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 ${ONLY}."
39 local form="%-13s %-17s %s\n"
40 echo "$(printf "$form" "status" "script" "skipped tests")"
41 echo "------------------------------------------------------------------------------------"
42 for O in $TESTSUITE_LIST; do
44 local o=$(echo $O | tr "[:upper:]" "[:lower:]")
47 local log=${TMP}/${o}.log
48 [ -f $log ] && skipped=$(cat $log | awk '{ printf " %s", $3 }' | sed 's/test_//g')
49 [ "${!O}" = "done" ] && \
50 echo "$(printf "$form" "Done" "$O" "$skipped")"
53 for O in $TESTSUITE_LIST; do
54 [ "${!O}" = "no" ] && \
55 echo "$(printf "$form" "Skipped" "$O" "")"
58 for O in $TESTSUITE_LIST; do
59 [ "${!O}" = "done" -o "${!O}" = "no" ] || \
60 echo "$(printf "$form" "UNFINISHED" "$O" "")"
65 export LUSTRE=`absolute_path $LUSTRE`
66 export TESTSUITE=`basename $0 .sh`
67 export LTESTDIR=${LTESTDIR:-$LUSTRE/../ltest}
69 [ -d /r ] && export ROOT=${ROOT:-/r}
70 export TMP=${TMP:-$ROOT/tmp}
71 export TESTSUITELOG=${TMP}/${TESTSUITE}.log
73 export PATH=:$PATH:$LUSTRE/utils:$LUSTRE/utils/gss:$LUSTRE/tests
74 export LCTL=${LCTL:-"$LUSTRE/utils/lctl"}
75 export LFS=${LFS:-"$LUSTRE/utils/lfs"}
76 [ ! -f "$LCTL" ] && export LCTL=$(which lctl)
77 export LFS=${LFS:-"$LUSTRE/utils/lfs"}
78 [ ! -f "$LFS" ] && export LFS=$(which lfs)
79 export MKFS=${MKFS:-"$LUSTRE/utils/mkfs.lustre"}
80 [ ! -f "$MKFS" ] && export MKFS=$(which mkfs.lustre)
81 export TUNEFS=${TUNEFS:-"$LUSTRE/utils/tunefs.lustre"}
82 [ ! -f "$TUNEFS" ] && export TUNEFS=$(which tunefs.lustre)
83 export CHECKSTAT="${CHECKSTAT:-"checkstat -v"} "
84 export FSYTPE=${FSTYPE:-"ldiskfs"}
85 export NAME=${NAME:-local}
86 export LPROC=/proc/fs/lustre
87 export LGSSD=${LGSSD:-"$LUSTRE/utils/gss/lgssd"}
88 export LSVCGSSD=${LSVCGSSD:-"$LUSTRE/utils/gss/lsvcgssd"}
89 export KRB5DIR=${KRB5DIR:-"/usr/kerberos"}
92 if [ "$ACCEPTOR_PORT" ]; then
93 export PORT_OPT="--port $ACCEPTOR_PORT"
98 echo "Using GSS/krb5 ptlrpc security flavor"
104 # Paths on remote nodes, if different
105 export RLUSTRE=${RLUSTRE:-$LUSTRE}
106 export RPWD=${RPWD:-$PWD}
107 export I_MOUNTED=${I_MOUNTED:-"no"}
111 while getopts "rvf:" opt $*; do
114 r) REFORMAT=--reformat;;
120 shift $((OPTIND - 1))
123 [ "$TESTSUITELOG" ] && rm -f $TESTSUITELOG || true
131 BASE=`basename $module $EXT`
132 lsmod | grep -q ${BASE} || \
133 if [ -f ${LUSTRE}/${module}${EXT} ]; then
134 insmod ${LUSTRE}/${module}${EXT} $@
136 # must be testing a "make install" or "rpm" installation
137 # note failed to load ptlrpc_gss is considered not fatal
138 if [ "$BASE" == "ptlrpc_gss" ]; then
139 modprobe $BASE $@ 2>/dev/null || echo "gss/krb5 is not supported"
147 if [ -n "$MODPROBE" ]; then
151 if [ "$HAVE_MODULES" = true ]; then
157 echo Loading modules from $LUSTRE
158 load_module ../lnet/libcfs/libcfs
159 [ -z "$LNETOPTS" ] && \
160 LNETOPTS=$(awk '/^options lnet/ { print $0}' /etc/modprobe.conf | sed 's/^options lnet //g')
161 echo "lnet options: '$LNETOPTS'"
162 # note that insmod will ignore anything in modprobe.conf
163 load_module ../lnet/lnet/lnet $LNETOPTS
164 LNETLND=${LNETLND:-"socklnd/ksocklnd"}
165 load_module ../lnet/klnds/$LNETLND
166 [ "$FSTYPE" = "ldiskfs" ] && load_module ../ldiskfs/ldiskfs/ldiskfs
167 load_module lvfs/lvfs
168 load_module obdclass/obdclass
169 load_module lvfs/fsfilt_$FSTYPE
170 load_module ptlrpc/ptlrpc
171 load_module ptlrpc/gss/ptlrpc_gss
172 # Now, some modules depend on lquota without USE_QUOTA check,
173 # will fix later. Disable check "$USE_QUOTA" = "yes" temporary.
