. $EXCEPT_LIST_FILE
fi
-[ -z "$MODPROBECONF" -a -f /etc/modprobe.conf ] && MODPROBECONF=/etc/modprobe.conf
-[ -z "$MODPROBECONF" -a -f /etc/modprobe.d/Lustre ] && MODPROBECONF=/etc/modprobe.d/Lustre
+# check config files for options in decreasing order of preference
+[ -z "$MODPROBECONF" -a -f /etc/modprobe.d/lustre.conf ] &&
+ MODPROBECONF=/etc/modprobe.d/lustre.conf
+[ -z "$MODPROBECONF" -a -f /etc/modprobe.d/Lustre ] &&
+ MODPROBECONF=/etc/modprobe.d/Lustre
+[ -z "$MODPROBECONF" -a -f /etc/modprobe.conf ] &&
+ MODPROBECONF=/etc/modprobe.conf
assert_DIR () {
local failed=""
export LFSCK_ALWAYS=${LFSCK_ALWAYS:-"no"} # check fs after each test suite
export FSCK_MAX_ERR=4 # File system errors left uncorrected
- # This is used by a small number of tests to share state between the client
- # running the tests, or in some cases between the servers (e.g. lfsck.sh).
- # It needs to be a non-lustre filesystem that is available on all the nodes.
- export SHARED_DIRECTORY=${SHARED_DIRECTORY:-"/tmp"}
- export MDSDB=${MDSDB:-$SHARED_DIRECTORY/mdsdb}
- export OSTDB=${OSTDB:-$SHARED_DIRECTORY/ostdb}
-
#[ -d /r ] && export ROOT=${ROOT:-/r}
export TMP=${TMP:-$ROOT/tmp}
export TESTSUITELOG=${TMP}/${TESTSUITE}.log
- if [[ -z $LOGDIRSET ]]; then
- export LOGDIR=${LOGDIR:-${TMP}/test_logs/}/$(date +%s)
- export LOGDIRSET=true
- fi
- export HOSTNAME=${HOSTNAME:-`hostname`}
+ export LOGDIR=${LOGDIR:-${TMP}/test_logs/$(date +%s)}
+ export TESTLOG_PREFIX=$LOGDIR/$TESTSUITE
+
+ export HOSTNAME=${HOSTNAME:-$(hostname -s)}
if ! echo $PATH | grep -q $LUSTRE/utils; then
export PATH=$LUSTRE/utils:$PATH
fi
rm -f $TMP/*active
}
+kernel_version() {
+ echo -n $((($1 << 16) | ($2 << 8) | $3))
+}
+
+export LINUX_VERSION=$(uname -r | sed -e "s/[-.]/ /3" -e "s/ .*//")
+export LINUX_VERSION_CODE=$(kernel_version ${LINUX_VERSION//\./ })
+
case `uname -r` in
2.4.*) EXT=".o"; USE_QUOTA=no; [ ! "$CLIENTONLY" ] && FSTYPE=ext3;;
*) EXT=".ko"; USE_QUOTA=yes;;
load_module lov/lov
load_module mgc/mgc
if ! client_only; then
- grep -q crc16 /proc/kallsyms || { modprobe crc16 2>/dev/null || true; }
- grep -q -w jbd /proc/kallsyms || { modprobe jbd 2>/dev/null || true; }
- grep -q -w jbd2 /proc/kallsyms || { modprobe jbd2 2>/dev/null || true; }
- [ "$FSTYPE" = "ldiskfs" ] && load_module ../ldiskfs/ldiskfs/ldiskfs
+ SYMLIST=/proc/kallsyms
+ grep -q crc16 $SYMLIST || { modprobe crc16 2>/dev/null || true; }
+ grep -q -w jbd $SYMLIST || { modprobe jbd 2>/dev/null || true; }
+ grep -q -w jbd2 $SYMLIST || { modprobe jbd2 2>/dev/null || true; }
+ if [ "$FSTYPE" = "ldiskfs" ]; then
+ grep -q exportfs_decode_fh $SYMLIST ||
+ { modprobe exportfs 2> /dev/null || true; }
+ load_module ../ldiskfs/ldiskfs/ldiskfs
+ fi
load_module mgs/mgs
load_module mds/mds
load_module mdd/mdd
set_debug_size () {
local dz=${1:-$DEBUG_SIZE}
- local cpus=$(getconf _NPROCESSORS_CONF)
+ local cpus=$(($(cut -d "-" -f 2 /sys/devices/system/cpu/possible)+1))
# bug 19944, adjust size to be -gt num_possible_cpus()
# promise 2MB for every cpu at least
set_default_debug () {
local debug=${1:-"$PTLDEBUG"}
- local subsystem_debug=${2:-"$SUBSYSTEM"}
+ local subsys=${2:-"$SUBSYSTEM"}
