
394 [ -z "$label" ] && echo no label for ${device} && exit 1
395 eval export ${facet}_svc=${label}
396 eval export ${facet}_dev=${device}
397 eval export ${facet}_opt=\"$@\"
398 echo Started ${label}
407 HOST=`facet_active_host $facet`
408 [ -z $HOST ] && echo stop: no host for $facet && return 0
410 running=$(do_facet ${facet} "grep -c ${MOUNT%/*}/${facet}' ' /proc/mounts") || true
411 if [ ${running} -ne 0 ]; then
412 echo "Stopping ${MOUNT%/*}/${facet} (opts:$@)"
413 do_facet ${facet} umount -d $@ ${MOUNT%/*}/${facet}
416 # umount should block, but we should wait for unrelated obd's
417 # like the MGS or MGC to also stop.
418 wait_exit_ST ${facet}
425 # Only supply -o to mount if we have options
426 if [ -n "$MOUNTOPT" ]; then
427 OPTIONS="-o $MOUNTOPT"
429 local device=$MGSNID:/$FSNAME
430 if [ -z "$mnt" -o -z "$FSNAME" ]; then
431 echo Bad zconf mount command: opt=$OPTIONS dev=$device mnt=$mnt
435 echo "Starting client: $OPTIONS $device $mnt"
436 do_node $client mkdir -p $mnt
437 do_node $client mount -t lustre $OPTIONS $device $mnt || return 1
439 do_node $client "sysctl -w lnet.debug=$PTLDEBUG;
440 sysctl -w lnet.subsystem_debug=${SUBSYSTEM# };
441 sysctl -w lnet.debug_mb=${DEBUG_SIZE}"
442 [ -d /r ] && $LCTL modules > /r/tmp/ogdb-`hostname`
450 local running=$(do_node $client "grep -c $mnt' ' /proc/mounts") || true
451 if [ $running -ne 0 ]; then
452 echo "Stopping client $mnt (opts:$force)"
453 do_node $client umount $force $mnt
459 if [ "$FAILURE_MODE" = HARD ]; then
460 $POWER_DOWN `facet_active_host $facet`
462 elif [ "$FAILURE_MODE" = SOFT ]; then
469 if [ "$FAILURE_MODE" = HARD ]; then
470 $POWER_UP `facet_active_host $facet`
476 # verify that lustre actually cleaned up properly
478 [ -f $CATASTROPHE ] && [ `cat $CATASTROPHE` -ne 0 ] && \
479 error "LBUG/LASSERT detected"
480 BUSY=`dmesg | grep -i destruct || true`
483 [ -e $TMP/debug ] && mv $TMP/debug $TMP/debug-busy.`date +%s`
486 LEAK_LUSTRE=`dmesg | tail -n 30 | grep "obd mem.*leaked" || true`
487 LEAK_PORTALS=`dmesg | tail -n 20 | grep "Portals memory leaked" || true`
488 if [ "$LEAK_LUSTRE" -o "$LEAK_PORTALS" ]; then
489 echo "$0: $LEAK_LUSTRE" 1>&2
490 echo "$0: $LEAK_PORTALS" 1>&2
491 echo "$0: Memory leak(s) detected..." 1>&2
492 mv $TMP/debug $TMP/debug-leak.`date +%s`
496 [ "`lctl dl 2> /dev/null | wc -l`" -gt 0 ] && lctl dl && \
497 echo "$0: lustre didn't clean up..." 1>&2 && return 202 || true
499 if [ "`/sbin/lsmod 2>&1 | egrep 'lnet|libcfs'`" ]; then
500 echo "$0: modules still loaded..." 1>&2
507 wait_delete_completed () {
508 local TOTALPREV=`awk 'BEGIN{total=0}; {total+=$1}; END{print total}' \
509 $LPROC/osc/*/kbytesavail`
513 while [ "$WAIT" -ne "$MAX_WAIT" ]; do
515 TOTAL=`awk 'BEGIN{total=0}; {total+=$1}; END{print total}' \
516 $LPROC/osc/*/kbytesavail`
517 [ "$TOTAL" -eq "$TOTALPREV" ] && break
518 echo "Waiting delete completed ... prev: $TOTALPREV current: $TOTAL "
522 echo "Delete completed."