174 #[ "$USE_QUOTA" = "yes" ] && load_module quota/lquota
175 load_module quota/lquota
188 load_module obdfilter/obdfilter
189 load_module llite/lustre
190 load_module llite/llite_lloop
193 rm -f $TMP/ogdb-`hostname`
194 $LCTL modules > $TMP/ogdb-`hostname`
195 # 'mount' doesn't look in $PATH, just sbin
196 [ -f $LUSTRE/utils/mount.lustre ] && cp $LUSTRE/utils/mount.lustre /sbin/. || true
200 if [ `uname -r | cut -c 3` -eq 4 ]; then
208 MODULES=$($LCTL modules | awk '{ print $2 }')
209 while [ -n "$MODULES" ]; do
211 $RMMOD $MODULES > /dev/null 2>&1 || true
212 MODULES=$($LCTL modules | awk '{ print $2 }')
213 if [ -z "$MODULES" ]; then
217 echo "waiting, $((MAX - WAIT)) secs left"
219 if [ $WAIT -eq $MAX ]; then
220 echo "LNET modules $MODULES will not unload"
228 wait_exit_ST client # bug 12845
230 lsmod | grep lnet > /dev/null && $LCTL dl && $LCTL dk $TMP/debug
231 local MODULES=$($LCTL modules | awk '{ print $2 }')
232 $RMMOD $MODULES > /dev/null 2>&1 || true
233 # do it again, in case we tried to unload ksocklnd too early
234 MODULES=$($LCTL modules | awk '{ print $2 }')
235 [ -n "$MODULES" ] && $RMMOD $MODULES > /dev/null 2>&1 || true
236 MODULES=$($LCTL modules | awk '{ print $2 }')
237 if [ -n "$MODULES" ]; then
238 echo "Modules still loaded: "
240 if [ -e $LPROC ]; then
241 echo "Lustre still loaded"
242 cat $LPROC/devices || true
246 echo "Lustre stopped but LNET is still loaded, waiting..."
247 wait_for_lnet || return 3
252 LEAK_LUSTRE=$(dmesg | tail -n 30 | grep "obd mem.*leaked" || true)
253 LEAK_PORTALS=$(dmesg | tail -n 20 | grep "Portals memory leaked" || true)
254 if [ "$LEAK_LUSTRE" -o "$LEAK_PORTALS" ]; then
255 echo "$LEAK_LUSTRE" 1>&2
256 echo "$LEAK_PORTALS" 1>&2
257 mv $TMP/debug $TMP/debug-leak.`date +%s` || true
258 echo "Memory leaks detected"
259 [ -n "$IGNORE_LEAK" ] && echo "ignoring leaks" && return 0
262 echo "modules unloaded."
266 check_gss_daemon_facet() {
270 num=`do_facet $facet ps -o cmd -C $dname | grep $dname | wc -l`
271 if [ $num -ne 1 ]; then
272 echo "$num instance of $dname on $facet"
281 do_facet $facet "killall -2 $@ 2>/dev/null || true"
284 start_gss_daemons() {
286 for num in `seq $MDSCOUNT`; do
287 do_facet mds$num "$LSVCGSSD -v"
289 do_facet mds$num "$LGSSD -v"
293 for num in `seq $OSTCOUNT`; do
294 do_facet ost$num "$LSVCGSSD -v"
297 # FIXME: is "client" the right facet name?
299 do_facet client "$LGSSD -v"
302 # wait daemons entering "stable" status
306 # check daemons are running
308 for num in `seq $MDSCOUNT`; do
309 check_gss_daemon_facet mds$num lsvcgssd
311 check_gss_daemon_facet mds$num lgssd
314 for num in `seq $OSTCOUNT`; do
315 check_gss_daemon_facet ost$num lsvcgssd
318 check_gss_daemon_facet client lgssd
323 for num in `seq $MDSCOUNT`; do
324 send_sigint mds$num lsvcgssd lgssd
326 for num in `seq $OSTCOUNT`; do
327 send_sigint ost$num lsvcgssd
329 send_sigint client lgssd
333 if [ ! -z $SEC ]; then
334 MDS_MOUNT_OPTS=$MDS_MOUNT_OPTS,sec=$SEC
335 OST_MOUNT_OPTS=$OST_MOUNT_OPTS,sec=$SEC
346 # maybe cleanup credential cache?
