export PARALLEL=${PARALLEL:-"no"}
TRACE=${TRACE:-""}
-LUSTRE_TESTS_API_DIR=${LUSTRE_TESTS_API_DIR:-${LUSTRE}/tests/clientapi}
LUSTRE=${LUSTRE:-$(dirname $0)/..}
+LUSTRE_TESTS_API_DIR=${LUSTRE_TESTS_API_DIR:-${LUSTRE}/tests/clientapi}
. $LUSTRE/tests/test-framework.sh
init_test_env $@
ALWAYS_EXCEPT="$SANITY_EXCEPT "
# bug number for skipped test: LU-9693 LU-6493 LU-9693
ALWAYS_EXCEPT+=" 42a 42b 42c "
-# bug number: LU-8411 LU-9054
-ALWAYS_EXCEPT+=" 407 312 "
+# bug number: LU-8411 LU-9054 LU-13314
+ALWAYS_EXCEPT+=" 407 312 56ob"
if $SHARED_KEY; then
# bug number: LU-9795 LU-9795 LU-9795 LU-9795
- ALWAYS_EXCEPT+=" 17n 60a 133g 300f "
+ ALWAYS_EXCEPT+=" 17n 60a 133g 300f"
+fi
+
+selinux_status=$(getenforce)
+if [ "$selinux_status" != "Disabled" ]; then
+ # bug number:
+ ALWAYS_EXCEPT+=""
fi
# skip the grant tests for ARM until they are fixed
if [[ $(uname -m) = aarch64 ]]; then
# bug number: LU-11596
ALWAYS_EXCEPT+=" $GRANT_CHECK_LIST"
- # bug number: LU-11671 LU-11667 LU-4398
- ALWAYS_EXCEPT+=" 45 317 817"
+ # bug number: LU-11671 LU-11667
+ ALWAYS_EXCEPT+=" 45 317"
+fi
+
+# skip nfs tests on kernels >= 4.14.0 until they are fixed
+if [ $LINUX_VERSION_CODE -ge $(version_code 4.14.0) ]; then
+ # bug number: LU-12661
+ ALWAYS_EXCEPT+=" 817"
+fi
+# skip cgroup tests on RHEL8.1 kernels until they are fixed
+if (( $LINUX_VERSION_CODE >= $(version_code 4.18.0) &&
+ $LINUX_VERSION_CODE < $(version_code 5.4.0) )); then
+ # bug number: LU-13063
+ ALWAYS_EXCEPT+=" 411"
fi
-# 5 12 (min)"
-[ "$SLOW" = "no" ] && EXCEPT_SLOW="27m 64b 68 71 115 300o"
+# 5 12 8 12 (min)"
+[ "$SLOW" = "no" ] && EXCEPT_SLOW="27m 64b 68 71 115 135 136 300o"
if [ "$mds1_FSTYPE" = "zfs" ]; then
# bug number for skipped test:
[ $MDS1_VERSION -le $(version_code 2.3.55) ] &&
TESTS="4094 4095"
- # skip long symlink name for rhel6.5.
- # rhel6.5 has a limit (PATH_MAX - sizeof(struct filename))
- grep -q '6.5' /etc/redhat-release &>/dev/null &&
- TESTS="59 60 61 4062 4063"
-
for i in $TESTS; do
local SYMNAME=$(str_repeat 'x' $i)
ln -s $SYMNAME $DIR/$tdir/f$i || error "failed $i-char symlink"
}
run_test 27ce "more stripes than OSTs with -o"
+test_27cf() {
+ local osp_proc="osp.$FSNAME-OST0000-osc-MDT000*.active"
+ local pid=0
+
+ test_mkdir -p $DIR/$tdir || error "failed to mkdir $DIR/$tdir"
+ do_facet $SINGLEMDS "$LCTL set_param -n $osp_proc=0"
+ stack_trap "do_facet $SINGLEMDS $LCTL set_param -n $osp_proc=1" EXIT
+ wait_update_facet $SINGLEMDS "$LCTL get_param -n $osp_proc | grep 1" ||
+ error "failed to set $osp_proc=0"
+
+ $LFS setstripe -o 0 $DIR/$tdir/$tfile &
+ pid=$!
+ sleep 1
+ do_facet $SINGLEMDS "$LCTL set_param -n $osp_proc=1"
+ wait_update_facet $SINGLEMDS "$LCTL get_param -n $osp_proc | grep 0" ||
+ error "failed to set $osp_proc=1"
+ wait $pid
+ [[ $pid -ne 0 ]] ||
+ error "should return error due to $osp_proc=0"
+}
+run_test 27cf "'setstripe -o' on inactive OSTs should return error"
+
test_27d() {
test_mkdir $DIR/$tdir
$LFS setstripe -c 0 -i -1 -S 0 $DIR/$tdir/$tfile ||
run_test 27I "check that root dir striping does not break parent dir one"
test_27J() {
- [[ $(lustre_version_code $SINGLEMDS) -le $(version_code 2.12.51) ]] &&
+ [[ $MDS1_VERSION -le $(version_code 2.12.51) ]] &&
skip "Need MDS version newer than 2.12.51"
test_mkdir $DIR/$tdir
run_test 27J "basic ops on file with foreign LOV"
test_27K() {
- [[ $(lustre_version_code $SINGLEMDS) -le $(version_code 2.12.49) ]] &&
+ [[ $MDS1_VERSION -le $(version_code 2.12.49) ]] &&
skip "Need MDS version newer than 2.12.49"
test_mkdir $DIR/$tdir
}
run_test 33g "nonroot user create already existing root created file"
+test_33h() {
+ [ $MDSCOUNT -lt 2 ] && skip_env "needs >= 2 MDTs"
+ [ $MDS1_VERSION -lt $(version_code 2.13.50) ] &&
+ skip "Need MDS version at least 2.13.50"
+
+ test_mkdir -c $MDSCOUNT -H crush $DIR/$tdir ||
+ error "mkdir $tdir failed"
+ touch $DIR/$tdir/$tfile || error "touch $tfile failed"
+
+ local index=$($LFS getstripe -m $DIR/$tdir/$tfile)
+ local index2
+
+ for fname in $DIR/$tdir/$tfile.bak \
+ $DIR/$tdir/$tfile.SAV \
+ $DIR/$tdir/$tfile.orig \
+ $DIR/$tdir/$tfile~; do
+ touch $fname || error "touch $fname failed"
+ index2=$($LFS getstripe -m $fname)
+ [ $index -eq $index2 ] ||
+ error "$fname MDT index mismatch $index != $index2"
+ done
+
+ local failed=0
+ for i in {1..50}; do
+ for fname in $(mktemp -u $DIR/$tdir/.$tfile.XXXXXX) \
+ $(mktemp $DIR/$tdir/$tfile.XXXXXXXX); do
+ touch $fname || error "touch $fname failed"
+ index2=$($LFS getstripe -m $fname)
+ if [[ $index != $index2 ]]; then
+ failed=$((failed + 1))
+ echo "$fname MDT index mismatch $index != $index2"
+ fi
+ done
+ done
+ echo "$failed MDT index mismatches"
+ (( failed < 4 )) || error "MDT index mismatch $failed times"
+
+}
+run_test 33h "temp file is located on the same MDT as target"
+
TEST_34_SIZE=${TEST_34_SIZE:-2000000000000}
test_34a() {
rm -f $DIR/f34
}
run_test 39p "remote directory cached attributes updated after create ========"
+test_39r() {
+ [ $OST1_VERSION -ge $(version_code 2.13.52) ] ||
+ skip "no atime update on old OST"
+ if [ "$ost1_FSTYPE" != ldiskfs ]; then
+ skip_env "ldiskfs only test"
+ fi
+
+ local saved_adiff
+ saved_adiff=$(do_facet ost1 \
+ lctl get_param -n obdfilter.*OST0000.atime_diff)
+ stack_trap "do_facet ost1 \
+ lctl set_param obdfilter.*.atime_diff=$saved_adiff"
+
+ do_facet ost1 "lctl set_param obdfilter.*.atime_diff=5"
+
+ $LFS setstripe -i 0 $DIR/$tfile
+ dd if=/dev/zero of=$DIR/$tfile bs=4k count=1 ||
+ error "can't write initial file"
+ cancel_lru_locks osc
+
+ # exceed atime_diff and access file
+ sleep 6
+ dd if=$DIR/$tfile of=/dev/null || error "can't udpate atime"
+
+ local atime_cli=$(stat -c %X $DIR/$tfile)
+ echo "client atime: $atime_cli"
+ # allow atime update to be written to device
+ do_facet ost1 "$LCTL set_param -n osd*.*OST*.force_sync 1"
+ sleep 5
+
+ local ostdev=$(ostdevname 1)
+ local fid=($(lfs getstripe -y $DIR/$tfile |
+ awk '/l_fid:/ { print $2 }' | tr ':' ' '))
+ local objpath="O/0/d$((${fid[1]} % 32))/$((${fid[1]}))"
+ local cmd="debugfs -c -R \\\"stat $objpath\\\" $ostdev"
+
+ echo "OST atime: $(do_facet ost1 "$cmd" |& grep atime)"
+ local atime_ost=$(do_facet ost1 "$cmd" |&
+ awk -F'[: ]' '/atime:/ { print $4 }')
+ (( atime_cli == atime_ost )) ||
+ error "atime on client $atime_cli != ost $atime_ost"
+}
+run_test 39r "lazy atime update on OST"
test_39q() { # LU-8041
local testdir=$DIR/$tdir
[ $PARALLEL == "yes" ] && skip "skip parallel run"
remote_ost_nodsh && skip "remote OST with nodsh"
- # get ost1 size - lustre-OST0000
+ # get ost1 size - $FSNAME-OST0000
ost1_size=$(do_facet ost1 $LFS df | grep ${ost1_svc} |
awk '{ print $4 }')
# write 800M at maximum
}
run_test 56ob "check lfs find -atime -mtime -ctime with units"
+test_newerXY_base() {
+ local x=$1
+ local y=$2
+ local dir=$DIR/$tdir
+ local ref
+ local negref
+
+ if [ $y == "t" ]; then
+ ref="\"$(date +"%Y-%m-%d %H:%M:%S")\""
+ else
+ ref=$DIR/$tfile.newer
+ touch $ref || error "touch $ref failed"
+ fi
+ sleep 2
+ setup_56 $dir $NUMFILES $NUMDIRS "-i0 -c1" "-i0 -c1"
+ sleep 2
+ if [ $y == "t" ]; then
+ negref="\"$(date +"%Y-%m-%d %H:%M:%S")\""
+ else
+ negref=$DIR/$tfile.newerneg
+ touch $negref || error "touch $negref failed"
+ fi
+
+ local cmd="$LFS find $dir -newer$x$y $ref"
+ local nums=$(eval $cmd | wc -l)
+ local expected=$(((NUMFILES + 2) * NUMDIRS + 1))
+
+ [ $nums -eq $expected ] ||
+ error "'$cmd' wrong: found $nums, expected $expected"
+
+ cmd="$LFS find $dir ! -newer$x$y $negref"
+ nums=$(eval $cmd | wc -l)
+ [ $nums -eq $expected ] ||
+ error "'$cmd' wrong: found $nums, expected $expected"
+
+ cmd="$LFS find $dir -newer$x$y $ref ! -newer$x$y $negref"
+ nums=$(eval $cmd | wc -l)
+ [ $nums -eq $expected ] ||
+ error "'$cmd' wrong: found $nums, expected $expected"
+
+ rm -rf $DIR/*
+}
+
+test_56oc() {
+ test_newerXY_base "a" "a"
+ test_newerXY_base "a" "m"
+ test_newerXY_base "a" "c"
+ test_newerXY_base "m" "a"
+ test_newerXY_base "m" "m"
+ test_newerXY_base "m" "c"
+ test_newerXY_base "c" "a"
+ test_newerXY_base "c" "m"
+ test_newerXY_base "c" "c"
+ test_newerXY_base "a" "t"
+ test_newerXY_base "m" "t"
+ test_newerXY_base "c" "t"
+}
+run_test 56oc "check lfs find -newerXY work"
+
test_56p() {
[ $RUNAS_ID -eq $UID ] &&
skip_env "RUNAS_ID = UID = $UID -- skipping"
}
run_test 56ra "check lfs find -size -lazy works for data on OSTs"
+test_56rb() {
+ local dir=$DIR/$tdir
+ local tmp=$TMP/$tfile.log
+ local mdt_idx;
+
+ test_mkdir -p $dir || error "failed to mkdir $dir"
+ $LFS setstripe -c 1 -i 0 $dir/$tfile ||
+ error "failed to setstripe $dir/$tfile"
+ dd if=/dev/zero of=$dir/$tfile bs=1M count=1
+
+ stack_trap "rm -f $tmp" EXIT
+ $LFS find --size +100K --ost 0 $dir 2>&1 | tee $tmp
+ [ -z "$(cat $tmp | grep "obd_uuid: ")" ] ||
+ error "failed to find --size +100K --ost 0 $dir"
+ $LFS find --size +100K --mdt $mdt_idx $dir 2>&1 | tee $tmp
+ [ -z "$(cat $tmp | grep "obd_uuid: ")" ] ||
+ error "failed to find --size +100K --mdt $mdt_idx $dir"
+}
+run_test 56rb "check lfs find --size --ost/--mdt works"
+
test_56s() { # LU-611 #LU-9369
[[ $OSTCOUNT -lt 2 ]] && skip_env "need at least 2 OSTs"
echo "done."
echo -n "Removing test file from pool '$pool'..."