527 check_network "$HOST" 900
528 while ! do_node $HOST "ls -d $LUSTRE " > /dev/null; do sleep 5; done
533 HOST=`facet_active_host $facet`
537 wait_mds_recovery_done () {
538 local timeout=`do_facet mds cat /proc/sys/lustre/timeout`
539 #define OBD_RECOVERY_TIMEOUT (obd_timeout * 5 / 2)
540 # as we are in process of changing obd_timeout in different ways
541 # let's set MAX longer than that
542 MAX=$(( timeout * 4 ))
544 while [ $WAIT -lt $MAX ]; do
545 STATUS=`do_facet mds grep status /proc/fs/lustre/mdt/*-MDT*/recovery_status`
546 echo $STATUS | grep COMPLETE && return 0
549 echo "Waiting $(($MAX - $WAIT)) secs for MDS recovery done"
551 echo "MDS recovery not done in $MAX sec"
560 # conf-sanity 31 takes a long time cleanup
561 while [ $WAIT -lt 300 ]; do
562 running=$(do_facet ${facet} "[ -e $LPROC ] && grep ST' ' $LPROC/devices") || true
563 [ -z "${running}" ] && return 0
564 echo "waited $WAIT for${running}"
565 [ $INTERVAL -lt 64 ] && INTERVAL=$((INTERVAL + INTERVAL))
567 WAIT=$((WAIT + INTERVAL))
569 echo "service didn't stop after $WAIT seconds. Still running:"
575 # not every config has many clients
576 if [ ! -z "$CLIENTS" ]; then
577 $PDSH $CLIENTS "df $MOUNT" > /dev/null
582 uname -n >> $MOUNT/recon
583 if [ ! -z "$CLIENTS" ]; then
584 $PDSH $CLIENTS "df $MOUNT; uname -n >> $MOUNT/recon" > /dev/null
586 echo Connected clients:
588 ls -l $MOUNT/recon > /dev/null
594 echo "Failing $facet on node `facet_active_host $facet`"
595 shutdown_facet $facet
599 echo "df pid is $DFPID"
601 TO=`facet_active_host $facet`
602 echo "Failover $facet to $TO"
604 local dev=${facet}_dev
605 local opt=${facet}_opt
606 start $facet ${!dev} ${!opt} || error "Restart of $facet failed"
617 local svc=${facet}_svc
618 do_facet $facet $LCTL --device %${!svc} readonly
619 do_facet $facet $LCTL --device %${!svc} notransno
620 do_facet $facet $LCTL mark "$facet REPLAY BARRIER on ${!svc}"
621 $LCTL mark "local REPLAY BARRIER on ${!svc}"
624 replay_barrier_nodf() {
625 local facet=$1 echo running=${running}
627 local svc=${facet}_svc
628 echo Replay barrier on ${!svc}
629 do_facet $facet $LCTL --device %${!svc} readonly
630 do_facet $facet $LCTL --device %${!svc} notransno
631 do_facet $facet $LCTL mark "$facet REPLAY BARRIER on ${!svc}"
632 $LCTL mark "local REPLAY BARRIER on ${!svc}"
636 UUID=`cat /proc/fs/lustre/mdc/${mds1_svc}-mdc-*/uuid`
637 do_facet mds1 "echo $UUID > /proc/fs/lustre/mdt/${mds1_svc}/evict_client"
641 UUID=`grep ${ost1_svc}-osc- $LPROC/devices | egrep -v 'MDT' | awk '{print $5}'`
642 do_facet ost1 "echo $UUID > /proc/fs/lustre/obdfilter/${ost1_svc}/evict_client"
646 facet_failover $* || error "failover: $?"