local debug_size=${3:-$DEBUG_SIZE}
- lctl set_param debug="$debug"
- lctl set_param subsystem_debug="${subsystem_debug# }"
+ [ -n "$debug" ] && lctl set_param debug="$debug" >/dev/null
+ [ -n "$subsys" ] && lctl set_param subsystem_debug="${subsys# }" >/dev/null
- set_debug_size $debug_size
- sync
+ [ -n "$debug_size" ] && set_debug_size $debug_size > /dev/null
}
set_default_debug_nodes () {
local nodes=$1
if [[ ,$nodes, = *,$HOSTNAME,* ]]; then
- nodes=$(exclude_items_from_list "$nodes" "$HOSTNAME")
- set_default_debug
+ nodes=$(exclude_items_from_list "$nodes" "$HOSTNAME")
+ set_default_debug
fi
[[ -n $nodes ]] && do_rpc_nodes $nodes set_default_debug \
- \\\"$PTLDEBUG\\\" \\\"$SUBSYSTEM\\\" $DEBUG_SIZE || true
+ \\\"$PTLDEBUG\\\" \\\"$SUBSYSTEM\\\" $DEBUG_SIZE || true
}
set_default_debug_facet () {
local mntpt=$(facet_mntpt $facet)
running=$(do_facet ${facet} "grep -c $mntpt' ' /proc/mounts") || true
if [ ${running} -ne 0 ]; then
- echo "Stopping $mntpt (opts:$@)"
+ echo "Stopping $mntpt (opts:$@) on $HOST"
do_facet ${facet} umount -d $@ $mntpt
fi
do_nodes $clients "running=\\\$(grep -c $mnt' ' /proc/mounts);
if [ \\\$running -ne 0 ] ; then
echo Stopping client \\\$(hostname) $mnt opts:$force;
-lsof -t $mnt || need_kill=no;
+lsof $mnt || need_kill=no;
if [ "x$force" != "x" -a "x\\\$need_kill" != "xno" ]; then
pids=\\\$(lsof -t $mnt | sort -u);
if [ -n \\\"\\\$pids\\\" ]; then
kill -9 \\\$pids;
fi
fi;
-busy=\\\$(umount $force $mnt 2>&1 | grep -c "busy");
-if [ \\\$busy -ne 0 ] ; then
+while umount $force $mnt 2>&1 | grep -q "busy"; do
echo "$mnt is still busy, wait one second" && sleep 1;
- umount $force $mnt;
-fi
+done;
fi"
}
BREAK_ON_ERROR=$BREAK_ON_ERROR \
END_RUN_FILE=$END_RUN_FILE \
LOAD_PID_FILE=$LOAD_PID_FILE \
- TESTSUITELOG=$TESTSUITELOG \
+ TESTLOG_PREFIX=$TESTLOG_PREFIX \
run_${load}.sh" &
- CLIENT_LOAD_PIDS="$CLIENT_LOAD_PIDS $!"
+ local ppid=$!
log "Started client load: ${load} on $client"
+ # get the children process IDs
+ local pids=$(ps --ppid $ppid -o pid= | xargs)
+ CLIENT_LOAD_PIDS="$CLIENT_LOAD_PIDS $ppid $pids"
return 0
}
local RESULT
local WAIT=0
- local sleep=5
+ local sleep=1
+ local print=10
while [ true ]; do
RESULT=$(do_node $node "$TEST")
if [ "$RESULT" == "$FINAL" ]; then
- echo "Updated after $WAIT sec: wanted '$FINAL' got '$RESULT'"
+ [ -z "$RESULT" -o $WAIT -le $sleep ] ||
+ echo "Updated after ${WAIT}s: wanted '$FINAL' got '$RESULT'"
return 0
fi
[ $WAIT -ge $MAX ] && break
- echo "Waiting $((MAX - WAIT)) secs for update"
+ [ $((WAIT % print)) -eq 0 ] &&
+ echo "Waiting $((MAX - WAIT)) secs for update"
WAIT=$((WAIT + sleep))
sleep $sleep
done
- echo "Update not seen after $MAX sec: wanted '$FINAL' got '$RESULT'"
+ echo "Update not seen after ${MAX}s: wanted '$FINAL' got '$RESULT'"
return 3
}
echo "Failing $facet on node $host"
+ # Make sure the client data is synced to disk. LU-924
+ #
+ # We don't write client data synchrnously (to avoid flooding sync writes
+ # when there are many clients connecting), so if the server reboots before
+ # the client data reachs disk, the client data will be lost and the client
+ # will be evicted after recovery, which is not what we expected.