- $LFS migrate $file1 &> /dev/null ||
+ # "lfs migrate $file" won't remove the file from the pool
+ # until some striping information is changed.
+ $LFS migrate -c 1 $file1 &> /dev/null ||
error "cannot remove from pool"
[ "$($LFS getstripe -p $file1)" ] &&
error "pool still set"
test_56wc() {
local file1="$DIR/$tdir/file1"
+ local parent_ssize
+ local parent_scount
+ local cur_ssize
+ local cur_scount
+ local orig_ssize
echo -n "Creating test dir..."
test_mkdir $DIR/$tdir &> /dev/null || error "cannot create dir"
- local def_stripe_size=$($LFS getstripe -S $DIR/$tdir 2>/dev/null)
$LFS setstripe -S 1M -c 1 "$DIR/$tdir" &> /dev/null ||
- error "cannot set stripe"
+ error "cannot set stripe by '-S 1M -c 1'"
echo "done"
echo -n "Setting initial stripe for test file..."
$LFS setstripe -S 512K -c 1 "$file1" &> /dev/null ||
error "cannot set stripe"
- [ $($LFS getstripe -S "$file1") -eq 524288 ] ||
- error "stripe size not set"
+ cur_ssize=$($LFS getstripe -S "$file1")
+ [ $cur_ssize -eq 524288 ] || error "setstripe -S $cur_ssize != 524288"
echo "done."
# File currently set to -S 512K -c 1
echo -n "Verifying -S option is passed through to lfs migrate..."
$LFS_MIGRATE -y -S 1M "$file1" &> /dev/null ||
error "migration failed"
- [ $($LFS getstripe -S "$file1") -eq 1048576 ] ||
- error "file was not restriped"
+ cur_ssize=$($LFS getstripe -S "$file1")
+ [ $cur_ssize -eq 1048576 ] || error "migrate -S $cur_ssize != 1048576"
echo "done."
# File currently set to -S 1M -c 1
error "long option without argument not supported"
$LFS_MIGRATE -y --stripe-size 512K "$file1" &> /dev/null ||
error "long option with argument not supported"
- [ $($LFS getstripe -S "$file1") -eq 524288 ] ||
- error "file not restriped with --stripe-size option"
+ cur_ssize=$($LFS getstripe -S "$file1")
+ [ $cur_ssize -eq 524288 ] ||
+ error "migrate --stripe-size $cur_ssize != 524288"
echo "done."
# File currently set to -S 512K -c 1
echo -n "Verifying explicit stripe count can be set..."
$LFS_MIGRATE -y -c 2 "$file1" &> /dev/null ||
error "migrate failed"
- [ $($LFS getstripe -c "$file1") -eq 2 ] ||
- error "file not restriped to explicit count"
+ cur_scount=$($LFS getstripe -c "$file1")
+ [ $cur_scount -eq 2 ] || error "migrate -c $cur_scount != 2"
echo "done."
fi
# Ensure parent striping is used if -R is set, and no stripe
# count or size is specified
echo -n "Setting stripe for parent directory..."
- $LFS setstripe -S 1M -c 1 "$DIR/$tdir" &> /dev/null ||
- error "cannot set stripe"
+ $LFS setstripe -S 2M -c 1 "$DIR/$tdir" &> /dev/null ||
+ error "cannot set stripe '-S 2M -c 1'"
echo "done."
echo -n "Verifying restripe option uses parent stripe settings..."
+ parent_ssize=$($LFS getstripe -S $DIR/$tdir 2>/dev/null)
+ parent_scount=$($LFS getstripe -c $DIR/$tdir 2>/dev/null)
$LFS_MIGRATE -y -R "$file1" &> /dev/null ||
error "migrate failed"
- [ $($LFS getstripe -S "$file1") -eq $def_stripe_size ] ||
- error "file not restriped to parent settings"
- [ $($LFS getstripe -c "$file1") -eq 1 ] ||
- error "file not restriped to parent settings"
+ cur_ssize=$($LFS getstripe -S "$file1")
+ [ $cur_ssize -eq $parent_ssize ] ||
+ error "migrate -R stripe_size $cur_ssize != $parent_ssize"
+ cur_scount=$($LFS getstripe -c "$file1")
+ [ $cur_scount -eq $parent_scount ] ||
+ error "migrate -R stripe_count $cur_scount != $parent_scount"
echo "done."
# File currently set to -S 1M -c 1
# Ensure striping is preserved if -R is not set, and no stripe
# count or size is specified
echo -n "Verifying striping size preserved when not specified..."
- local orig_stripe_size=$($LFS getstripe -S "$file1" 2>/dev/null)
+ orig_ssize=$($LFS getstripe -S "$file1" 2>/dev/null)
$LFS setstripe -S 2M -c 1 "$DIR/$tdir" &> /dev/null ||
error "cannot set stripe on parent directory"
$LFS_MIGRATE -y "$file1" &> /dev/null ||
error "migrate failed"
- [ $($LFS getstripe -S "$file1") -eq $orig_stripe_size ] ||
- error "file was restriped"
+ cur_ssize=$($LFS getstripe -S "$file1")
+ [ $cur_ssize -eq $orig_ssize ] ||
+ error "migrate by default $cur_ssize != $orig_ssize"
echo "done."
# Ensure file name properly detected when final option has no argument
}
run_test 56xc "lfs migration autostripe"
+test_56xd() {
+ [[ $OSTCOUNT -lt 2 ]] && skip_env "needs >= 2 OSTs"
+
+ local dir=$DIR/$tdir
+ local f_mgrt=$dir/$tfile.mgrt
+ local f_yaml=$dir/$tfile.yaml
+ local f_copy=$dir/$tfile.copy
+ local layout_yaml="-E 1M -S 512K -c 1 -E -1 -S 1M -c 2 -i 0"
+ local layout_copy="-c 2 -S 2M -i 1"
+ local yamlfile=$dir/yamlfile
+ local layout_before;
+ local layout_after;
+
+ test_mkdir "$dir" || error "cannot create dir $dir"
+ $LFS setstripe $layout_yaml $f_yaml ||
+ error "cannot setstripe $f_yaml with layout $layout_yaml"
+ $LFS getstripe --yaml $f_yaml > $yamlfile
+ $LFS setstripe $layout_copy $f_copy ||
+ error "cannot setstripe $f_copy with layout $layout_copy"
+ touch $f_mgrt
+ dd if=/dev/zero of=$f_mgrt bs=1M count=4
+
+ # 1. test option --yaml
+ $LFS_MIGRATE -y --yaml $yamlfile $f_mgrt ||
+ error "cannot migrate $f_mgrt with --yaml $yamlfile"
+ layout_before=$(get_layout_param $f_yaml)
+ layout_after=$(get_layout_param $f_mgrt)
+ [ "$layout_after" == "$layout_before" ] ||
+ error "lfs_migrate --yaml: $layout_after != $layout_before"
+
+ # 2. test option --copy
+ $LFS_MIGRATE -y --copy $f_copy $f_mgrt ||
+ error "cannot migrate $f_mgrt with --copy $f_copy"
+ layout_before=$(get_layout_param $f_copy)
+ layout_after=$(get_layout_param $f_mgrt)
+ [ "$layout_after" == "$layout_before" ] ||
+ error "lfs_migrate --copy: $layout_after != $layout_before"
+}
+run_test 56xd "check lfs_migrate --yaml and --copy support"
+
+test_56xe() {
+ [[ $OSTCOUNT -lt 2 ]] && skip_env "needs >= 2 OSTs"
+
+ local dir=$DIR/$tdir
+ local f_comp=$dir/$tfile
+ local layout="-E 1M -S 512K -c 1 -E -1 -S 1M -c 2 -i 0"
+ local layout_before=""
+ local layout_after=""
+
+ test_mkdir "$dir" || error "cannot create dir $dir"
+ $LFS setstripe $layout $f_comp ||
+ error "cannot setstripe $f_comp with layout $layout"
+ layout_before=$(get_layout_param $f_comp)
+ dd if=/dev/zero of=$f_comp bs=1M count=4
+
+ # 1. migrate a comp layout file by lfs_migrate
+ $LFS_MIGRATE -y $f_comp || error "cannot migrate $f_comp by lfs_migrate"
+ layout_after=$(get_layout_param $f_comp)
+ [ "$layout_before" == "$layout_after" ] ||
+ error "lfs_migrate: $layout_before != $layout_after"
+
+ # 2. migrate a comp layout file by lfs migrate
+ $LFS migrate $f_comp || error "cannot migrate $f_comp by lfs migrate"
+ layout_after=$(get_layout_param $f_comp)
+ [ "$layout_before" == "$layout_after" ] ||
+ error "lfs migrate: $layout_before != $layout_after"
+}
+run_test 56xe "migrate a composite layout file"
+
test_56y() {
[ $MDS1_VERSION -lt $(version_code 2.4.53) ] &&
skip "No HSM $(lustre_build_version $SINGLEMDS) MDS < 2.4.53"
test_64a () {
[ $PARALLEL == "yes" ] && skip "skip parallel run"
- df $DIR
- lctl get_param -n osc.*[oO][sS][cC][_-]*.cur* | grep "[0-9]"
+ lfs df $DIR
+ lctl get_param osc.*[oO][sS][cC][_-]*.cur* | grep "=[1-9]"
}
run_test 64a "verify filter grant calculations (in kernel) ====="
test_65n() {
[ -n "$FILESET" ] && skip "Not functional for FILESET set"
- [[ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.12.50) ]] ||
+ [[ $MDS1_VERSION -ge $(version_code 2.12.50) ]] ||
skip "Need MDS version at least 2.12.50"
[[ $PARALLEL != "yes" ]] || skip "skip parallel run"
local dir3=$MOUNT/$tdir-3
mkdir $dir3 || error "mkdir $dir3 failed"
- ! getfattr -n trusted.lov $dir3 &> /dev/null ||
- error "$dir3 shouldn't have LOV EA"
+ # $dir3 shouldn't have LOV EA, but "lfs getstripe -d $dir3" should show
+ # the root layout, which is the actual default layout that will be used
+ # when new files are created in $dir3.