352 local device=`mdsdevname $num`
353 local label=`do_facet mds$num "e2label ${device}" | grep -v "CMD: "`
359 local device=`ostdevname $num`
360 local label=`do_facet ost$num "e2label ${device}" | grep -v "CMD: "`
365 # start facet device options
371 echo "Starting ${facet}: $@ ${device} ${MOUNT%/*}/${facet}"
372 do_facet ${facet} mkdir -p ${MOUNT%/*}/${facet}
373 do_facet ${facet} mount -t lustre $@ ${device} ${MOUNT%/*}/${facet}
375 if [ $RC -ne 0 ]; then
376 echo mount -t lustre $@ ${device} ${MOUNT%/*}/${facet}
377 echo Start of ${device} on ${facet} failed ${RC}
379 do_facet ${facet} "sysctl -w lnet.debug=$PTLDEBUG; \
380 sysctl -w lnet.subsystem_debug=${SUBSYSTEM# }; \
381 sysctl -w lnet.debug_mb=${DEBUG_SIZE}"
383 do_facet ${facet} sync
384 label=$(do_facet ${facet} "e2label ${device}")
385 [ -z "$label" ] && echo no label for ${device} && exit 1
386 eval export ${facet}_svc=${label}
387 eval export ${facet}_dev=${device}
388 eval export ${facet}_opt=\"$@\"
389 echo Started ${label}
398 HOST=`facet_active_host $facet`
399 [ -z $HOST ] && echo stop: no host for $facet && return 0
401 running=$(do_facet ${facet} "grep -c ${MOUNT%/*}/${facet}' ' /proc/mounts") || true
402 if [ ${running} -ne 0 ]; then
403 echo "Stopping ${MOUNT%/*}/${facet} (opts:$@)"
404 do_facet ${facet} umount -d $@ ${MOUNT%/*}/${facet}
407 # umount should block, but we should wait for unrelated obd's
408 # like the MGS or MGC to also stop.
409 wait_exit_ST ${facet}
416 # Only supply -o to mount if we have options
417 if [ -n "$MOUNTOPT" ]; then
418 OPTIONS="-o $MOUNTOPT"
420 local device=$MGSNID:/$FSNAME
421 if [ -z "$mnt" -o -z "$FSNAME" ]; then
422 echo Bad zconf mount command: opt=$OPTIONS dev=$device mnt=$mnt
426 echo "Starting client: $OPTIONS $device $mnt"
427 do_node $client mkdir -p $mnt
428 do_node $client mount -t lustre $OPTIONS $device $mnt || return 1
430 do_node $client "sysctl -w lnet.debug=$PTLDEBUG;
431 sysctl -w lnet.subsystem_debug=${SUBSYSTEM# };
432 sysctl -w lnet.debug_mb=${DEBUG_SIZE}"
433 [ -d /r ] && $LCTL modules > /r/tmp/ogdb-`hostname`
441 local running=$(do_node $client "grep -c $mnt' ' /proc/mounts") || true
442 if [ $running -ne 0 ]; then
443 echo "Stopping client $mnt (opts:$force)"
444 do_node $client umount $force $mnt
450 if [ "$FAILURE_MODE" = HARD ]; then
451 $POWER_DOWN `facet_active_host $facet`
453 elif [ "$FAILURE_MODE" = SOFT ]; then
460 if [ "$FAILURE_MODE" = HARD ]; then
461 $POWER_UP `facet_active_host $facet`
467 # verify that lustre actually cleaned up properly
469 [ -f $CATASTROPHE ] && [ `cat $CATASTROPHE` -ne 0 ] && \
470 error "LBUG/LASSERT detected"
471 BUSY=`dmesg | grep -i destruct || true`
474 [ -e $TMP/debug ] && mv $TMP/debug $TMP/debug-busy.`date +%s`
477 LEAK_LUSTRE=`dmesg | tail -n 30 | grep "obd mem.*leaked" || true`
478 LEAK_PORTALS=`dmesg | tail -n 20 | grep "Portals memory leaked" || true`
479 if [ "$LEAK_LUSTRE" -o "$LEAK_PORTALS" ]; then
480 echo "$0: $LEAK_LUSTRE" 1>&2
481 echo "$0: $LEAK_PORTALS" 1>&2
482 echo "$0: Memory leak(s) detected..." 1>&2
483 mv $TMP/debug $TMP/debug-leak.`date +%s`
487 [ "`lctl dl 2> /dev/null | wc -l`" -gt 0 ] && lctl dl && \
488 echo "$0: lustre didn't clean up..." 1>&2 && return 202 || true
490 if [ "`/sbin/lsmod 2>&1 | egrep 'lnet|libcfs'`" ]; then
491 echo "$0: modules still loaded..." 1>&2
498 wait_delete_completed () {
499 local TOTALPREV=`awk 'BEGIN{total=0}; {total+=$1}; END{print total}' \
500 $LPROC/osc/*/kbytesavail`
504 while [ "$WAIT" -ne "$MAX_WAIT" ]; do
506 TOTAL=`awk 'BEGIN{total=0}; {total+=$1}; END{print total}' \
507 $LPROC/osc/*/kbytesavail`
508 [ "$TOTAL" -eq "$TOTALPREV" ] && break
509 echo "Waiting delete completed ... prev: $TOTALPREV current: $TOTAL "
513 echo "Delete completed."