647 df $MOUNT || error "post-failover df: $?"
654 local svc=${facet}_svc
655 local dev=${facet}_dev
656 local opt=${facet}_opt
657 start $facet ${!dev} ${!opt}
658 do_facet $facet lctl --device %${!svc} abort_recovery
659 df $MOUNT || echo "first df failed: $?"
661 df $MOUNT || error "post-failover df: $?"
665 echo There is no lmc. This is mountconf, baby.
670 if [ "$1" = "client" -o "$1" = "'*'" ]; then echo \'*\'; else
671 ID=`$PDSH $1 $GMNALNID -l | cut -d\ -f2`
677 if [ "$1" = "client" -o "$1" = "'*'" ]; then echo \'*\'; else
683 if [ "$1" = "client" -o "$1" = "'*'" ]; then echo \'*\'; else
684 ID=`xtprocadmin -n $1 2>/dev/null | egrep -v 'NID' | awk '{print $1}'`
685 if [ -z "$ID" ]; then
686 echo "Could not get a ptl id for $1..."
695 h2name_or_ip "$1" "tcp"
700 if [ "$1" = "client" -o "$1" = "'*'" ]; then echo \'*\'; else
701 if type __h2elan >/dev/null 2>&1; then
704 ID=`echo $1 | sed 's/[^0-9]*//g'`
712 h2name_or_ip "$1" "openib"
717 h2name_or_ip "$1" "o2ib"
723 varname=${facet}_HOST
724 if [ -z "${!varname}" ]; then
725 if [ "${facet:0:3}" == "ost" ]; then
726 eval ${facet}_HOST=${ost_HOST}
734 local activevar=${facet}active
736 if [ -f ./${facet}active ] ; then
737 source ./${facet}active
741 if [ -z "$active" ] ; then
748 facet_active_host() {
750 local active=`facet_active $facet`
751 if [ "$facet" == client ]; then
754 echo `facet_host $active`
760 failover=${facet}failover
761 host=`facet_host $failover`
762 [ -z "$host" ] && return
763 curactive=`facet_active $facet`
764 if [ -z "${curactive}" -o "$curactive" == "$failover" ] ; then
765 eval export ${facet}active=$facet
767 eval export ${facet}active=$failover
769 # save the active host for this facet
770 activevar=${facet}active
771 echo "$activevar=${!activevar}" > ./$activevar
778 if [ "$HOST" = "$(hostname)" ]; then
782 echo "CMD: $HOST $@" >&2
783 $myPDSH $HOST $LCTL mark "$@" > /dev/null 2>&1 || :
785 $myPDSH $HOST "(PATH=\$PATH:$RLUSTRE/utils:$RLUSTRE/tests:/sbin:/usr/sbin; cd $RPWD; sh -c \"$@\")" | sed "s/^${HOST}: //"
786 return ${PIPESTATUS[0]}
792 HOST=`facet_active_host $facet`
793 [ -z $HOST ] && echo No host defined for facet ${facet} && exit 1
800 # make sure its not already running
803 do_facet ${facet} $MKFS $*
809 #if $OSTDEVn isn't defined, default is $OSTDEVBASE + num
810 eval DEVPTR=${!DEVNAME:=${OSTDEVBASE}${num}}
817 #if $MDSDEVn isn't defined, default is $MDSDEVBASE + num
818 eval DEVPTR=${!DEVNAME:=${MDSDEVBASE}${num}}
826 # make sure we are using the primary server, so test-framework will
827 # be able to clean up properly.