+ local dir3_layout=$(get_layout_param $dir3)
+ local root_dir_layout=$(get_layout_param $MOUNT)
+ [[ "$dir3_layout" = "$root_dir_layout" ]] ||
+ error "$dir3 should show the default layout from $MOUNT"
# set OST pool on root directory
local pool=$TESTNAME
local dir4=$MOUNT/$tdir-4
mkdir $dir4 || error "mkdir $dir4 failed"
- ! getfattr -n trusted.lov $dir4 &> /dev/null ||
- error "$dir4 shouldn't have LOV EA"
+ local dir4_layout=$(get_layout_param $dir4)
+ root_dir_layout=$(get_layout_param $MOUNT)
+ echo "$LFS getstripe -d $dir4"
+ $LFS getstripe -d $dir4
+ echo "$LFS getstripe -d $MOUNT"
+ $LFS getstripe -d $MOUNT
+ [[ "$dir4_layout" = "$root_dir_layout" ]] ||
+ error "$dir4 should show the default layout from $MOUNT"
# new file created in $dir4 should inherit the pool from
# the filesystem default
local dir5=$dir4/$tdir-5
mkdir $dir5 || error "mkdir $dir5 failed"
- local dir4_layout=$(get_layout_param $dir4)
+ dir4_layout=$(get_layout_param $dir4)
local dir5_layout=$(get_layout_param $dir5)
[[ "$dir4_layout" = "$dir5_layout" ]] ||
error "$dir5 should inherit the default layout from $dir4"
test_76() { # Now for bug 20433, added originally in bug 1443
[ $PARALLEL == "yes" ] && skip "skip parallel run"
- local CPUS=$(getconf _NPROCESSORS_ONLN 2>/dev/null)
-
cancel_lru_locks osc
- BEFORE_INODES=$(num_inodes)
- echo "before inodes: $BEFORE_INODES"
- local COUNT=1000
- [ "$SLOW" = "no" ] && COUNT=100
- for i in $(seq $COUNT); do
+ local cpus=$(getconf _NPROCESSORS_ONLN 2>/dev/null)
+ local before=$(num_inodes)
+ local count=$((512 * cpus))
+ [ "$SLOW" = "no" ] && count=$((64 * cpus))
+
+ echo "before inodes: $before"
+ for i in $(seq $count); do
touch $DIR/$tfile
rm -f $DIR/$tfile
done
cancel_lru_locks osc
- AFTER_INODES=$(num_inodes)
- echo "after inodes: $AFTER_INODES"
- local wait=0
- while [[ $((AFTER_INODES-1*${CPUS:-1})) -gt $BEFORE_INODES ]]; do
- sleep 2
- AFTER_INODES=$(num_inodes)
- wait=$((wait+2))
- echo "wait $wait seconds inodes: $AFTER_INODES"
- if [ $wait -gt 30 ]; then
- error "inode slab grew from $BEFORE_INODES to $AFTER_INODES"
+ local after=$(num_inodes)
+ echo "after inodes: $after"
+ while (( after > before + 8 * ${cpus:-1} )); do
+ sleep 1
+ after=$(num_inodes)
+ wait=$((wait + 1))
+ (( wait % 5 == 0 )) && echo "wait $wait seconds inodes: $after"
+ if (( wait > 30 )); then
+ error "inode slab grew from $before to $after"
fi
done
}
[ $PARALLEL == "yes" ] && skip "skip parallel run"
# relax strong synchronous semantics for slow backends like ZFS
- local soc="obdfilter.*.sync_on_lock_cancel"
- local soc_old=$(do_facet ost1 lctl get_param -n $soc | head -n1)
- local hosts=
- if [ "$soc_old" != "never" ] &&
- [ "$ost1_FSTYPE" != "ldiskfs" ]; then
- hosts=$(for host in $(seq -f "ost%g" 1 $OSTCOUNT); do
- facet_active_host $host; done | sort -u)
- do_nodes $hosts lctl set_param $soc=never
- fi
+ if [ "$ost1_FSTYPE" != "ldiskfs" ]; then
+ local soc="obdfilter.*.sync_lock_cancel"
+ local save=$(do_facet ost1 $LCTL get_param -n $soc | head -n1)
- dd if=/dev/zero of=$DIR/$tfile bs=1M count=1 seek=1M
- sync; sleep 1; sync
- local BEFORE=`date +%s`
- cancel_lru_locks osc
- local AFTER=`date +%s`
- local DIFF=$((AFTER-BEFORE))
- if [ $DIFF -gt 1 ] ; then
- error "elapsed for 1M@1T = $DIFF"
- fi
+ # "sync_on_lock_cancel" was broken by v2_11_55_0-26-g7059644e9a
+ if [ -z "$save" ]; then
+ soc="obdfilter.*.sync_on_lock_cancel"
+ save=$(do_facet ost1 $LCTL get_param -n $soc | head -n1)
+ fi
- [ -n "$hosts" ] && do_nodes $hosts lctl set_param $soc=$soc_old
+ if [ "$save" != "never" ]; then
+ local hosts=$(comma_list $(osts_nodes))
- rm -f $DIR/$tfile
+ do_nodes $hosts $LCTL set_param $soc=never
+ stack_trap "do_nodes $hosts $LCTL set_param $soc=$save"
+ fi
+ fi
+
+ dd if=/dev/zero of=$DIR/$tfile bs=1M count=1 seek=1M
+ sync; sleep 1; sync
+ local before=$(date +%s)
+ cancel_lru_locks osc
+ local after=$(date +%s)
+ local diff=$((after - before))
+ [ $diff -le 1 ] || error "elapsed for 1M@1T = $diff"
+
+ rm -f $DIR/$tfile
}
-run_test 80 "Page eviction is equally fast at high offsets too ===="
+run_test 80 "Page eviction is equally fast at high offsets too"
test_81a() { # LU-456
[ $PARALLEL == "yes" ] && skip "skip parallel run"
}
run_test 101c "check stripe_size aligned read-ahead ================="
-set_read_ahead() {
- $LCTL get_param -n llite.*.max_read_ahead_mb | head -n 1
- $LCTL set_param -n llite.*.max_read_ahead_mb $1 > /dev/null 2>&1
-}
-
test_101d() {
[ $PARALLEL == "yes" ] && skip "skip parallel run"
local file=$DIR/$tfile
- local sz_MB=${FILESIZE_101d:-500}
+ local sz_MB=${FILESIZE_101d:-80}
local ra_MB=${READAHEAD_MB:-40}
local free_MB=$(($(df -P $DIR | tail -n 1 | awk '{ print $4 }') / 1024))
cancel_lru_locks osc
echo Disable read-ahead
- local old_READAHEAD=$(set_read_ahead 0)
+ local old_RA=$($LCTL get_param -n llite.*.max_read_ahead_mb | head -n 1)
+ $LCTL set_param -n llite.*.max_read_ahead_mb=0
+ stack_trap "$LCTL set_param -n llite.*.max_read_ahead_mb $old_RA" EXIT
+ $LCTL get_param -n llite.*.max_read_ahead_mb
echo Reading the test file $file with read-ahead disabled
- local raOFF=$(do_and_time "dd if=$file of=/dev/null bs=1M count=$sz_MB")
+ local sz_KB=$((sz_MB * 1024 / 4))
+ local raOFF=$(do_and_time "dd if=$file of=/dev/null bs=4k count=$sz_KB")
echo Cancel LRU locks on lustre client to flush the client cache
cancel_lru_locks osc
echo Enable read-ahead with ${ra_MB}MB
- set_read_ahead $ra_MB
+ $LCTL set_param -n llite.*.max_read_ahead_mb=$ra_MB
echo Reading the test file $file with read-ahead enabled
- local raON=$(do_and_time "dd if=$file of=/dev/null bs=1M count=$sz_MB")
+ local raON=$(do_and_time "dd if=$file of=/dev/null bs=4k count=$sz_KB")
echo "read-ahead disabled time read $raOFF"
- echo "read-ahead enabled time read $raON"
+ echo "read-ahead enabled time read $raON"
- set_read_ahead $old_READAHEAD
rm -f $file
wait_delete_completed
}
run_test 101h "Readahead should cover current read window"
+test_101i() {
+ dd if=/dev/zero of=$DIR/$tfile bs=1M count=10 ||
+ error "dd 10M file failed"
+
+ local max_per_file_mb=$($LCTL get_param -n \
+ llite.*.max_read_ahead_per_file_mb 2>/dev/null)
+ cancel_lru_locks osc
+ stack_trap "$LCTL set_param llite.*.max_read_ahead_per_file_mb=$max_per_file_mb"
+ $LCTL set_param llite.*.max_read_ahead_per_file_mb=1 ||
+ error "set max_read_ahead_per_file_mb to 1 failed"
+
+ echo "Reset readahead stats"
+ $LCTL set_param llite.*.read_ahead_stats=0
+
+ dd if=$DIR/$tfile of=/dev/null bs=2M
+
+ $LCTL get_param llite.*.read_ahead_stats
+ local miss=$($LCTL get_param -n llite.*.read_ahead_stats |
+ awk '/misses/ { print $2 }')
+ [ $miss -eq 5 ] || error "expected misses 5 but got $miss"
+ rm -f $DIR/$tfile
+}
+run_test 101i "allow current readahead to exceed reservation"
+
+test_101j() {
+ $LFS setstripe -i 0 -c 1 $DIR/$tfile ||
+ error "setstripe $DIR/$tfile failed"
+ local file_size=$((1048576 * 16))
+ local old_ra=$($LCTL get_param -n llite.*.max_read_ahead_mb | head -n 1)
+ stack_trap "$LCTL set_param -n llite.*.max_read_ahead_mb $old_ra" EXIT
+
+ echo Disable read-ahead
+ $LCTL set_param -n llite.*.max_read_ahead_mb=0
+
+ dd if=/dev/zero of=$DIR/$tfile bs=1M count=$(($file_size / 1048576))
+ for blk in $PAGE_SIZE 1048576 $file_size; do
+ cancel_lru_locks osc
+ echo "Reset readahead stats"
+ $LCTL set_param -n llite.*.read_ahead_stats=0
+ local count=$(($file_size / $blk))
+ dd if=$DIR/$tfile bs=$blk count=$count of=/dev/null
+ local miss=$($LCTL get_param -n llite.*.read_ahead_stats |
+ get_named_value 'failed to fast read' |
+ cut -d" " -f1 | calc_total)
+ $LCTL get_param -n llite.*.read_ahead_stats
+ [ $miss -eq $count ] || error "expected $count got $miss"
+ done
+
+ rm -f $p $DIR/$tfile
+}
+run_test 101j "A complete read block should be submitted when no RA"
+
setup_test102() {
test_mkdir $DIR/$tdir
chown $RUNAS_ID $DIR/$tdir
}
run_test 123b "not panic with network error in statahead enqueue (bug 15027)"
+test_123c() {
+ [[ $MDSCOUNT -lt 2 ]] && skip_env "needs >= 2 MDTs"
+
+ test_mkdir -i 0 -c 1 $DIR/$tdir.0
+ test_mkdir -i 1 -c 1 $DIR/$tdir.1
+ touch $DIR/$tdir.1/{1..3}
+ mv $DIR/$tdir.1/{1..3} $DIR/$tdir.0
+
+ remount_client $MOUNT
+
+ $MULTIOP $DIR/$tdir.0 Q
+
+ # let statahead to complete
+ ls -l $DIR/$tdir.0 > /dev/null
+
+ testid=$(echo $TESTNAME | tr '_' ' ')
+ dmesg | tac | sed "/$testid/,$ d" | grep "Can not initialize inode" &&
+ error "statahead warning" || true
+}
+run_test 123c "Can not initialize inode warning on DNE statahead"
+
test_124a() {
[ $PARALLEL == "yes" ] && skip "skip parallel run"
$LCTL get_param -n mdc.*.connect_flags | grep -q lru_resize ||
test_127a() { # bug 15521
[ $PARALLEL == "yes" ] && skip "skip parallel run"
+ local name count samp unit min max sum sumsq
$LFS setstripe -i 0 -c 1 $DIR/$tfile || error "setstripe failed"
+ echo "stats before reset"
+ $LCTL get_param osc.*.stats
$LCTL set_param osc.*.stats=0
- FSIZE=$((2048 * 1024))
- dd if=/dev/zero of=$DIR/$tfile bs=$FSIZE count=1
+ local fsize=$((2048 * 1024))
+
+ dd if=/dev/zero of=$DIR/$tfile bs=$fsize count=1
cancel_lru_locks osc
- dd if=$DIR/$tfile of=/dev/null bs=$FSIZE
-
- $LCTL get_param osc.*0000-osc-*.stats | grep samples > $DIR/${tfile}.tmp
- while read NAME COUNT SAMP UNIT MIN MAX SUM SUMSQ; do
- echo "got $COUNT $NAME"
- [ ! $MIN ] && error "Missing min value for $NAME proc entry"
- eval $NAME=$COUNT || error "Wrong proc format"
-
- case $NAME in
- read_bytes|write_bytes)
- [ $MIN -lt 4096 ] && error "min is too small: $MIN"
- [ $MIN -gt $FSIZE ] && error "min is too big: $MIN"
- [ $MAX -lt 4096 ] && error "max is too small: $MAX"
- [ $MAX -gt $FSIZE ] && error "max is too big: $MAX"
- [ $SUM -ne $FSIZE ] && error "sum is wrong: $SUM"
- [ $SUMSQ -lt $(((FSIZE /4096) * (4096 * 4096))) ] &&
- error "sumsquare is too small: $SUMSQ"
- [ $SUMSQ -gt $((FSIZE * FSIZE)) ] &&
- error "sumsquare is too big: $SUMSQ"
- ;;
- *) ;;
- esac
- done < $DIR/${tfile}.tmp
-
- #check that we actually got some stats
- [ "$read_bytes" ] || error "Missing read_bytes stats"
- [ "$write_bytes" ] || error "Missing write_bytes stats"
- [ "$read_bytes" != 0 ] || error "no read done"
- [ "$write_bytes" != 0 ] || error "no write done"
+ dd if=$DIR/$tfile of=/dev/null bs=$fsize
+
+ $LCTL get_param osc.*0000-osc-*.stats | grep samples > $DIR/$tfile.tmp
+ stack_trap "rm -f $TMP/$tfile.tmp"
+ while read name count samp unit min max sum sumsq; do
+ echo "got name=$name count=$count unit=$unit min=$min max=$max"
+ [ ! $min ] && error "Missing min value for $name proc entry"
+ eval $name=$count || error "Wrong proc format"
+
+ case $name in
+ read_bytes|write_bytes)
+ [[ "$unit" =~ "bytes" ]] ||
+ error "unit is not 'bytes': $unit"
+ (( $min >= 4096 )) || error "min is too small: $min"
+ (( $min <= $fsize )) || error "min is too big: $min"
+ (( $max >= 4096 )) || error "max is too small: $max"
+ (( $max <= $fsize )) || error "max is too big: $max"
+ (( $sum == $fsize )) || error "sum is wrong: $sum"
+ (( $sumsq >= ($fsize / 4096) * (4096 * 4096) )) ||
+ error "sumsquare is too small: $sumsq"
+ (( $sumsq <= $fsize * $fsize )) ||
+ error "sumsquare is too big: $sumsq"
+ ;;
+ ost_read|ost_write)
+ [[ "$unit" =~ "usec" ]] ||
+ error "unit is not 'usec': $unit"
+ ;;
+ *) ;;
+ esac
+ done < $DIR/$tfile.tmp
+
+ #check that we actually got some stats
+ [ "$read_bytes" ] || error "Missing read_bytes stats"
+ [ "$write_bytes" ] || error "Missing write_bytes stats"
+ [ "$read_bytes" != 0 ] || error "no read done"
+ [ "$write_bytes" != 0 ] || error "no write done"
}
run_test 127a "verify the client stats are sane"
[ $PARALLEL == "yes" ] && skip "skip parallel run"
local name count samp unit min max sum sumsq
+ echo "stats before reset"