518 check_network "$HOST" 900
519 while ! do_node $HOST "ls -d $LUSTRE " > /dev/null; do sleep 5; done
524 HOST=`facet_active_host $facet`
528 wait_mds_recovery_done () {
529 local timeout=`do_facet mds cat /proc/sys/lustre/timeout`
530 #define OBD_RECOVERY_TIMEOUT (obd_timeout * 5 / 2)
531 # as we are in process of changing obd_timeout in different ways
532 # let's set MAX longer than that
533 MAX=$(( timeout * 4 ))
535 while [ $WAIT -lt $MAX ]; do
536 STATUS=`do_facet mds grep status /proc/fs/lustre/mdt/*-MDT*/recovery_status`
537 echo $STATUS | grep COMPLETE && return 0
540 echo "Waiting $(($MAX - $WAIT)) secs for MDS recovery done"
542 echo "MDS recovery not done in $MAX sec"
551 # conf-sanity 31 takes a long time cleanup
552 while [ $WAIT -lt 300 ]; do
553 running=$(do_facet ${facet} "[ -e $LPROC ] && grep ST' ' $LPROC/devices") || true
554 [ -z "${running}" ] && return 0
555 echo "waited $WAIT for${running}"
556 [ $INTERVAL -lt 64 ] && INTERVAL=$((INTERVAL + INTERVAL))
558 WAIT=$((WAIT + INTERVAL))
560 echo "service didn't stop after $WAIT seconds. Still running:"
566 # not every config has many clients
567 if [ ! -z "$CLIENTS" ]; then
568 $PDSH $CLIENTS "df $MOUNT" > /dev/null
573 uname -n >> $MOUNT/recon
574 if [ ! -z "$CLIENTS" ]; then
575 $PDSH $CLIENTS "df $MOUNT; uname -n >> $MOUNT/recon" > /dev/null
577 echo Connected clients:
579 ls -l $MOUNT/recon > /dev/null
585 echo "Failing $facet on node `facet_active_host $facet`"
586 shutdown_facet $facet
590 echo "df pid is $DFPID"
592 TO=`facet_active_host $facet`
593 echo "Failover $facet to $TO"
595 local dev=${facet}_dev
596 local opt=${facet}_opt
597 start $facet ${!dev} ${!opt} || error "Restart of $facet failed"
608 local svc=${facet}_svc
609 do_facet $facet $LCTL --device %${!svc} readonly
610 do_facet $facet $LCTL --device %${!svc} notransno
611 do_facet $facet $LCTL mark "$facet REPLAY BARRIER on ${!svc}"
612 $LCTL mark "local REPLAY BARRIER on ${!svc}"
615 replay_barrier_nodf() {
616 local facet=$1 echo running=${running}
618 local svc=${facet}_svc
619 echo Replay barrier on ${!svc}
620 do_facet $facet $LCTL --device %${!svc} readonly
621 do_facet $facet $LCTL --device %${!svc} notransno
622 do_facet $facet $LCTL mark "$facet REPLAY BARRIER on ${!svc}"
623 $LCTL mark "local REPLAY BARRIER on ${!svc}"
627 UUID=`cat /proc/fs/lustre/mdc/${mds1_svc}-mdc-*/uuid`
628 do_facet mds1 "echo $UUID > /proc/fs/lustre/mdt/${mds1_svc}/evict_client"
632 UUID=`grep ${ost1_svc}-osc- $LPROC/devices | egrep -v 'MDT' | awk '{print $5}'`
633 do_facet ost1 "echo $UUID > /proc/fs/lustre/obdfilter/${ost1_svc}/evict_client"
637 facet_failover $* || error "failover: $?"
638 df $MOUNT || error "post-failover df: $?"
645 local svc=${facet}_svc
646 local dev=${facet}_dev
647 local opt=${facet}_opt
648 start $facet ${!dev} ${!opt}
649 do_facet $facet lctl --device %${!svc} abort_recovery
650 df $MOUNT || echo "first df failed: $?"