+ do_facet $facet "sync; sync; sync"
+
local affected=$(affected_facets $facet)
shutdown_facet $facet
exit 1
}
+host_nids_address() {
+ local nodes=$1
+ local kind=$2
+
+ if [ -n "$kind" ]; then
+ nids=$(do_nodes $nodes "$LCTL list_nids | grep $kind | cut -f 1 -d '@'")
+ else
+ nids=$(do_nodes $nodes "$LCTL list_nids all | cut -f 1 -d '@'")
+ fi
+ echo $nids
+}
+
h2name_or_ip() {
if [ "$1" = "client" -o "$1" = "'*'" ]; then echo \'*\'; else
echo $1"@$2"
}
declare -fx h2o2ib
+# This enables variables in cfg/"setup".sh files to support the pdsh HOSTLIST
+# expressions format. As a bonus we can then just pass in those variables
+# to pdsh. What this function does is take a HOSTLIST type string and
+# expand it into a space deliminated list for us.
+hostlist_expand() {
+ local hostlist=$1
+ local offset=$2
+ local myList
+ local item
+ local list
+
+ [ -z "$hostlist" ] && return
+
+ # Translate the case of [..],..,[..] to [..] .. [..]
+ list="${hostlist/],/] }"
+ front=${list%%[*}
+ [[ "$front" == *,* ]] && {
+ new="${list%,*} "
+ old="${list%,*},"
+ list=${list/${old}/${new}}
+ }
+
+ for item in $list; do
+ # Test if we have any []'s at all
+ if [ "$item" != "${item/\[/}" ]; then {
+ # Expand the [*] into list
+ name=${item%%[*}
+ back=${item#*]}
+
+ if [ "$name" != "$item" ]; then
+ group=${item#$name[*}
+ group=${group%%]*}
+
+ for range in ${group//,/ }; do
+ begin=${range%-*}
+ end=${range#*-}
+
+ # Number of leading zeros
+ padlen=${#begin}
+ padlen2=${#end}
+ end=$(echo $end | sed 's/0*//')
+ [[ -z "$end" ]] && end=0
+ [[ $padlen2 -gt $padlen ]] && {
+ [[ $padlen2 -eq ${#end} ]] && padlen2=0
+ padlen=$padlen2
+ }
+ begin=$(echo $begin | sed 's/0*//')
+ [ -z $begin ] && begin=0
+
+ for num in $(seq -f "%0${padlen}g" $begin $end); do
+ value="${name#*,}${num}${back}"
+ [ "$value" != "${value/\[/}" ] && {
+ value=$(hostlist_expand "$value")
+ }
+ myList="$myList $value"
+ done
+ done
+ fi
+ } else {
+ myList="$myList $item"
+ } fi
+ done
+ myList=${myList//,/ }
+ myList=${myList:1} # Remove first character which is a space
+
+ # Filter any duplicates without sorting
+ list="$myList "
+ myList="${list%% *}"
+
+ while [[ "$list" != ${myList##* } ]]; do
+ list=${list//${list%% *} /}
+ myList="$myList ${list%% *}"
+ done
+ myList="${myList%* }";
+
+ # We can select an object at a offset in the list
+ [ $# -eq 2 ] && {
+ cnt=0
+ for item in $myList; do
+ let cnt=cnt+1
+ [ $cnt -eq $offset ] && {
+ myList=$item
+ }
+ done
+ [ $(get_node_count $myList) -ne 1 ] && myList=""
+ }
+ echo $myList
+}
+
facet_host() {
local facet=$1
return 0
}
+# verify a directory is shared among nodes.