+ $LCTL get_param llite.*.stats
$LCTL set_param llite.*.stats=0
# perform 2 reads and writes so MAX is different from SUM.
dd if=$DIR/$tfile of=/dev/null bs=$PAGE_SIZE count=1
$LCTL get_param llite.*.stats | grep samples > $TMP/$tfile.tmp
+ stack_trap "rm -f $TMP/$tfile.tmp"
while read name count samp unit min max sum sumsq; do
- echo "got $count $name"
+ echo "got name=$name count=$count unit=$unit min=$min max=$max"
eval $name=$count || error "Wrong proc format"
case $name in
- read_bytes)
- [ $count -ne 2 ] && error "count is not 2: $count"
- [ $min -ne $PAGE_SIZE ] &&
+ read_bytes|write_bytes)
+ [[ "$unit" =~ "bytes" ]] ||
+ error "unit is not 'bytes': $unit"
+ (( $count == 2 )) || error "count is not 2: $count"
+ (( $min == $PAGE_SIZE )) ||
error "min is not $PAGE_SIZE: $min"
- [ $max -ne $PAGE_SIZE ] &&
- error "max is incorrect: $max"
- [ $sum -ne $((PAGE_SIZE * 2)) ] &&
- error "sum is wrong: $sum"
+ (( $max == $PAGE_SIZE )) ||
+ error "max is not $PAGE_SIZE: $max"
+ (( $sum == $PAGE_SIZE * 2 )) ||
+ error "sum is not $((PAGE_SIZE * 2)): $sum"
;;
- write_bytes)
- [ $count -ne 2 ] && error "count is not 2: $count"
- [ $min -ne $PAGE_SIZE ] &&
- error "min is not $PAGE_SIZE: $min"
- [ $max -ne $PAGE_SIZE ] &&
- error "max is incorrect: $max"
- [ $sum -ne $((PAGE_SIZE * 2)) ] &&
- error "sum is wrong: $sum"
+ read|write)
+ [[ "$unit" =~ "usec" ]] ||
+ error "unit is not 'usec': $unit"
;;
- *) ;;
+ *) ;;
esac
done < $TMP/$tfile.tmp
[ "$write_bytes" ] || error "Missing write_bytes stats"
[ "$read_bytes" != 0 ] || error "no read done"
[ "$write_bytes" != 0 ] || error "no write done"
-
- rm -f $TMP/${tfile}.tmp
}
run_test 127b "verify the llite client stats are sane"
remote_mds_nodsh && skip "remote MDS with nodsh"
remote_ost_nodsh && skip "remote OST with nodsh"
- # eventually, this can also be replaced with "lctl get_param -R",
- # but not until that option is always available on the server
local facet
for facet in mds1 ost1; do
- [ $(lustre_version_code $facet) -le $(version_code 2.5.54) ] &&
- skip_noexit "Too old lustre on $facet"
- local facet_proc_dirs=$(do_facet $facet \
- \\\ls -d $proc_regexp 2>/dev/null)
- echo "${facet}_proc_dirs='$facet_proc_dirs'"
- [ -z "$facet_proc_dirs" ] && error "no proc_dirs on $facet"
- do_facet $facet find $facet_proc_dirs \
- ! -name req_history \
- -exec cat '{}' \\\; &> /dev/null
-
- do_facet $facet find $facet_proc_dirs \
- ! -name req_history \
- -type f \
- -exec cat '{}' \\\; &> /dev/null ||
- error "proc file read failed"
-
- do_facet $facet find $facet_proc_dirs \
- -ignore_readdir_race \
- -type f \
- -not -name force_lbug \
- -not -name changelog_mask \
- -exec badarea_io '{}' \\\; ||
- error_133 "$facet find $facet_proc_dirs failed"
+ local facet_ver=$(lustre_version_code $facet)
+ if [ $facet_ver -ge $(version_code 2.7.65) ]; then
+ do_facet $facet "$LCTL get_param -R '*'" &> /dev/null
+ else
+ log "$facet: too old lustre for get_param -R"
+ fi
+ if [ $facet_ver -ge $(version_code 2.5.54) ]; then
+ do_facet $facet "$LCTL list_param -R '*' | grep '=' |
+ tr -d = | egrep -v 'force_lbug|changelog_mask' |
+ xargs badarea_io" ||
+ error_133 "$facet badarea_io failed"
+ else
+ skip_noexit "$facet: too old lustre for get_param -R"
+ fi
done
# remount the FS in case writes/reads /proc break the FS
}
run_test 134b "Server rejects lock request when reaching lock_limit_mb"
+test_135() {
+ remote_mds_nodsh && skip "remote MDS with nodsh"
+ [[ $MDS1_VERSION -lt $(version_code 2.13.50) ]] &&
+ skip "Need MDS version at least 2.13.50"
+ local fname
+
+ mkdir -p $DIR/$tdir || error "failed to create $DIR/$tdir"
+
+#define OBD_FAIL_PLAIN_RECORDS 0x1319
+ #set only one record at plain llog
+ do_facet $SINGLEMDS $LCTL set_param fail_loc=0x1319 fail_val=1
+
+ #fill already existed plain llog each 64767
+ #wrapping whole catalog
+ createmany -o -u $DIR/$tdir/$tfile- $((64767 * 1))
+
+ createmany -o $DIR/$tdir/$tfile_ 64700
+ for (( i = 0; i < 64700; i = i + 2 ))
+ do
+ rm $DIR/$tdir/$tfile_$i &
+ rm $DIR/$tdir/$tfile_$((i + 1)) &
+ local pid=$!
+ wait $pid
+ done
+
+ #waiting osp synchronization
+ wait_delete_completed
+}
+run_test 135 "Race catalog processing"
+
+test_136() {
+ remote_mds_nodsh && skip "remote MDS with nodsh"
+ [[ $MDS1_VERSION -lt $(version_code 2.13.50) ]] &&
+ skip "Need MDS version at least 2.13.50"
+ local fname
+
+ mkdir -p $DIR/$tdir || error "failed to create $DIR/$tdir"
+ $LFS setstripe -c 1 -i 0 $DIR/$tdir || error "failed to set striping"
+ #set only one record at plain llog
+#define OBD_FAIL_CATALOG_FULL_CHECK 0x131a
+ do_facet $SINGLEMDS $LCTL set_param fail_loc=0x131a fail_val=1
+
+ #fill already existed 2 plain llogs each 64767
+ #wrapping whole catalog
+ createmany -o -u $DIR/$tdir/$tfile- $((64767 * 1))
+ createmany -o -u $DIR/$tdir/$tfile- $((64767 * 3 / 2))
+ wait_delete_completed
+
+ createmany -o $DIR/$tdir/$tfile_ 10
+ sleep 25
+
+ do_facet $SINGLEMDS $LCTL set_param fail_val=3
+ for (( i = 0; i < 10; i = i + 3 ))
+ do
+ rm $DIR/$tdir/$tfile_$i &
+ rm $DIR/$tdir/$tfile_$((i + 1)) &
+ local pid=$!
+ wait $pid
+ sleep 7
+ rm $DIR/$tdir/$tfile_$((i + 2)) &
+ done
+
+ #waiting osp synchronization
+ wait_delete_completed
+}
+run_test 136 "Race catalog processing 2"
+
test_140() { #bug-17379
[ $PARALLEL == "yes" ] && skip "skip parallel run"
error "NOT IN CACHE: before: $BEFORE, after: $AFTER"
fi
- # the following read invalidates the cache
cancel_lru_locks osc
+ # invalidates OST cache
+ do_nodes $list "echo 1 > /proc/sys/vm/drop_caches"
set_osd_param $list '' read_cache_enable 0
cat $DIR/$tfile >/dev/null
cat $file >/dev/null
AFTER=$(roc_hit)
if ! let "AFTER - BEFORE == CPAGES"; then
- error "NOT IN CACHE: before: $BEFORE, after: $AFTER"
+ error "NOT IN CACHE (2): before: $BEFORE, after: $AFTER"
else
- log "cache hits:: before: $BEFORE, after: $AFTER"
+ log "cache hits: before: $BEFORE, after: $AFTER"
fi
log "Read again; it should be satisfied from the cache."
cat $file >/dev/null
AFTER=$(roc_hit)
if ! let "AFTER - BEFORE == CPAGES"; then
- error "NOT IN CACHE: before: $BEFORE, after: $AFTER"
+ error "NOT IN CACHE (3): before: $BEFORE, after: $AFTER"
else
log "cache hits:: before: $BEFORE, after: $AFTER"
fi
cat $file >/dev/null
AFTER=$(roc_hit)
if ! let "AFTER - BEFORE == CPAGES"; then
- error "NOT IN CACHE: before: $BEFORE, after: $AFTER"
+ error "NOT IN CACHE (4): before: $BEFORE, after: $AFTER"
else
log "cache hits:: before: $BEFORE, after: $AFTER"
fi
- log "Read again; it should not be satisfied from the cache."
- BEFORE=$AFTER
- cancel_lru_locks osc
- cat $file >/dev/null
- AFTER=$(roc_hit)
- if ! let "AFTER - BEFORE == 0"; then
- error "IN CACHE: before: $BEFORE, after: $AFTER"
- else
- log "cache hits:: before: $BEFORE, after: $AFTER"
+ if [ $OST1_VERSION -lt $(version_code 2.12.55) ]; then
+ # > 2.12.56 uses pagecache if cached
+ log "Read again; it should not be satisfied from the cache."