652 df $MOUNT || error "post-failover df: $?"
656 echo There is no lmc. This is mountconf, baby.
661 if [ "$1" = "client" -o "$1" = "'*'" ]; then echo \'*\'; else
662 ID=`$PDSH $1 $GMNALNID -l | cut -d\ -f2`
668 if [ "$1" = "client" -o "$1" = "'*'" ]; then echo \'*\'; else
674 if [ "$1" = "client" -o "$1" = "'*'" ]; then echo \'*\'; else
675 ID=`xtprocadmin -n $1 2>/dev/null | egrep -v 'NID' | awk '{print $1}'`
676 if [ -z "$ID" ]; then
677 echo "Could not get a ptl id for $1..."
686 h2name_or_ip "$1" "tcp"
691 if [ "$1" = "client" -o "$1" = "'*'" ]; then echo \'*\'; else
692 if type __h2elan >/dev/null 2>&1; then
695 ID=`echo $1 | sed 's/[^0-9]*//g'`
703 h2name_or_ip "$1" "openib"
708 h2name_or_ip "$1" "o2ib"
714 varname=${facet}_HOST
715 if [ -z "${!varname}" ]; then
716 if [ "${facet:0:3}" == "ost" ]; then
717 eval ${facet}_HOST=${ost_HOST}
725 local activevar=${facet}active
727 if [ -f ./${facet}active ] ; then
728 source ./${facet}active
732 if [ -z "$active" ] ; then
739 facet_active_host() {
741 local active=`facet_active $facet`
742 if [ "$facet" == client ]; then
745 echo `facet_host $active`
751 failover=${facet}failover
752 host=`facet_host $failover`
753 [ -z "$host" ] && return
754 curactive=`facet_active $facet`
755 if [ -z "${curactive}" -o "$curactive" == "$failover" ] ; then
756 eval export ${facet}active=$facet
758 eval export ${facet}active=$failover
760 # save the active host for this facet
761 activevar=${facet}active
762 echo "$activevar=${!activevar}" > ./$activevar
769 if [ "$HOST" = "$(hostname)" ]; then
773 echo "CMD: $HOST $@" >&2
774 $myPDSH $HOST $LCTL mark "$@" > /dev/null 2>&1 || :
776 $myPDSH $HOST "(PATH=\$PATH:$RLUSTRE/utils:$RLUSTRE/tests:/sbin:/usr/sbin; cd $RPWD; sh -c \"$@\")" | sed "s/^${HOST}: //"
777 return ${PIPESTATUS[0]}
783 HOST=`facet_active_host $facet`
784 [ -z $HOST ] && echo No host defined for facet ${facet} && exit 1
791 # make sure its not already running
794 do_facet ${facet} $MKFS $*
800 #if $OSTDEVn isn't defined, default is $OSTDEVBASE + num
801 eval DEVPTR=${!DEVNAME:=${OSTDEVBASE}${num}}
808 #if $MDSDEVn isn't defined, default is $MDSDEVBASE + num
809 eval DEVPTR=${!DEVNAME:=${MDSDEVBASE}${num}}
817 # make sure we are using the primary server, so test-framework will
818 # be able to clean up properly.
819 activemds=`facet_active mds1`
820 if [ $activemds != "mds1" ]; then
824 # assume client mount is local
825 grep " $MOUNT " /proc/mounts && zconf_umount `hostname` $MOUNT $*
826 grep " $MOUNT2 " /proc/mounts && zconf_umount `hostname` $MOUNT2 $*
827 for num in `seq $MDSCOUNT`; do
830 for num in `seq $OSTCOUNT`; do
845 test $nr = 1 && echo -n $MDS_MKFS_OPTS || echo -n $MDSn_MKFS_OPTS
849 [ "$FSTYPE" ] && FSTYPE_OPT="--backfstype $FSTYPE"
852 # We need ldiskfs here, may as well load them all
854 [ "$CLIENTONLY" ] && return
855 echo "Formatting mdts, osts"
856 for num in `seq $MDSCOUNT`; do
857 echo "Format mds$num: $(mdsdevname $num)"
859 add mds$num `mdsmkfsopts $num` $FSTYPE_OPT --reformat `mdsdevname $num` || exit 9
861 add mds$num `mdsmkfsopts $num` $FSTYPE_OPT --reformat `mdsdevname $num` > /dev/null || exit 9
865 for num in `seq $OSTCOUNT`; do
866 echo "Format ost$num: $(ostdevname $num)"
868 add ost$num $OST_MKFS_OPTS --reformat `ostdevname $num` || exit 10