828 activemds=`facet_active mds1`
829 if [ $activemds != "mds1" ]; then
833 # assume client mount is local
834 grep " $MOUNT " /proc/mounts && zconf_umount `hostname` $MOUNT $*
835 grep " $MOUNT2 " /proc/mounts && zconf_umount `hostname` $MOUNT2 $*
836 [ "$CLIENTONLY" ] && return
837 for num in `seq $MDSCOUNT`; do
840 for num in `seq $OSTCOUNT`; do
855 test $nr = 1 && echo -n $MDS_MKFS_OPTS || echo -n $MDSn_MKFS_OPTS
859 [ "$FSTYPE" ] && FSTYPE_OPT="--backfstype $FSTYPE"
862 # We need ldiskfs here, may as well load them all
864 [ "$CLIENTONLY" ] && return
865 echo "Formatting mdts, osts"
866 for num in `seq $MDSCOUNT`; do
867 echo "Format mds$num: $(mdsdevname $num)"
869 add mds$num `mdsmkfsopts $num` $FSTYPE_OPT --reformat `mdsdevname $num` || exit 9
871 add mds$num `mdsmkfsopts $num` $FSTYPE_OPT --reformat `mdsdevname $num` > /dev/null || exit 9
875 for num in `seq $OSTCOUNT`; do
876 echo "Format ost$num: $(ostdevname $num)"
878 add ost$num $OST_MKFS_OPTS --reformat `ostdevname $num` || exit 10
880 add ost$num $OST_MKFS_OPTS --reformat `ostdevname $num` > /dev/null || exit 10
886 grep " $1 " /proc/mounts || zconf_mount `hostname` $*
892 if [ -z "$CLIENTONLY" ]; then
893 echo "Setup mdts, osts"
894 for num in `seq $MDSCOUNT`; do
895 DEVNAME=$(mdsdevname $num)
896 echo $REFORMAT | grep -q "reformat" \
897 || do_facet mds$num "$TUNEFS --writeconf $DEVNAME"
898 start mds$num $DEVNAME $MDS_MOUNT_OPTS
900 for num in `seq $OSTCOUNT`; do
901 DEVNAME=$(ostdevname $num)
902 start ost$num $DEVNAME $OST_MOUNT_OPTS
905 [ "$DAEMONFILE" ] && $LCTL debug_daemon start $DAEMONFILE $DAEMONSIZE
907 if [ "$MOUNT_2" ]; then
913 mounted_lustre_filesystems() {
914 awk '($3 ~ "lustre" && $1 ~ ":") { print $2 }' /proc/mounts
917 check_and_setup_lustre() {
918 MOUNTED="`mounted_lustre_filesystems`"
919 if [ -z "$MOUNTED" ]; then
920 [ "$REFORMAT" ] && formatall
922 MOUNTED="`mounted_lustre_filesystems`"
923 [ -z "$MOUNTED" ] && error "NAME=$NAME not mounted"
926 if [ "$ONLY" == "setup" ]; then
931 cleanup_and_setup_lustre() {
932 if [ "$ONLY" == "cleanup" -o "`mount | grep $MOUNT`" ]; then
933 sysctl -w lnet.debug=0 || true
935 if [ "$ONLY" == "cleanup" ]; then
939 check_and_setup_lustre
942 check_and_cleanup_lustre() {
943 if [ "`mount | grep $MOUNT`" ]; then
944 rm -rf $DIR/[Rdfs][0-9]*
945 rm -f $DIR/${TESTSUITE}/[Rdfs][1-9]*
947 if [ "$I_MOUNTED" = "yes" ]; then
948 cleanupall -f || error "cleanup failed"
960 while [ $NETWORK -eq 0 ]; do
961 ping -c 1 -w 3 $1 > /dev/null
962 if [ $? -eq 0 ]; then
966 echo "waiting for $1, $((MAX - WAIT)) secs left"
969 if [ $WAIT -gt $MAX ]; then
970 echo "Network not available"
976 while( !($DSH2 $1 "netstat -tna | grep -q $2") ) ; do
987 # the sed converts spaces to commas, but leaves the last space
988 # alone, so the line doesn't end with a comma.
989 echo "$*" | tr -s " " "\n" | sort -b -u | tr "\n" " " | sed 's/ \([^$]\)/,\1/g'
993 (cd `dirname $1`; echo $PWD/`basename $1`)
996 ##################################
1000 # OBD_FAIL_MDS_ALL_REQUEST_NET
1002 do_facet mds sysctl -w lustre.fail_loc=0x123
1003 do_facet client "$1" || RC=$?