+ BEFORE=$AFTER
+ cancel_lru_locks osc
+ cat $file >/dev/null
+ AFTER=$(roc_hit)
+ if ! let "AFTER - BEFORE == 0"; then
+ error "IN CACHE (5): before: $BEFORE, after: $AFTER"
+ else
+ log "cache hits:: before: $BEFORE, after: $AFTER"
+ fi
fi
log "Write data and read it back."
cat $file >/dev/null
AFTER=$(roc_hit)
if ! let "AFTER - BEFORE == CPAGES"; then
- error "NOT IN CACHE: before: $BEFORE, after: $AFTER"
+ error "NOT IN CACHE (6): before: $BEFORE, after: $AFTER"
else
log "cache hits:: before: $BEFORE, after: $AFTER"
fi
- log "Read again; it should not be satisfied from the cache."
- BEFORE=$AFTER
- cancel_lru_locks osc
- cat $file >/dev/null
- AFTER=$(roc_hit)
- if ! let "AFTER - BEFORE == 0"; then
- error "IN CACHE: before: $BEFORE, after: $AFTER"
- else
- log "cache hits:: before: $BEFORE, after: $AFTER"
+ if [ $OST1_VERSION -lt $(version_code 2.12.55) ]; then
+ # > 2.12.56 uses pagecache if cached
+ log "Read again; it should not be satisfied from the cache."
+ BEFORE=$AFTER
+ cancel_lru_locks osc
+ cat $file >/dev/null
+ AFTER=$(roc_hit)
+ if ! let "AFTER - BEFORE == 0"; then
+ error "IN CACHE (7): before: $BEFORE, after: $AFTER"
+ else
+ log "cache hits:: before: $BEFORE, after: $AFTER"
+ fi
fi
log "Turn off read and write cache"
cat $file >/dev/null
AFTER=$(roc_hit)
if ! let "AFTER - BEFORE == 0"; then
- error_ignore bz20762 "IN CACHE: before: $BEFORE, after: $AFTER"
+ error_ignore bz20762 "IN CACHE (8):before:$BEFORE,after:$AFTER"
else
log "cache hits:: before: $BEFORE, after: $AFTER"
fi
cat $file >/dev/null
AFTER=$(roc_hit)
if ! let "AFTER - BEFORE == 0"; then
- error_ignore bz20762 "IN CACHE: before: $BEFORE, after: $AFTER"
+ error_ignore bz20762 "IN CACHE (9):before:$BEFORE,after:$AFTER"
else
log "cache hits:: before: $BEFORE, after: $AFTER"
fi
cat $file >/dev/null
AFTER=$(roc_hit)
if ! let "AFTER - BEFORE == CPAGES"; then
- error "NOT IN CACHE: before: $BEFORE, after: $AFTER"
+ error "NOT IN CACHE (1): before: $BEFORE, after: $AFTER"
else
log "cache hits:: before: $BEFORE, after: $AFTER"
fi
local i
# generate some changelog records to accumulate on each MDT
- test_mkdir -c $MDSCOUNT $DIR/$tdir || error "test_mkdir $tdir failed"
+ # use fnv1a because created files should be evenly distributed
+ test_mkdir -c $MDSCOUNT -H fnv_1a_64 $DIR/$tdir ||
+ error "test_mkdir $tdir failed"
+ log "$(date +%s): creating first files"
createmany -m $DIR/$tdir/$tfile $((MDSCOUNT * 2)) ||
error "create $DIR/$tdir/$tfile failed"
# check changelogs have been generated
+ local start=$SECONDS
+ local idle_time=$((MDSCOUNT * 5 + 5))
local nbcl=$(changelog_dump | wc -l)
[[ $nbcl -eq 0 ]] && error "no changelogs found"
- for param in "changelog_max_idle_time=10" \
+ for param in "changelog_max_idle_time=$idle_time" \
"changelog_gc=1" \
"changelog_min_gc_interval=2" \
"changelog_min_free_cat_entries=3"; do
do_nodes $mdts $LCTL set_param mdd.*.$param
done
- # force cl_user2 to be idle (1st part)
- sleep 9
+ # force cl_user2 to be idle (1st part), but also cancel the
+ # cl_user1 records so that it is not evicted later in the test.
+ local sleep1=$((idle_time / 2))
+ echo "$(date +%s): sleep1 $sleep1/${idle_time}s"
+ sleep $sleep1
# simulate changelog catalog almost full
#define OBD_FAIL_CAT_FREE_RECORDS 0x1313
"$user_rec1, but is $user_rec2"
done
- # force cl_user2 to be idle (2nd part) and to reach
- # changelog_max_idle_time
- sleep 2
+ # force cl_user2 idle (2nd part) to just exceed changelog_max_idle_time
+ local sleep2=$((idle_time - (SECONDS - start) + 1))
+ echo "$(date +%s): sleep2 $sleep2/${idle_time}s"
+ sleep $sleep2
- # generate one more changelog to trigger fail_loc
- createmany -m $DIR/$tdir/${tfile}bis $((MDSCOUNT * 2)) ||
- error "create $DIR/$tdir/${tfile}bis failed"
+ # Generate one more changelog to trigger GC at fail_loc for cl_user2.
+ # cl_user1 should be OK because it recently processed records.
+ echo "$(date +%s): creating $((MDSCOUNT * 2)) files"
+ createmany -m $DIR/$tdir/${tfile}b $((MDSCOUNT * 2)) ||
+ error "create $DIR/$tdir/${tfile}b failed"
# ensure gc thread is done
for i in $(mdts_nodes); do
local i
# generate some changelog records to accumulate on each MDT
- test_mkdir -c $MDSCOUNT $DIR/$tdir || error "mkdir $tdir failed"
+ # use fnv1a because created files should be evenly distributed
+ test_mkdir -c $MDSCOUNT -H fnv_1a_64 $DIR/$tdir ||
+ error "mkdir $tdir failed"
createmany -m $DIR/$tdir/$tfile $((MDSCOUNT * 2)) ||
error "create $DIR/$tdir/$tfile failed"
local i
# generate some changelog records to accumulate on each MDT
- test_mkdir -c $MDSCOUNT $DIR/$tdir || error "test_mkdir $tdir failed"
+ # use fnv1a because created files should be evenly distributed
+ test_mkdir -c $MDSCOUNT -H fnv_1a_64 $DIR/$tdir ||
+ error "test_mkdir $tdir failed"
createmany -m $DIR/$tdir/$tfile $((MDSCOUNT * 2)) ||
error "create $DIR/$tdir/$tfile failed"
changelog_register || error "first changelog_register failed"
# generate some changelog records to accumulate on each MDT
- test_mkdir -c $MDSCOUNT $DIR/$tdir || error "mkdir $tdir failed"
+ # use fnv1a because created files should be evenly distributed
+ test_mkdir -c $MDSCOUNT -H fnv_1a_64 $DIR/$tdir ||
+ error "mkdir $tdir failed"
createmany -m $DIR/$tdir/$tfile $((MDSCOUNT * 2)) ||
error "create $DIR/$tdir/$tfile failed"
skip "Need MDS version at least 2.12.56"
mount_client $MOUNT2 || error "mount_client on $MOUNT2 failed"
+ stack_trap "umount $MOUNT2" EXIT
changelog_register || error "first changelog_register failed"
+ stack_trap "changelog_deregister" EXIT
# generate some changelog
- test_mkdir -c $MDSCOUNT $DIR/$tdir || error "mkdir $tdir failed"
+ # use fnv1a because created files should be evenly distributed
+ test_mkdir -c $MDSCOUNT -H fnv_1a_64 $DIR/$tdir ||
+ error "mkdir $tdir failed"
createmany -m $DIR/$tdir/${tfile}bis $((MDSCOUNT * 2)) ||
error "create $DIR/$tdir/${tfile}bis failed"
# open the changelog device
exec 3>/dev/changelog-$FSNAME-MDT0000
+ stack_trap "exec 3>&-" EXIT
exec 4</dev/changelog-$FSNAME-MDT0000
+ stack_trap "exec 4<&-" EXIT
# umount the first lustre mount
umount $MOUNT
+ stack_trap "mount_client $MOUNT" EXIT
# read changelog
cat <&4 >/dev/null || error "read changelog failed"
error "User $cl_user not found in changelog_users"
printf 'clear:'$cl_user':0' >&3
-
- # close
- exec 3>&-
- exec 4<&-
-
- # cleanup
- changelog_deregister || error "changelog_deregister failed"
-
- umount $MOUNT2
- mount_client $MOUNT || error "mount_client on $MOUNT failed"
}
run_test 160j "client can be umounted while its chanangelog is being used"
"$FSNAME.sys.jobid_var" $new_jobenv
}
-test_205() { # Job stats
+test_205a() { # Job stats
[ $PARALLEL == "yes" ] && skip "skip parallel run"
[[ $MDS1_VERSION -ge $(version_code 2.7.1) ]] ||
skip "Need MDS version with at least 2.7.1"
verify_jobstats "touch $DIR/$tfile" $SINGLEMDS
}
-run_test 205 "Verify job stats"
+run_test 205a "Verify job stats"
+
+# LU-13117
+test_205b() {
+ $LCTL set_param jobid_var=USER jobid_name="%e.%u"
+ env -i USERTESTJOBSTATS=foolish touch $DIR/$tfile.1
+ do_facet $SINGLEMDS $LCTL get_param mdt.*.job_stats |
+ grep job_id: | grep foolish &&
+ error "Unexpected jobid found"
+ true
+}
+run_test 205b "Verify job stats jobid parsing"
# LU-1480, LU-1773 and LU-1657
test_206() {
save_writethrough $p
set_cache writethrough on
- local pages_per_rpc=$($LCTL get_param \
- osc.*.max_pages_per_rpc)
+ local pages_per_rpc=$($LCTL get_param osc.*.max_pages_per_rpc)
local at_max=$($LCTL get_param -n at_max)
local timeout=$($LCTL get_param -n timeout)
local test_at="at_max"
ln -s $migrate_dir/$tfile $migrate_dir/${tfile}_ln
ln -s $other_dir/$tfile $migrate_dir/${tfile}_ln_other
+ local len
+ local lnktgt
+
+ # inline symlink
+ for len in 58 59 60; do
+ lnktgt=$(str_repeat 'l' $len)
+ touch $migrate_dir/$lnktgt
+ ln -s $lnktgt $migrate_dir/${len}char_ln
+ done
+
+ # PATH_MAX
+ for len in 4094 4095; do
+ lnktgt=$(str_repeat 'l' $len)
+ ln -s $lnktgt $migrate_dir/${len}char_ln
+ done
+
+ # NAME_MAX
+ for len in 254 255; do
+ touch $migrate_dir/$(str_repeat 'l' $len)
+ done
+
$LFS migrate -m $MDTIDX $migrate_dir ||
error "fails on migrating remote dir to MDT1"
for ((i = 0; i < 10; i++)); do
for file in $(find $migrate_dir/dir_${i}); do
mdt_index=$($LFS getstripe -m $file)
- [ $mdt_index == $MDTIDX ] ||
+ # broken symlink getstripe will fail
+ [ $mdt_index -ne $MDTIDX ] && stat -L $file &&
error "$file is not on MDT${MDTIDX}"
done
done
echo "migrate back to MDT0, checking.."