870 add ost$num $OST_MKFS_OPTS --reformat `ostdevname $num` > /dev/null || exit 10
876 grep " $1 " /proc/mounts || zconf_mount `hostname` $*
882 if [ -z "$CLIENTONLY" ]; then
883 echo "Setup mdts, osts"
884 for num in `seq $MDSCOUNT`; do
885 DEVNAME=$(mdsdevname $num)
886 echo $REFORMAT | grep -q "reformat" \
887 || do_facet mds$num "$TUNEFS --writeconf $DEVNAME"
888 start mds$num $DEVNAME $MDS_MOUNT_OPTS
890 for num in `seq $OSTCOUNT`; do
891 DEVNAME=$(ostdevname $num)
892 start ost$num $DEVNAME $OST_MOUNT_OPTS
895 [ "$DAEMONFILE" ] && $LCTL debug_daemon start $DAEMONFILE $DAEMONSIZE
897 if [ "$MOUNT_2" ]; then
903 mounted_lustre_filesystems() {
904 awk '($3 ~ "lustre" && $1 ~ ":") { print $2 }' /proc/mounts
907 check_and_setup_lustre() {
908 MOUNTED="`mounted_lustre_filesystems`"
909 if [ -z "$MOUNTED" ]; then
910 [ "$REFORMAT" ] && formatall
912 MOUNTED="`mounted_lustre_filesystems`"
913 [ -z "$MOUNTED" ] && error "NAME=$NAME not mounted"
916 if [ "$ONLY" == "setup" ]; then
921 cleanup_and_setup_lustre() {
922 if [ "$ONLY" == "cleanup" -o "`mount | grep $MOUNT`" ]; then
923 sysctl -w lnet.debug=0 || true
925 if [ "$ONLY" == "cleanup" ]; then
929 check_and_setup_lustre
932 check_and_cleanup_lustre() {
933 if [ "`mount | grep $MOUNT`" ]; then
934 rm -rf $DIR/[Rdfs][0-9]*
935 rm -f $DIR/${TESTSUITE}/[Rdfs][1-9]*
937 if [ "$I_MOUNTED" = "yes" ]; then
938 cleanupall -f || error "cleanup failed"
950 while [ $NETWORK -eq 0 ]; do
951 ping -c 1 -w 3 $1 > /dev/null
952 if [ $? -eq 0 ]; then
956 echo "waiting for $1, $((MAX - WAIT)) secs left"
959 if [ $WAIT -gt $MAX ]; then
960 echo "Network not available"
966 while( !($DSH2 $1 "netstat -tna | grep -q $2") ) ; do
977 # the sed converts spaces to commas, but leaves the last space
978 # alone, so the line doesn't end with a comma.
979 echo "$*" | tr -s " " "\n" | sort -b -u | tr "\n" " " | sed 's/ \([^$]\)/,\1/g'
983 (cd `dirname $1`; echo $PWD/`basename $1`)
986 ##################################
990 # OBD_FAIL_MDS_ALL_REQUEST_NET
992 do_facet mds sysctl -w lustre.fail_loc=0x123
993 do_facet client "$1" || RC=$?
994 do_facet mds sysctl -w lustre.fail_loc=0
999 # OBD_FAIL_MDS_ALL_REPLY_NET
1001 do_facet mds sysctl -w lustre.fail_loc=0x122
1002 do_facet client "$@" || RC=$?
1003 do_facet mds sysctl -w lustre.fail_loc=0
1007 drop_reint_reply() {
1008 # OBD_FAIL_MDS_REINT_NET_REP
1010 do_facet mds sysctl -w lustre.fail_loc=0x119
1011 do_facet client "$@" || RC=$?
1012 do_facet mds sysctl -w lustre.fail_loc=0
1017 #define OBD_FAIL_OST_BRW_PAUSE_BULK 0x214
1019 do_facet ost1 sysctl -w lustre.fail_loc=0x214
1020 do_facet client "$1" || RC=$?
1021 do_facet client "sync"
1022 do_facet ost1 sysctl -w lustre.fail_loc=0
1026 drop_ldlm_cancel() {
1027 #define OBD_FAIL_LDLM_CANCEL 0x304
1029 do_facet client sysctl -w lustre.fail_loc=0x304
1030 do_facet client "$@" || RC=$?
1031 do_facet client sysctl -w lustre.fail_loc=0
1035 drop_bl_callback() {
1036 #define OBD_FAIL_LDLM_BL_CALLBACK 0x305
1038 do_facet client sysctl -w lustre.fail_loc=0x305
1039 do_facet client "$@" || RC=$?
1040 do_facet client sysctl -w lustre.fail_loc=0
1045 #define OBD_FAIL_LDLM_REPLY 0x30c
1047 do_facet mds sysctl -w lustre.fail_loc=0x30c
1048 do_facet client "$@" || RC=$?