+check_shared_dir() {
+ local dir=$1
+
+ [ -z "$dir" ] && return 1
+ do_rpc_nodes $(comma_list $(nodes_list)) check_logdir $dir
+ check_write_access $dir || return 1
+ return 0
+}
+
# Run e2fsck on MDT and OST(s) to generate databases used for lfsck.
generate_db() {
local i
local ostidx
local dev
- local tmp_file
- [ $MDSCOUNT -eq 1 ] || error "CMD is not supported"
- tmp_file=$(mktemp -p $SHARED_DIRECTORY ||
- error "fail to create file in $SHARED_DIRECTORY")
+ check_shared_dir $SHARED_DIRECTORY ||
+ error "$SHARED_DIRECTORY isn't a shared directory"
- # make sure everything gets to the backing store
- local list=$(comma_list $CLIENTS $(facet_host $SINGLEMDS) $(osts_nodes))
- do_nodes $list "sync; sleep 2; sync"
+ export MDSDB=$SHARED_DIRECTORY/mdsdb
+ export OSTDB=$SHARED_DIRECTORY/ostdb
- do_nodes $list ls $tmp_file || \
- error "$SHARED_DIRECTORY is not a shared directory"
- rm $tmp_file
+ [ $MDSCOUNT -eq 1 ] || error "CMD is not supported"
run_e2fsck $(mdts_nodes) $MDTDEV "--mdsdb $MDSDB"
# We need to dump the logs on all nodes
if $dump; then
- gather_logs $(comma_list $(nodes_list))
+ gather_logs $(comma_list $(nodes_list)) 0
fi
debugrestore
echo
log_sub_test_begin test_${1}
- (run_one $1 "$2") 2>&1 | tee $test_log
+ (run_one $1 "$2") 2>&1 | tee -i $test_log
local RC=${PIPESTATUS[0]}
[ $RC -ne 0 ] && [ ! -f $LOGDIR/err ] && \
ostuuid_from_index()
{
- $LFS osts $2 | awk '/^'$1'/ { print $2 }'
+ $LFS osts $2 | sed -ne "/^$1: /s/.* \(.*\) .*$/\1/p"
+}
+
+ostname_from_index() {
+ local uuid=$(ostuuid_from_index $1)
+ echo ${uuid/_UUID/}
+}
+
+index_from_ostuuid()
+{
+ $LFS osts $2 | sed -ne "/${1}/s/\(.*\): .* .*$/\1/p"
+}
+
+mdtuuid_from_index()
+{
+ $LFS mdts $2 | awk '/^'$1'/ { print $2 }'
}
remote_node () {
init_clients_lists () {
# Sanity check: exclude the local client from RCLIENTS
- local rclients=$(echo " $RCLIENTS " | sed -re "s/\s+$HOSTNAME\s+/ /g")
+ local clients=$(hostlist_expand "$RCLIENTS")
+ local rclients=$(exclude_items_from_list "$clients" $HOSTNAME)
# Sanity check: exclude the dup entries
- rclients=$(for i in $rclients; do echo $i; done | sort -u)
+ RCLIENTS=$(for i in ${rclients//,/ }; do echo $i; done | sort -u)
- local clients="$SINGLECLIENT $HOSTNAME $rclients"
+ clients="$SINGLECLIENT $HOSTNAME $RCLIENTS"
# Sanity check: exclude the dup entries from CLIENTS
# for those configs which has SINGLCLIENT set to local client
clients=$(for i in $clients; do echo $i; done | sort -u)
- CLIENTS=`comma_list $clients`
- local -a remoteclients=($rclients)
+ CLIENTS=$(comma_list $clients)
+ local -a remoteclients=($RCLIENTS)
for ((i=0; $i<${#remoteclients[@]}; i++)); do
varname=CLIENT$((i + 2))
eval $varname=${remoteclients[i]}
done
}
-# One client request could be timeouted because server was not ready
+# One client request could be timed out because server was not ready
# when request was sent by client.
# The request timeout calculation details :
# ptl_send_rpc ()
local maxtime=$(( 2 * $(request_timeout $facet)))
if ! do_rpc_nodes $(facet_host $facet) \
-_wait_import_state $expected $param $maxtime; then
+ _wait_import_state $expected $param $maxtime; then
error "import is not in ${expected} state"
return 1
fi
gather_logs () {
local list=$1
+ local tar_logs=$2
local ts=$(date +%s)
local docp=true
# dump lustre logs, dmesg
- prefix="$LOGDIR/${TESTSUITE}.${TESTNAME}"
+ prefix="$TESTLOG_PREFIX.$TESTNAME"
suffix="$ts.log"
echo "Dumping lctl log to ${prefix}.*.${suffix}"
do_nodes $list rsync -az "${prefix}.*.${suffix}" $HOSTNAME:$LOGDIR
fi
- local archive=$LOGDIR/${TESTSUITE}-$ts.tar.bz2
- tar -jcf $archive $LOGDIR/*$ts* $LOGDIR/*${TESTSUITE}*
-
- echo $archive
-}
-
-cleanup_logs () {
- local list=${1:-$(comma_list $(nodes_list))}
+ if [ $tar_logs == 1 ]; then
+ local archive=$LOGDIR/${TESTSUITE}-$ts.tar.bz2
+ tar -jcf $archive $LOGDIR/*$ts* $LOGDIR/*${TESTSUITE}*
- [ -n ${TESTSUITE} ] && do_nodes $list "rm -f $TMP/*${TESTSUITE}*" || true
+ echo $archive
+ fi
}
do_ls () {
echo -n "checking $dir..."