for file in $(find $migrate_dir); do
mdt_index=$($LFS getstripe -m $file)
- [ $mdt_index == $MDTIDX ] ||
+ [ $mdt_index -ne $MDTIDX ] && stat -L $file &&
error "$file is not on MDT${MDTIDX}"
done
old_count=$((MDSCOUNT - old_index))
new_index=$((RANDOM % MDSCOUNT))
new_count=$((MDSCOUNT - new_index))
- new_hash="all_char"
+ new_hash=1 # for all_char
[ $old_count -gt 1 ] && old_count=$((old_count - RANDOM % old_count))
[ $new_count -gt 1 ] && new_count=$((new_count - RANDOM % new_count))
}
run_test 230l "readdir between MDTs won't crash"
+test_230m() {
+ [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
+ [ $MDS1_VERSION -lt $(version_code 2.11.56) ] &&
+ skip "Need MDS version at least 2.11.56"
+
+ local MDTIDX=1
+ local mig_dir=$DIR/$tdir/migrate_dir
+ local longstr="aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa"
+ local shortstr="b"
+ local val
+
+ echo "Creating files and dirs with xattrs"
+ test_mkdir $DIR/$tdir
+ test_mkdir -i0 -c1 $mig_dir
+ mkdir $mig_dir/dir
+ setfattr -n user.attr1 -v $longstr $mig_dir/dir ||
+ error "cannot set xattr attr1 on dir"
+ setfattr -n user.attr2 -v $shortstr $mig_dir/dir ||
+ error "cannot set xattr attr2 on dir"
+ touch $mig_dir/dir/f0
+ setfattr -n user.attr1 -v $longstr $mig_dir/dir/f0 ||
+ error "cannot set xattr attr1 on file"
+ setfattr -n user.attr2 -v $shortstr $mig_dir/dir/f0 ||
+ error "cannot set xattr attr2 on file"
+ sync ; sync ; echo 3 > /proc/sys/vm/drop_caches
+ val=$(getfattr --only-values -n user.attr1 $mig_dir/dir 2>/dev/null)
+ [ "$val" = $longstr ] || error "xattr attr1 not set properly on dir"
+ val=$(getfattr --only-values -n user.attr2 $mig_dir/dir 2>/dev/null)
+ [ "$val" = $shortstr ] || error "xattr attr2 not set properly on dir"
+ val=$(getfattr --only-values -n user.attr1 $mig_dir/dir/f0 2>/dev/null)
+ [ "$val" = $longstr ] || error "xattr attr1 not set properly on file"
+ val=$(getfattr --only-values -n user.attr2 $mig_dir/dir/f0 2>/dev/null)
+ [ "$val" = $shortstr ] || error "xattr attr2 not set properly on file"
+
+ echo "Migrating to MDT1"
+ $LFS migrate -m $MDTIDX $mig_dir ||
+ error "fails on migrating dir to MDT1"
+
+ sync ; sync ; echo 3 > /proc/sys/vm/drop_caches
+ echo "Checking xattrs"
+ val=$(getfattr --only-values -n user.attr1 $mig_dir/dir 2>/dev/null)
+ [ "$val" = $longstr ] ||
+ error "expecting xattr1 $longstr on dir, found $val"
+ val=$(getfattr --only-values -n user.attr2 $mig_dir/dir 2>/dev/null)
+ [ "$val" = $shortstr ] ||
+ error "expecting xattr2 $shortstr on dir, found $val"
+ val=$(getfattr --only-values -n user.attr1 $mig_dir/dir/f0 2>/dev/null)
+ [ "$val" = $longstr ] ||
+ error "expecting xattr1 $longstr on file, found $val"
+ val=$(getfattr --only-values -n user.attr2 $mig_dir/dir/f0 2>/dev/null)
+ [ "$val" = $shortstr ] ||
+ error "expecting xattr2 $shortstr on file, found $val"
+}
+run_test 230m "xattrs not changed after dir migration"
+
+test_230n() {
+ [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
+ [[ $MDS1_VERSION -ge $(version_code 2.13.53) ]] ||
+ skip "Need MDS version at least 2.13.53"
+
+ $LFS mkdir -i 0 $DIR/$tdir || error "mkdir $tdir failed"
+ cat /etc/hosts > $DIR/$tdir/$tfile
+ $LFS mirror extend -N1 $DIR/$tdir/$tfile || error "Mirroring failed"
+ $LFS migrate -m 1 $DIR/$tdir || error "Migration failed"
+
+ cmp /etc/hosts $DIR/$tdir/$tfile ||
+ error "File data mismatch after migration"
+}
+run_test 230n "Dir migration with mirrored file"
+
test_231a()
{
# For simplicity this test assumes that max_pages_per_rpc
}
run_test 245 "check mdc connection flag/data: multiple modify RPCs"
-test_246() { # LU-7371
- remote_ost_nodsh && skip "remote OST with nodsh"
- [ $OST1_VERSION -lt $(version_code 2.7.62) ] &&
- skip "Need OST version >= 2.7.62"
-
- do_facet ost1 $LCTL set_param fail_val=4095
-#define OBD_FAIL_OST_READ_SIZE 0x234
- do_facet ost1 $LCTL set_param fail_loc=0x234
- $LFS setstripe $DIR/$tfile -i 0 -c 1
- dd if=/dev/zero of=$DIR/$tfile bs=4095 count=1 > /dev/null 2>&1
- cancel_lru_locks $FSNAME-OST0000
- dd if=$DIR/$tfile of=/dev/null bs=1048576 || error "Read failed"
-}
-run_test 246 "Read file of size 4095 should return right length"
-
cleanup_247() {
local submount=$1
}
run_test 247e "mount .. as fileset"
-test_248() {
+test_248a() {
local fast_read_sav=$($LCTL get_param -n llite.*.fast_read 2>/dev/null)
[ -z "$fast_read_sav" ] && skip "no fast read support"
$LCTL set_param -n llite.*.fast_read=$fast_read_sav
rm -f $DIR/$tfile
}
-run_test 248 "fast read verification"
+run_test 248a "fast read verification"
+
+test_248b() {
+ # Default short_io_bytes=16384, try both smaller and larger sizes.
+ # Lustre O_DIRECT read and write needs to be a multiple of PAGE_SIZE.
+ # 6017024 = 2^12*13*113 = 47008*128 = 11752*512 = 4096*1469 = 53248*113
+ echo "bs=53248 count=113 normal buffered write"
+ dd if=/dev/urandom of=$TMP/$tfile.0 bs=53248 count=113 ||
+ error "dd of initial data file failed"
+ stack_trap "rm -f $DIR/$tfile.[0-3] $TMP/$tfile.[0-3]" EXIT
+
+ echo "bs=47008 count=128 oflag=dsync normal write $tfile.0"
+ dd if=$TMP/$tfile.0 of=$DIR/$tfile.0 bs=47008 count=128 oflag=dsync ||
+ error "dd with sync normal writes failed"
+ cmp $TMP/$tfile.0 $DIR/$tfile.0 || error "compare $DIR/$tfile.0 failed"
+
+ echo "bs=11752 count=512 oflag=dsync small write $tfile.1"
+ dd if=$TMP/$tfile.0 of=$DIR/$tfile.1 bs=11752 count=512 oflag=dsync ||
+ error "dd with sync small writes failed"
+ cmp $TMP/$tfile.0 $DIR/$tfile.1 || error "compare $DIR/$tfile.1 failed"
+
+ cancel_lru_locks osc
+
+ # calculate the small O_DIRECT size and count for the client PAGE_SIZE
+ local num=$((13 * 113 / (PAGE_SIZE / 4096)))
+ echo "bs=$PAGE_SIZE count=$num iflag=direct small read $tfile.1"
+ dd if=$DIR/$tfile.1 of=$TMP/$tfile.1 bs=$PAGE_SIZE count=$num \
+ iflag=direct || error "dd with O_DIRECT small read failed"
+ # adjust bytes checked to handle larger PAGE_SIZE for ARM/PPC
+ cmp --bytes=$((PAGE_SIZE * num)) $TMP/$tfile.0 $TMP/$tfile.1 ||
+ error "compare $TMP/$tfile.1 failed"
+
+ local save=$($LCTL get_param -n osc.*OST000*.short_io_bytes | head -n 1)
+ stack_trap "$LCTL set_param osc.$FSNAME-*.short_io_bytes=$save" EXIT
+
+ # just to see what the maximum tunable value is, and test parsing
+ echo "test invalid parameter 2MB"
+ $LCTL set_param osc.$FSNAME-OST0000*.short_io_bytes=2M &&
+ error "too-large short_io_bytes allowed"
+ echo "test maximum parameter 512KB"
+ # if we can set a larger short_io_bytes, run test regardless of version
+ if ! $LCTL set_param osc.$FSNAME-OST0000*.short_io_bytes=512K; then
+ # older clients may not allow setting it this large, that's OK
+ [ $CLIENT_VERSION -ge $(version_code 2.13.50) ] ||
+ skip "Need at least client version 2.13.50"
+ error "medium short_io_bytes failed"
+ fi
+ $LCTL get_param osc.$FSNAME-OST0000*.short_io_bytes
+ size=$($LCTL get_param -n osc.$FSNAME-OST0000*.short_io_bytes)
+
+ echo "test large parameter 64KB"
+ $LCTL set_param osc.$FSNAME-*.short_io_bytes=65536
+ $LCTL get_param osc.$FSNAME-OST0000*.short_io_bytes
+
+ echo "bs=47008 count=128 oflag=dsync large write $tfile.2"
+ dd if=$TMP/$tfile.0 of=$DIR/$tfile.2 bs=47008 count=128 oflag=dsync ||
+ error "dd with sync large writes failed"
+ cmp $TMP/$tfile.0 $DIR/$tfile.2 || error "compare $DIR/$tfile.2 failed"
+
+ # calculate the large O_DIRECT size and count for the client PAGE_SIZE
+ local size=$(((4096 * 13 + PAGE_SIZE - 1) / PAGE_SIZE * PAGE_SIZE))
+ num=$((113 * 4096 / PAGE_SIZE))
+ echo "bs=$size count=$num oflag=direct large write $tfile.3"
+ dd if=$TMP/$tfile.0 of=$DIR/$tfile.3 bs=$size count=$num oflag=direct ||
+ error "dd with O_DIRECT large writes failed"
+ cmp --bytes=$((size * num)) $TMP/$tfile.0 $DIR/$tfile.3 ||
+ error "compare $DIR/$tfile.3 failed"
+
+ cancel_lru_locks osc
+
+ echo "bs=$size count=$num iflag=direct large read $tfile.2"
+ dd if=$DIR/$tfile.2 of=$TMP/$tfile.2 bs=$size count=$num iflag=direct ||
+ error "dd with O_DIRECT large read failed"
+ cmp --bytes=$((size * num)) $TMP/$tfile.0 $TMP/$tfile.2 ||
+ error "compare $TMP/$tfile.2 failed"
+
+ echo "bs=$size count=$num iflag=direct large read $tfile.3"
+ dd if=$DIR/$tfile.3 of=$TMP/$tfile.3 bs=$size count=$num iflag=direct ||
+ error "dd with O_DIRECT large read failed"
+ cmp --bytes=$((size * num)) $TMP/$tfile.0 $TMP/$tfile.3 ||
+ error "compare $TMP/$tfile.3 failed"
+}
+run_test 248b "test short_io read and write for both small and large sizes"
test_249() { # LU-7890
[ $MDS1_VERSION -lt $(version_code 2.8.53) ] &&
error "md5sum differ: $old_md5, $new_md5"
[ $($LFS getstripe -c $dom) -eq 2 ] ||
- error "migrate stripe count bad: $(LFS getstripe -c $dom) != 2"
+ error "bad final stripe count: $($LFS getstripe -c $dom) != 2"
}
run_test 272a "DoM migration: new layout with the same DOM component"
}
run_test 278 "Race starting MDS between MDTs stop/start"
+test_280() {
+ [ $MGS_VERSION -lt $(version_code 2.13.52) ] &&
+ skip "Need MGS version at least 2.13.52"
+ [ $PARALLEL == "yes" ] && skip "skip parallel run"
+ combined_mgs_mds || skip "needs combined MGS/MDT"
+
+ umount_client $MOUNT
+#define OBD_FAIL_MDS_LLOG_UMOUNT_RACE 0x15e
+ do_facet mgs $LCTL set_param fail_loc=0x8000015e fail_val=0
+
+ mount_client $MOUNT &
+ sleep 1
+ stop mgs || error "stop mgs failed"
+ #for a race mgs would crash
+ start mgs $(mgsdevname) $MGS_MOUNT_OPTS || error "start mgs failed"
+ mount_client $MOUNT || error "mount client failed"
+}
+run_test 280 "Race between MGS umount and client llog processing"
+
cleanup_test_300() {
trap 0
umask $SAVE_UMASK
#local striped directory
$LFS setdirstripe -i 0 -c 2 -H all_char $DIR/$tdir/striped_dir ||
error "set striped dir error"
+ #look at the directories for debug purposes
+ ls -l $DIR/$tdir
+ $LFS getdirstripe $DIR/$tdir
+ ls -l $DIR/$tdir/striped_dir
+ $LFS getdirstripe $DIR/$tdir/striped_dir
createmany -o $DIR/$tdir/striped_dir/f 10 ||
error "create 10 files failed"
#remote striped directory
$LFS setdirstripe -i 1 -c 2 $DIR/$tdir/remote_striped_dir ||
error "set striped dir error"
+ #look at the directories for debug purposes
+ ls -l $DIR/$tdir
+ $LFS getdirstripe $DIR/$tdir
+ ls -l $DIR/$tdir/remote_striped_dir
+ $LFS getdirstripe $DIR/$tdir/remote_striped_dir
createmany -o $DIR/$tdir/remote_striped_dir/f 10 ||
error "create 10 files failed"
# this test needs a huge transaction
local kb
- kb=$(do_facet $SINGLEMDS lctl get_param -n osd*.lustre-MDT0000.kbytestotal)
- [ $kb -lt $((1024*1024)) ] && skip "too small mds: $kb"
+ kb=$(do_facet $SINGLEMDS "$LCTL get_param -n \
+ osd*.$FSNAME-MDT0000.kbytestotal")
+ [ $kb -lt $((1024*1024)) ] && skip "MDT0 too small: $kb"
local stripe_count
local file
run_test 300q "create remote directory under orphan directory"
test_300r() {
- [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.7.55) ] &&
+ [ $MDS1_VERSION -lt $(version_code 2.7.55) ] &&
skip "Need MDS version at least 2.7.55" && return
[ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
chown nobody $DIR/$tdir/d
touch $DIR/$tdir/d/file
- $LFS mv -M1 $DIR/$tdir/d || error "lfs mv failed"
+ $LFS mv -m1 $DIR/$tdir/d || error "lfs mv failed"
}
run_test 316 "lfs mv"
llite.*.max_read_ahead_async_active 2>/dev/null)
[ $max_active -ne 256 ] && error "expected 256 but got $max_active"
- # currently reset to 0 is unsupported, leave it 512 for now.