1049 do_facet mds sysctl -w lustre.fail_loc=0
1057 echo "clearing fail_loc on $facet"
1058 do_facet $facet "sysctl -w lustre.fail_loc=0"
1061 cancel_lru_locks() {
1062 $LCTL mark "cancel_lru_locks $1 start"
1063 for d in `find $LPROC/ldlm/namespaces | egrep -i $1`; do
1064 [ -f $d/lru_size ] && echo clear > $d/lru_size
1065 [ -f $d/lock_unused_count ] && grep [1-9] $d/lock_unused_count /dev/null
1067 $LCTL mark "cancel_lru_locks $1 stop"
1072 for a in /proc/fs/lustre/llite/*/dump_page_cache; do
1073 if [ `wc -l $a | awk '{print $1}'` -gt 1 ]; then
1074 echo there is still data in page cache $a ?
1083 DEBUGSAVE="$(sysctl -n lnet.debug)"
1087 [ -n "$DEBUGSAVE" ] && sysctl -w lnet.debug="${DEBUGSAVE}"
1091 ##################################
1093 ##################################
1096 local FAIL_ON_ERROR=${FAIL_ON_ERROR:-true}
1097 local TYPE=${TYPE:-"FAIL"}
1099 sysctl -w lustre.fail_loc=0 2> /dev/null || true
1100 log " ${TESTSUITE} ${TESTNAME}: @@@@@@ ${TYPE}: $@ "
1101 ERRLOG=$TMP/lustre_${TESTSUITE}_${TESTNAME}.$(date +%s)
1102 echo "Dumping lctl log to $ERRLOG"
1103 # We need to dump the logs on all nodes
1104 local NODES=$(nodes_list)
1105 for NODE in $NODES; do
1106 do_node $NODE $LCTL dk $ERRLOG
1109 [ "$TESTSUITELOG" ] && echo "$0: ${TYPE}: $TESTNAME $@" >> $TESTSUITELOG
1110 if $FAIL_ON_ERROR; then
1115 # use only if we are ignoring failures for this test, bugno required.
1116 # (like ALWAYS_EXCEPT, but run the test and ignore the results.)
1117 # e.g. error_ignore 5494 "your message"
1119 FAIL_ON_ERROR=false TYPE="IGNORE (bz$1)" error $2
1123 log " SKIP: ${TESTSUITE} ${TESTNAME} $@"
1124 [ "$TESTSUITELOG" ] && echo "${TESTSUITE}: SKIP: $TESTNAME $@" >> $TESTSUITELOG
1127 build_test_filter() {
1128 [ "$ONLY" ] && log "only running test `echo $ONLY`"
1132 [ "$EXCEPT$ALWAYS_EXCEPT" ] && \
1133 log "skipping tests: `echo $EXCEPT $ALWAYS_EXCEPT`"
1134 for E in $EXCEPT $ALWAYS_EXCEPT; do
1135 eval EXCEPT_${E}=true
1137 for G in $GRANT_CHECK_LIST; do
1138 eval GCHECK_ONLY_${G}=true
1147 IFS=abcdefghijklmnopqrstuvwxyz _basetest $1
1151 export base=`basetest $1`
1152 if [ ! -z "$ONLY" ]; then
1154 if [ ${!testname}x != x ]; then
1159 if [ ${!testname}x != x ]; then
1167 if [ ${!testname}x != x ]; then
1168 TESTNAME=test_$1 skip "skipping excluded test $1"
1171 testname=EXCEPT_$base
1172 if [ ${!testname}x != x ]; then
1173 TESTNAME=test_$1 skip "skipping excluded test $1 (base $base)"
1181 EQUALS="======================================================================"
1185 local suffixlen=$((${#EQUALS} - ${#msg}))
1186 [ $suffixlen -lt 5 ] && suffixlen=5
1187 log `echo $(printf '===== %s %.*s\n' "$msg" $suffixlen $EQUALS)`
1192 lsmod | grep lnet > /dev/null || load_modules
1201 local NODES=$(nodes_list)
1202 for NODE in $NODES; do
1203 do_node $NODE $LCTL mark "$MSG" 2> /dev/null || true
1209 strace -o $TMP/$1.strace -ttt $*
1211 log "FINISHED: $*: rc $RC"
1220 FFREE=`cat /proc/fs/lustre/mds/*/filesfree`
1221 FTOTAL=`cat /proc/fs/lustre/mds/*/filestotal`
1222 [ $FFREE -ge $FTOTAL ] && error "files free $FFREE > total $FTOTAL" || true
1229 export tdir=d${TESTSUITE}/d${base}
1230 local SAVE_UMASK=`umask`
1235 log "== test $testnum: $message ============ `date +%H:%M:%S` ($BEFORE)"
1237 export TESTNAME=test_$testnum
1238 test_${testnum} || error "test_$testnum failed with $?"