res=$(do_check_flavor $dir $flavor)
echo "found $res/$expect $flavor connections"
- [ $res -eq $expect ] && return 0
+ [ $res -ge $expect ] && return 0
sleep 4
done
return 1
fi
done
+ rm -f $dir/node.*.yml
return 0
}
mkdir -p $LOGDIR
init_clients_lists
- do_rpc_nodes $(comma_list $(nodes_list)) check_logdir $LOGDIR
- if check_write_access $LOGDIR; then
- touch $LOGDIR/shared
- echo "Logging to shared log directory: $LOGDIR"
- else
- echo "Logging to local directory: $LOGDIR"
- fi
+ if [ ! -f $YAML_LOG ]; then # If the yaml log already exists then we will just append to it
+ if check_shared_dir $LOGDIR; then
+ touch $LOGDIR/shared
+ echo "Logging to shared log directory: $LOGDIR"
+ else
+ echo "Logging to local directory: $LOGDIR"
+ fi
- yml_nodes_file $LOGDIR >> $YAML_LOG
- yml_results_file >> $YAML_LOG
+ yml_nodes_file $LOGDIR >> $YAML_LOG
+ yml_results_file >> $YAML_LOG
+ fi
umask $SAVE_UMASK
}
local mdtdev=$2
shift 2
local files="$@"
- local mntpt=${MOUNT%/*}/$facet
+ local mntpt=$(facet_mntpt $facet)
echo "removing files from $mdtdev on $facet: $files"
mount -t $FSTYPE $MDS_MOUNT_OPTS $mdtdev $mntpt || return $?
local mdtdev=$2
shift 2
local files="$@"
- local mntpt=${MOUNT%/*}/$facet
+ local mntpt=$(facet_mntpt $facet)
echo "duplicating files on $mdtdev on $facet: $files"
mkdir -p $mntpt || return $?
min_ost_size () {
$LCTL get_param -n osc.*.kbytesavail | sort -n | head -n1
}
+
+# Get the block size of the filesystem.
+get_block_size() {
+ local facet=$1
+ local device=$2
+ local size
+
+ size=$(do_facet $facet "$DUMPE2FS -h $device 2>&1" |
+ awk '/^Block size:/ {print $3}')
+ echo $size
+}
+
+# Check whether the "large_xattr" feature is enabled or not.
+large_xattr_enabled() {
+ local mds_dev=$(mdsdevname ${SINGLEMDS//mds/})
+
+ do_facet $SINGLEMDS "$DUMPE2FS -h $mds_dev 2>&1 | grep -q large_xattr"
+ return ${PIPESTATUS[0]}
+}
+
+# Get the maximum xattr size supported by the filesystem.
+max_xattr_size() {
+ local size
+
+ if large_xattr_enabled; then
+ # include/linux/limits.h: #define XATTR_SIZE_MAX 65536
+ size=65536
+ else
+ local mds_dev=$(mdsdevname ${SINGLEMDS//mds/})
+ local block_size=$(get_block_size $SINGLEMDS $mds_dev)
+
+ # maximum xattr size = size of block - size of header -
+ # size of 1 entry - 4 null bytes
+ size=$((block_size - 32 - 32 - 4))
+ fi
+
+ echo $size
+}
+
+# Dump the value of the named xattr from a file.
+get_xattr_value() {
+ local xattr_name=$1
+ local file=$2
+
+ echo "$(getfattr -n $xattr_name --absolute-names --only-values $file)"
+}
+
+# Generate a string with size of $size bytes.
+generate_string() {
+ local size=${1:-1024} # in bytes
+
+ echo "$(head -c $size < /dev/zero | tr '\0' y)"
+}
git://git.whamcloud.com / fs / lustre-release.git / blobdiff