- $LCTL set_param llite.*.max_read_ahead_async_active=0 &&
- error "set max_read_ahead_async_active should fail"
+ $LCTL set_param llite.*.max_read_ahead_async_active=0 ||
+ error "set max_read_ahead_async_active should succeed"
$LCTL set_param llite.*.max_read_ahead_async_active=512
max_active=$($LCTL get_param -n \
}
run_test 319 "lost lease lock on migrate error"
+test_398a() { # LU-4198
+ $LFS setstripe -c 1 -i 0 $DIR/$tfile
+ $LCTL set_param ldlm.namespaces.*.lru_size=clear
+
+ # request a new lock on client
+ dd if=/dev/zero of=$DIR/$tfile bs=1M count=1
+
+ dd if=/dev/zero of=$DIR/$tfile bs=1M count=1 oflag=direct conv=notrunc
+ local lock_count=$($LCTL get_param -n \
+ ldlm.namespaces.*-OST0000-osc-ffff*.lru_size)
+ [[ $lock_count -eq 0 ]] || error "lock should be cancelled by direct IO"
+
+ $LCTL set_param ldlm.namespaces.*-OST0000-osc-ffff*.lru_size=clear
+
+ # no lock cached, should use lockless IO and not enqueue new lock
+ dd if=/dev/zero of=$DIR/$tfile bs=1M count=1 oflag=direct conv=notrunc
+ lock_count=$($LCTL get_param -n \
+ ldlm.namespaces.*-OST0000-osc-ffff*.lru_size)
+ [[ $lock_count -eq 0 ]] || error "no lock should be held by direct IO"
+}
+run_test 398a "direct IO should cancel lock otherwise lockless"
+
+test_398b() { # LU-4198
+ which fio || skip_env "no fio installed"
+ $LFS setstripe -c -1 $DIR/$tfile
+
+ local size=12
+ dd if=/dev/zero of=$DIR/$tfile bs=1M count=$size
+
+ local njobs=4
+ echo "mix direct rw ${size}M to OST0 by fio with $njobs jobs..."
+ fio --name=rand-rw --rw=randrw --bs=$PAGE_SIZE --direct=1 \
+ --numjobs=$njobs --fallocate=none \
+ --iodepth=16 --allow_file_create=0 --size=$((size/njobs))M \
+ --filename=$DIR/$tfile &
+ bg_pid=$!
+
+ echo "mix buffer rw ${size}M to OST0 by fio with $njobs jobs..."
+ fio --name=rand-rw --rw=randrw --bs=$PAGE_SIZE \
+ --numjobs=$njobs --fallocate=none \
+ --iodepth=16 --allow_file_create=0 --size=$((size/njobs))M \
+ --filename=$DIR/$tfile || true
+ wait $bg_pid
+
+ rm -rf $DIR/$tfile
+}
+run_test 398b "DIO and buffer IO race"
+
+test_398c() { # LU-4198
+ which fio || skip_env "no fio installed"
+
+ saved_debug=$($LCTL get_param -n debug)
+ $LCTL set_param debug=0
+
+ local size=$(lctl get_param -n osc.$FSNAME-OST0000*.kbytesavail | head -1)
+ ((size /= 1024)) # by megabytes
+ ((size /= 2)) # write half of the OST at most
+ [ $size -gt 40 ] && size=40 #reduce test time anyway
+
+ $LFS setstripe -c 1 $DIR/$tfile
+
+ # it seems like ldiskfs reserves more space than necessary if the
+ # writing blocks are not mapped, so it extends the file firstly
+ dd if=/dev/zero of=$DIR/$tfile bs=1M count=$size && sync
+ cancel_lru_locks osc
+
+ # clear and verify rpc_stats later
+ $LCTL set_param osc.${FSNAME}-OST0000-osc-ffff*.rpc_stats=clear
+
+ local njobs=4
+ echo "writing ${size}M to OST0 by fio with $njobs jobs..."
+ fio --name=rand-write --rw=randwrite --bs=$PAGE_SIZE --direct=1 \
+ --numjobs=$njobs --fallocate=none --ioengine=libaio \
+ --iodepth=16 --allow_file_create=0 --size=$((size/njobs))M \
+ --filename=$DIR/$tfile
+ [ $? -eq 0 ] || error "fio write error"
+
+ [ $($LCTL get_param -n \
+ ldlm.namespaces.${FSNAME}-OST0000-osc-ffff*.lock_count) -eq 0 ] ||
+ error "Locks were requested while doing AIO"
+
+ # get the percentage of 1-page I/O
+ pct=$($LCTL get_param osc.${FSNAME}-OST0000-osc-ffff*.rpc_stats |
+ grep -A 1 'pages per rpc' | grep -v 'pages per rpc' |
+ awk '{print $7}')
+ [ $pct -le 50 ] || error "$pct% of I/O are 1-page"
+
+ echo "mix rw ${size}M to OST0 by fio with $njobs jobs..."
+ fio --name=rand-rw --rw=randrw --bs=$PAGE_SIZE --direct=1 \
+ --numjobs=$njobs --fallocate=none --ioengine=libaio \
+ --iodepth=16 --allow_file_create=0 --size=$((size/njobs))M \
+ --filename=$DIR/$tfile
+ [ $? -eq 0 ] || error "fio mixed read write error"
+
+ rm -rf $DIR/$tfile
+ $LCTL set_param debug="$saved_debug"
+}
+run_test 398c "run fio to test AIO"
+
test_fake_rw() {
local read_write=$1
if [ "$read_write" = "write" ]; then
$LFS setstripe -c 1 -i 0 $DIR/$tfile
- # get ost1 size - lustre-OST0000
+ # get ost1 size - $FSNAME-OST0000
local ost1_avail_size=$($LFS df | awk /${ost1_svc}/'{ print $4 }')
local blocks=$((ost1_avail_size/2/1024)) # half avail space by megabytes
[ $blocks -gt 1000 ] && blocks=1000 # 1G in maximum
local prefix=/usr/include/lustre
local prog
+ # Oleg removes c files in his test rig so test if any c files exist
+ [ -z "$(ls -A $LUSTRE_TESTS_API_DIR)" ] && \
+ skip_env "Needed c test files are missing"
+
if ! [[ -d $prefix ]]; then
# Assume we're running in tree and fixup the include path.
extra_flags+=" -I$LUSTRE/../lnet/include/uapi -I$LUSTRE/include/uapi -I$LUSTRE/include"
fi
for prog in $LUSTRE_TESTS_API_DIR/*.c; do
- $CC -Wall -Werror $extra_flags -o $out $prog -llustreapi ||
+ $CC -Wall -Werror -std=c99 $extra_flags -o $out $prog -llustreapi ||
error "client api broken"
done
rm -f $out
continue # lustre_ioctl.h is internal header
fi
- $CC -Wall -Werror -include $header -c -x c /dev/null -o $out ||
+ $CC -Wall -Werror -std=c99 -include $header -c -x c /dev/null -o $out ||
error "cannot compile '$header'"
done
rm -f $out
test_412() {
[ $MDSCOUNT -lt 2 ] && skip_env "needs >= 2 MDTs"
- if [ $(lustre_version_code mds1) -lt $(version_code 2.10.55) ]; then
+ if [ $MDS1_VERSION -lt $(version_code 2.10.55) ]; then
skip "Need server version at least 2.10.55"
fi
}
run_test 412 "mkdir on specific MDTs"
-test_413a() {
- [ $MDSCOUNT -lt 2 ] &&
- skip "We need at least 2 MDTs for this test"
-
- if [ $(lustre_version_code mds1) -lt $(version_code 2.10.55) ]; then
- skip "Need server version at least 2.10.55"
- fi
-
- mkdir $DIR/$tdir || error "mkdir failed"
-
- # find MDT that is the most full
- local max=$($LFS df | grep MDT |
- awk 'BEGIN { a=0 }
- { sub("%", "", $5)
- if (0+$5 >= a)
- {
- a = $5
- b = $6
- }
- }
- END { split(b, c, ":")
- sub("]", "", c[2])
- print c[2]
- }')
-
- for i in $(seq $((MDSCOUNT - 1))); do
- $LFS mkdir -c $i $DIR/$tdir/d$i ||
- error "mkdir d$i failed"
- $LFS getdirstripe $DIR/$tdir/d$i
- local stripe_index=$($LFS getdirstripe -i $DIR/$tdir/d$i)
- [ $stripe_index -ne $max ] ||
- error "don't expect $max"
- done
-}
-run_test 413a "mkdir on less full MDTs"
-
-test_413b() {
- [ $MDSCOUNT -lt 2 ] &&
- skip "We need at least 2 MDTs for this test"
-
- [ $MDS1_VERSION -lt $(version_code 2.12.52) ] &&
- skip "Need server version at least 2.12.52"
-
- mkdir $DIR/$tdir || error "mkdir failed"
- $LFS setdirstripe -D -i -1 -H space $DIR/$tdir ||
- error "setdirstripe failed"
+test_qos_mkdir() {
+ local mkdir_cmd=$1
+ local stripe_count=$2
+ local mdts=$(comma_list $(mdts_nodes))
- local qos_prio_free
- local qos_threshold_rr
+ local testdir
+ local lmv_qos_prio_free
+ local lmv_qos_threshold_rr
+ local lmv_qos_maxage
+ local lod_qos_prio_free
+ local lod_qos_threshold_rr
+ local lod_qos_maxage
local count
+ local i
- qos_prio_free=$($LCTL get_param -n lmv.*.qos_prio_free | head -n1)
- qos_prio_free=${qos_prio_free%%%}
- qos_threshold_rr=$($LCTL get_param -n lmv.*.qos_threshold_rr | head -n1)
- qos_threshold_rr=${qos_threshold_rr%%%}
- qos_maxage=$($LCTL get_param -n lmv.*.qos_maxage)
-
- stack_trap "$LCTL set_param lmv.*.qos_prio_free=$qos_prio_free" EXIT
- stack_trap "$LCTL set_param lmv.*.qos_threshold_rr=$qos_threshold_rr" \
+ lmv_qos_prio_free=$($LCTL get_param -n lmv.*.qos_prio_free | head -n1)
+ lmv_qos_prio_free=${lmv_qos_prio_free%%%}
+ lmv_qos_threshold_rr=$($LCTL get_param -n lmv.*.qos_threshold_rr |
+ head -n1)
+ lmv_qos_threshold_rr=${lmv_qos_threshold_rr%%%}
+ lmv_qos_maxage=$($LCTL get_param -n lmv.*.qos_maxage)
+ stack_trap "$LCTL set_param \
+ lmv.*.qos_prio_free=$lmv_qos_prio_free > /dev/null" EXIT
+ stack_trap "$LCTL set_param \
+ lmv.*.qos_threshold_rr=$lmv_qos_threshold_rr > /dev/null" EXIT
+ stack_trap "$LCTL set_param \
+ lmv.*.qos_maxage=$lmv_qos_maxage > /dev/null" EXIT
+
+ lod_qos_prio_free=$(do_facet mds1 $LCTL get_param -n \