1004 do_facet mds sysctl -w lustre.fail_loc=0
1009 # OBD_FAIL_MDS_ALL_REPLY_NET
1011 do_facet mds sysctl -w lustre.fail_loc=0x122
1012 do_facet client "$@" || RC=$?
1013 do_facet mds sysctl -w lustre.fail_loc=0
1017 drop_reint_reply() {
1018 # OBD_FAIL_MDS_REINT_NET_REP
1020 do_facet mds sysctl -w lustre.fail_loc=0x119
1021 do_facet client "$@" || RC=$?
1022 do_facet mds sysctl -w lustre.fail_loc=0
1027 #define OBD_FAIL_OST_BRW_PAUSE_BULK 0x214
1029 do_facet ost1 sysctl -w lustre.fail_loc=0x214
1030 do_facet client "$1" || RC=$?
1031 do_facet client "sync"
1032 do_facet ost1 sysctl -w lustre.fail_loc=0
1036 drop_ldlm_cancel() {
1037 #define OBD_FAIL_LDLM_CANCEL 0x304
1039 do_facet client sysctl -w lustre.fail_loc=0x304
1040 do_facet client "$@" || RC=$?
1041 do_facet client sysctl -w lustre.fail_loc=0
1045 drop_bl_callback() {
1046 #define OBD_FAIL_LDLM_BL_CALLBACK 0x305
1048 do_facet client sysctl -w lustre.fail_loc=0x305
1049 do_facet client "$@" || RC=$?
1050 do_facet client sysctl -w lustre.fail_loc=0
1055 #define OBD_FAIL_LDLM_REPLY 0x30c
1057 do_facet mds sysctl -w lustre.fail_loc=0x30c
1058 do_facet client "$@" || RC=$?
1059 do_facet mds sysctl -w lustre.fail_loc=0
1067 echo "clearing fail_loc on $facet"
1068 do_facet $facet "sysctl -w lustre.fail_loc=0"
1071 cancel_lru_locks() {
1072 $LCTL mark "cancel_lru_locks $1 start"
1073 for d in `find $LPROC/ldlm/namespaces | egrep -i $1`; do
1074 [ -f $d/lru_size ] && echo clear > $d/lru_size
1075 [ -f $d/lock_unused_count ] && grep [1-9] $d/lock_unused_count /dev/null
1077 $LCTL mark "cancel_lru_locks $1 stop"
1082 for a in /proc/fs/lustre/llite/*/dump_page_cache; do
1083 if [ `wc -l $a | awk '{print $1}'` -gt 1 ]; then
1084 echo there is still data in page cache $a ?
1093 DEBUGSAVE="$(sysctl -n lnet.debug)"
1097 [ -n "$DEBUGSAVE" ] && sysctl -w lnet.debug="${DEBUGSAVE}"
1101 ##################################
1103 ##################################
1106 local FAIL_ON_ERROR=${FAIL_ON_ERROR:-true}
1107 local TYPE=${TYPE:-"FAIL"}
1109 sysctl -w lustre.fail_loc=0 2> /dev/null || true
1110 log " ${TESTSUITE} ${TESTNAME}: @@@@@@ ${TYPE}: $@ "
1111 ERRLOG=$TMP/lustre_${TESTSUITE}_${TESTNAME}.$(date +%s)
1112 echo "Dumping lctl log to $ERRLOG"
1113 # We need to dump the logs on all nodes
1114 local NODES=$(nodes_list)
1115 for NODE in $NODES; do
1116 do_node $NODE $LCTL dk $ERRLOG
1119 [ "$TESTSUITELOG" ] && echo "$0: ${TYPE}: $TESTNAME $@" >> $TESTSUITELOG
1120 if $FAIL_ON_ERROR; then
1125 # use only if we are ignoring failures for this test, bugno required.