1240 check_grant ${testnum} || error "check_grant $testnum failed with $?"
1241 [ -f $CATASTROPHE ] && [ `cat $CATASTROPHE` -ne 0 ] && \
1242 error "LBUG/LASSERT detected"
1243 pass "($((`date +%s` - $BEFORE))s)"
1244 rmdir ${DIR}/$tdir >/dev/null 2>&1 || true
1253 (cd `dirname $1`; echo $PWD/`basename $1`)
1257 [ -d $DIR1 ] && cd $DIR1 && sync; sleep 1; sync
1258 [ -d $DIR2 ] && cd $DIR2 && sync; sleep 1; sync
1263 export base=`basetest $1`
1264 [ "$CHECK_GRANT" == "no" ] && return 0
1266 testname=GCHECK_ONLY_${base}
1267 [ ${!testname}x == x ] && return 0
1269 echo -n "checking grant......"
1271 # write some data to sync client lost_grant
1272 rm -f $DIR1/${tfile}_check_grant_* 2>&1
1273 for i in `seq $OSTCOUNT`; do
1274 $LFS setstripe $DIR1/${tfile}_check_grant_$i 0 $(($i -1)) 1
1275 dd if=/dev/zero of=$DIR1/${tfile}_check_grant_$i bs=4k \
1276 count=1 > /dev/null 2>&1
1278 # sync all the data and make sure no pending data on server
1281 #get client grant and server grant
1283 for d in ${LPROC}/osc/*/cur_grant_bytes; do
1284 client_grant=$((client_grant + `cat $d`))
1287 for d in ${LPROC}/obdfilter/*/tot_granted; do
1288 server_grant=$((server_grant + `cat $d`))
1291 # cleanup the check_grant file
1292 for i in `seq $OSTCOUNT`; do
1293 rm $DIR1/${tfile}_check_grant_$i
1296 #check whether client grant == server grant
1297 if [ $client_grant != $server_grant ]; then
1298 echo "failed: client:${client_grant} server: ${server_grant}"
1305 ########################
1311 ost=`echo $1 | awk -F_ '{print $3}'`
1312 if [ -z $ost ]; then
1313 ost=`echo $1 | sed 's/-osc.*//'`
1320 [ ! -e /proc/fs/lustre/mdt/*MDT* ]
1325 [ $(grep -c obdfilter $LPROC/devices) -eq 0 ]
1329 local MDSNODES=$(facet_host $SINGLEMDS)
1331 # FIXME: Currenly we use only $SINGLEMDS,
1332 # should be fixed when we will start to test cmd.
1336 for num in `seq $MDSCOUNT`; do
1337 local myMDS=$(facet_host mds$num)
1338 [[ ! '\ '"$MDSNODES"'\ ' = *'\ '"$myMDS"'\ '* ]] && MDSNODES="$MDSNODES $myMDS"
1345 local OSTNODES=$(facet_host ost1)
1347 for num in `seq $OSTCOUNT`; do
1348 local myOST=$(facet_host ost$num)
1349 [[ ! '\ '"$OSTNODES"'\ ' = *'\ '"$myOST"'\ '* ]] && OSTNODES="$OSTNODES $myOST"
1356 # FIXME. We need a list of clients
1357 local myNODES=`hostname`
1359 local OSTNODES=$(osts_nodes)
1360 local myOSTNODES=`echo ' '"$OSTNODES"' ' | sed -e s/[\ ]$(hostname)[\ ]/\ /`
1361 [ -n "$myOSTNODES" ] && myNODES="$myNODES $myOSTNODES"
1363 local myNODES=${myNODES% }
1364 # Add to list only not listed mds nodes
1365 local MDSNODES=$(mdts_nodes)
1366 for myMDS in $MDSNODES; do
1367 [[ ! "'\ '$myNODES'\ '" = *'\ '"$myMDS"'\ '* ]] && myNODES="$myNODES $myMDS"
1375 grep -q patchless $LPROC/version
1382 mkdir $DIR/d0_runas_test
1384 chown $myRUNAS_ID:$myRUNAS_ID $DIR/d0_runas_test
1385 $myRUNAS touch $DIR/d0_runas_test/f$$ || \
1386 error "unable to write to $DIR/d0_runas_test as UID $myRUNAS_ID.
1387 Please set RUNAS_ID to some UID which exists on MDS and client or
1388 add user $myRUNAS_ID:$myRUNAS_ID on these nodes."
1389 rm -rf $DIR/d0_runas_test