1126 # (like ALWAYS_EXCEPT, but run the test and ignore the results.)
1127 # e.g. error_ignore 5494 "your message"
1129 FAIL_ON_ERROR=false TYPE="IGNORE (bz$1)" error $2
1133 log " SKIP: ${TESTSUITE} ${TESTNAME} $@"
1134 [ "$TESTSUITELOG" ] && echo "${TESTSUITE}: SKIP: $TESTNAME $@" >> $TESTSUITELOG
1137 build_test_filter() {
1138 [ "$ONLY" ] && log "only running test `echo $ONLY`"
1142 [ "$EXCEPT$ALWAYS_EXCEPT" ] && \
1143 log "skipping tests: `echo $EXCEPT $ALWAYS_EXCEPT`"
1144 [ "$EXCEPT_SLOW" ] && \
1145 log "skipping tests SLOW=no: `echo $EXCEPT_SLOW`"
1146 for E in $EXCEPT $ALWAYS_EXCEPT; do
1147 eval EXCEPT_${E}=true
1149 for E in $EXCEPT_SLOW; do
1150 eval EXCEPT_SLOW_${E}=true
1152 for G in $GRANT_CHECK_LIST; do
1153 eval GCHECK_ONLY_${G}=true
1162 IFS=abcdefghijklmnopqrstuvwxyz _basetest $1
1166 export base=`basetest $1`
1167 if [ ! -z "$ONLY" ]; then
1169 if [ ${!testname}x != x ]; then
1174 if [ ${!testname}x != x ]; then
1182 if [ ${!testname}x != x ]; then
1183 TESTNAME=test_$1 skip "skipping excluded test $1"
1186 testname=EXCEPT_$base
1187 if [ ${!testname}x != x ]; then
1188 TESTNAME=test_$1 skip "skipping excluded test $1 (base $base)"
1191 testname=EXCEPT_SLOW_$1
1192 if [ ${!testname}x != x ]; then
1193 TESTNAME=test_$1 skip "skipping SLOW test $1"
1196 testname=EXCEPT_SLOW_$base
1197 if [ ${!testname}x != x ]; then
1198 TESTNAME=test_$1 skip "skipping SLOW test $1 (base $base)"
1207 EQUALS="======================================================================"
1211 local suffixlen=$((${#EQUALS} - ${#msg}))
1212 [ $suffixlen -lt 5 ] && suffixlen=5
1213 log `echo $(printf '===== %s %.*s\n' "$msg" $suffixlen $EQUALS)`
1218 lsmod | grep lnet > /dev/null || load_modules
1227 local NODES=$(nodes_list)
1228 for NODE in $NODES; do
1229 do_node $NODE $LCTL mark "$MSG" 2> /dev/null || true
1235 strace -o $TMP/$1.strace -ttt $*
1237 log "FINISHED: $*: rc $RC"
1246 FFREE=`cat /proc/fs/lustre/mds/*/filesfree`
1247 FTOTAL=`cat /proc/fs/lustre/mds/*/filestotal`
1248 [ $FFREE -ge $FTOTAL ] && error "files free $FFREE > total $FTOTAL" || true
1255 export tdir=d${TESTSUITE}/d${base}
1256 local SAVE_UMASK=`umask`
1261 log "== test $testnum: $message ============ `date +%H:%M:%S` ($BEFORE)"
1263 export TESTNAME=test_$testnum
1264 test_${testnum} || error "test_$testnum failed with $?"
1266 check_grant ${testnum} || error "check_grant $testnum failed with $?"
1267 [ -f $CATASTROPHE ] && [ `cat $CATASTROPHE` -ne 0 ] && \
1268 error "LBUG/LASSERT detected"
1269 pass "($((`date +%s` - $BEFORE))s)"
1270 rmdir ${DIR}/$tdir >/dev/null 2>&1 || true
1279 (cd `dirname $1`; echo $PWD/`basename $1`)
1283 [ -d $DIR1 ] && cd $DIR1 && sync; sleep 1; sync
1284 [ -d $DIR2 ] && cd $DIR2 && sync; sleep 1; sync
1289 export base=`basetest $1`
1290 [ "$CHECK_GRANT" == "no" ] && return 0
1292 testname=GCHECK_ONLY_${base}
1293 [ ${!testname}x == x ] && return 0
1295 echo -n "checking grant......"
1297 # write some data to sync client lost_grant
1298 rm -f $DIR1/${tfile}_check_grant_* 2>&1
1299 for i in `seq $OSTCOUNT`; do
1300 $LFS setstripe $DIR1/${tfile}_check_grant_$i -i $(($i -1)) -c 1
1301 dd if=/dev/zero of=$DIR1/${tfile}_check_grant_$i bs=4k \
1302 count=1 > /dev/null 2>&1
1304 # sync all the data and make sure no pending data on server
1307 #get client grant and server grant
1309 for d in ${LPROC}/osc/*/cur_grant_bytes; do
1310 client_grant=$((client_grant + `cat $d`))
1313 for d in ${LPROC}/obdfilter/*/tot_granted; do
1314 server_grant=$((server_grant + `cat $d`))
1317 # cleanup the check_grant file
1318 for i in `seq $OSTCOUNT`; do
1319 rm $DIR1/${tfile}_check_grant_$i
1322 #check whether client grant == server grant
1323 if [ $client_grant != $server_grant ]; then
1324 echo "failed: client:${client_grant} server: ${server_grant}"
1331 ########################
1337 ost=`echo $1 | awk -F_ '{print $3}'`
1338 if [ -z $ost ]; then
1339 ost=`echo $1 | sed 's/-osc.*//'`
1346 [ ! -e /proc/fs/lustre/mdt/*MDT* ]
1351 [ $(grep -c obdfilter $LPROC/devices) -eq 0 ]
1355 local MDSNODES=$(facet_host $SINGLEMDS)
1357 # FIXME: Currenly we use only $SINGLEMDS,
1358 # should be fixed when we will start to test cmd.
1362 for num in `seq $MDSCOUNT`; do
1363 local myMDS=$(facet_host mds$num)
1364 [[ ! '\ '"$MDSNODES"'\ ' = *'\ '"$myMDS"'\ '* ]] && MDSNODES="$MDSNODES $myMDS"
1371 local OSTNODES=$(facet_host ost1)
1373 for num in `seq $OSTCOUNT`; do
1374 local myOST=$(facet_host ost$num)
1375 [[ ! '\ '"$OSTNODES"'\ ' = *'\ '"$myOST"'\ '* ]] && OSTNODES="$OSTNODES $myOST"
1382 # FIXME. We need a list of clients
1383 local myNODES=`hostname`
1385 local OSTNODES=$(osts_nodes)
1386 local myOSTNODES=`echo ' '"$OSTNODES"' ' | sed -e s/[\ ]$(hostname)[\ ]/\ /`
1387 [ -n "$myOSTNODES" ] && myNODES="$myNODES $myOSTNODES"
1389 local myNODES=${myNODES% }
1390 # Add to list only not listed mds nodes
1391 local MDSNODES=$(mdts_nodes)
1392 for myMDS in $MDSNODES; do
1393 [[ ! "'\ '$myNODES'\ '" = *'\ '"$myMDS"'\ '* ]] && myNODES="$myNODES $myMDS"
1401 grep -q patchless $LPROC/version
1408 mkdir $DIR/d0_runas_test
1410 chown $myRUNAS_ID:$myRUNAS_ID $DIR/d0_runas_test
1411 $myRUNAS touch $DIR/d0_runas_test/f$$ || \
1412 error "unable to write to $DIR/d0_runas_test as UID $myRUNAS_ID.
1413 Please set RUNAS_ID to some UID which exists on MDS and client or
1414 add user $myRUNAS_ID:$myRUNAS_ID on these nodes."
1415 rm -rf $DIR/d0_runas_test