+ lod.$FSNAME-MDT0000-mdtlov.mdt_qos_prio_free | head -n1)
+ lod_qos_prio_free=${lod_qos_prio_free%%%}
+ lod_qos_threshold_rr=$(do_facet mds1 $LCTL get_param -n \
+ lod.$FSNAME-MDT0000-mdtlov.mdt_qos_threshold_rr | head -n1)
+ lod_qos_threshold_rr=${lod_qos_threshold_rr%%%}
+ lod_qos_maxage=$(do_facet mds1 $LCTL get_param -n \
+ lod.$FSNAME-MDT0000-mdtlov.qos_maxage | awk '{ print $1 }')
+ stack_trap "do_nodes $mdts $LCTL set_param \
+ lod.*.mdt_qos_prio_free=$lod_qos_prio_free > /dev/null" EXIT
+ stack_trap "do_nodes $mdts $LCTL set_param \
+ lod.*.mdt_qos_threshold_rr=$lod_qos_threshold_rr > /dev/null" \
EXIT
- stack_trap "$LCTL set_param lmv.*.qos_maxage=$qos_maxage" EXIT
+ stack_trap "do_nodes $mdts $LCTL set_param \
+ lod.*.mdt_qos_maxage=$lod_qos_maxage > /dev/null" EXIT
+
+ echo
+ echo "Mkdir (stripe_count $stripe_count) roundrobin:"
- echo "mkdir with roundrobin"
+ $LCTL set_param lmv.*.qos_threshold_rr=100 > /dev/null
+ do_nodes $mdts $LCTL set_param lod.*.mdt_qos_threshold_rr=100 > /dev/null
+
+ testdir=$DIR/$tdir-s$stripe_count/rr
- $LCTL set_param lmv.*.qos_threshold_rr=100
for i in $(seq $((100 * MDSCOUNT))); do
- mkdir $DIR/$tdir/subdir$i || error "mkdir subdir$i failed"
+ eval $mkdir_cmd $testdir/subdir$i ||
+ error "$mkdir_cmd subdir$i failed"
done
+
for i in $(seq $MDSCOUNT); do
- count=$($LFS getdirstripe -i $DIR/$tdir/* | grep ^$((i - 1))$ |
- wc -w)
+ count=$($LFS getdirstripe -i $testdir/* |
+ grep ^$((i - 1))$ | wc -l)
echo "$count directories created on MDT$((i - 1))"
[ $count -eq 100 ] || error "subdirs are not evenly distributed"
+
+ if [ $stripe_count -gt 1 ]; then
+ count=$($LFS getdirstripe $testdir/* |
+ grep -P "^\s+$((i - 1))\t" | wc -l)
+ echo "$count stripes created on MDT$((i - 1))"
+ # deviation should < 5% of average
+ [ $count -lt $((95 * stripe_count)) ] ||
+ [ $count -gt $((105 * stripe_count)) ] &&
+ error "stripes are not evenly distributed"
+ fi
done
- rm -rf $DIR/$tdir/*
+ $LCTL set_param lmv.*.qos_threshold_rr=$lmv_qos_threshold_rr > /dev/null
+ do_nodes $mdts $LCTL set_param \
+ lod.*.mdt_qos_threshold_rr=$lod_qos_threshold_rr > /dev/null
- $LCTL set_param lmv.*.qos_threshold_rr=$qos_threshold_rr
- # Shorten statfs result age, so that it can be updated in time
- $LCTL set_param lmv.*.qos_maxage=1
- sleep_maxage
+ echo
+ echo "Check for uneven MDTs: "
local ffree
local bavail
# Check if we need to generate uneven MDTs
local threshold=50
- local diff=$(((max - min ) * 100 / min))
+ local diff=$(((max - min) * 100 / min))
local value="$(generate_string 1024)"
- local i
while [ $diff -lt $threshold ]; do
# generate uneven MDTs, create till $threshold% diff
error "mkdir $tdir-MDT$min_index failed"
for i in $(seq $count); do
$OPENFILE -f O_CREAT:O_LOV_DELAY_CREATE \
- $DIR/$tdir-MDT$min_index/f$i > /dev/null ||
- error "create f$i failed"
+ $DIR/$tdir-MDT$min_index/f$j_$i > /dev/null ||
+ error "create f$j_$i failed"
setfattr -n user.413b -v $value \
- $DIR/$tdir-MDT$min_index/f$i ||
- error "setfattr f$i failed"
+ $DIR/$tdir-MDT$min_index/f$j_$i ||
+ error "setfattr f$j_$i failed"
done
ffree=($(lctl get_param -n mdc.*[mM][dD][cC]-*.filesfree))
echo "MDT blocks available: ${bavail[@]}"
echo "weight diff=$diff%"
- echo "mkdir with balanced space usage"
- $LCTL set_param lmv.*.qos_prio_free=100
+ echo
+ echo "Mkdir (stripe_count $stripe_count) with balanced space usage:"
+
+ $LCTL set_param lmv.*.qos_prio_free=100 > /dev/null
+ do_nodes $mdts $LCTL set_param lod.*.mdt_qos_prio_free=100 > /dev/null
+ # decrease statfs age, so that it can be updated in time
+ $LCTL set_param lmv.*.qos_maxage=1 > /dev/null
+ do_nodes $mdts $LCTL set_param lod.*.mdt_qos_maxage=1 > /dev/null
+
+ sleep 1
+
+ testdir=$DIR/$tdir-s$stripe_count/qos
+
for i in $(seq $((100 * MDSCOUNT))); do
- mkdir $DIR/$tdir/subdir$i || error "mkdir subdir$i failed"
+ eval $mkdir_cmd $testdir/subdir$i ||
+ error "$mkdir_cmd subdir$i failed"
done
for i in $(seq $MDSCOUNT); do
- count=$($LFS getdirstripe -i $DIR/$tdir/* | grep ^$((i - 1))$ |
- wc -w)
+ count=$($LFS getdirstripe -i $testdir/* | grep ^$((i - 1))$ |
+ wc -l)
echo "$count directories created on MDT$((i - 1))"
+
+ if [ $stripe_count -gt 1 ]; then
+ count=$($LFS getdirstripe $testdir/* |
+ grep -P "^\s+$((i - 1))\t" | wc -l)
+ echo "$count stripes created on MDT$((i - 1))"
+ fi
done
- max=$($LFS getdirstripe -i $DIR/$tdir/* | grep ^$max_index$ | wc -l)
- min=$($LFS getdirstripe -i $DIR/$tdir/* | grep ^$min_index$ | wc -l)
+ max=$($LFS getdirstripe -i $testdir/* | grep ^$max_index$ | wc -l)
+ min=$($LFS getdirstripe -i $testdir/* | grep ^$min_index$ | wc -l)
+ # D-value should > 10% of averge
[ $((max - min)) -lt 10 ] &&
error "subdirs shouldn't be evenly distributed"
- which getfattr > /dev/null 2>&1 || skip_env "no getfattr command"
+ # ditto
+ if [ $stripe_count -gt 1 ]; then
+ max=$($LFS getdirstripe $testdir/* |
+ grep -P "^\s+$max_index\t" | wc -l)
+ min=$($LFS getdirstripe $testdir/* |
+ grep -P "^\s+$min_index\t" | wc -l)
+ [ $((max - min)) -le $((10 * stripe_count)) ] &&
+ error "stripes shouldn't be evenly distributed"|| true
+ fi
+}
+
+test_413a() {
+ [ $MDSCOUNT -lt 2 ] &&
+ skip "We need at least 2 MDTs for this test"
+
+ [ $MDS1_VERSION -lt $(version_code 2.12.52) ] &&
+ skip "Need server version at least 2.12.52"
+
+ local stripe_count
+
+ for stripe_count in $(seq 1 $((MDSCOUNT - 1))); do
+ mkdir $DIR/$tdir-s$stripe_count || error "mkdir failed"
+ mkdir $DIR/$tdir-s$stripe_count/rr || error "mkdir failed"
+ mkdir $DIR/$tdir-s$stripe_count/qos || error "mkdir failed"
+ test_qos_mkdir "$LFS mkdir -c $stripe_count" $stripe_count
+ done
+}
+run_test 413a "QoS mkdir with 'lfs mkdir -i -1'"
+
+test_413b() {
+ [ $MDSCOUNT -lt 2 ] &&
+ skip "We need at least 2 MDTs for this test"
+
+ [ $MDS1_VERSION -lt $(version_code 2.12.52) ] &&
+ skip "Need server version at least 2.12.52"
- $LFS setdirstripe -D -d $DIR/$tdir || error "setdirstripe -d failed"
- getfattr -n trusted.dmv $DIR/$tdir &&
- error "default dir layout exists" || true
+ local stripe_count
+
+ for stripe_count in $(seq 1 $((MDSCOUNT - 1))); do
+ mkdir $DIR/$tdir-s$stripe_count || error "mkdir failed"
+ mkdir $DIR/$tdir-s$stripe_count/rr || error "mkdir failed"
+ mkdir $DIR/$tdir-s$stripe_count/qos || error "mkdir failed"
+ $LFS setdirstripe -D -c $stripe_count \
+ $DIR/$tdir-s$stripe_count/rr ||
+ error "setdirstripe failed"
+ $LFS setdirstripe -D -c $stripe_count \
+ $DIR/$tdir-s$stripe_count/qos ||
+ error "setdirstripe failed"
+ test_qos_mkdir "mkdir" $stripe_count
+ done
}
-run_test 413b "mkdir with balanced space usage"
+run_test 413b "QoS mkdir under dir whose default LMV starting MDT offset is -1"
test_414() {
#define OBD_FAIL_PTLRPC_BULK_ATTACH 0x521
test_415() {
[ $PARALLEL == "yes" ] && skip "skip parallel run"
- [ $(lustre_version_code mds1) -lt $(version_code 2.11.52) ] &&
+ [ $MDS1_VERSION -lt $(version_code 2.11.52) ] &&
skip "Need server version at least 2.11.52"
# LU-11102
run_test 415 "lock revoke is not missing"
test_416() {
- [ $(lustre_version_code mds1) -lt $(version_code 2.11.55) ] &&
+ [ $MDS1_VERSION -lt $(version_code 2.11.55) ] &&
skip "Need server version at least 2.11.55"
# define OBD_FAIL_OSD_TXN_START 0x19a
}
run_test 422 "kill a process with RPC in progress"
+stat_test() {
+ df -h $MOUNT &
+ df -h $MOUNT &
+ df -h $MOUNT &
+ df -h $MOUNT &
+ df -h $MOUNT &
+ df -h $MOUNT &
+}
+
+test_423() {
+ local _stats
+ # ensure statfs cache is expired
+ sleep 2;
+
+ _stats=$(stat_test | grep $MOUNT | sort -u | wc -l)
+ [[ ${_stats} -ne 1 ]] && error "statfs wrong"
+
+ return 0
+}
+run_test 423 "statfs should return a right data"
+
prep_801() {
- [[ $(lustre_version_code mds1) -lt $(version_code 2.9.55) ]] ||
+ [[ $MDS1_VERSION -lt $(version_code 2.9.55) ]] ||
[[ $OST1_VERSION -lt $(version_code 2.9.55) ]] &&
skip "Need server version at least 2.9.55"
test_802a() {
[[ $mds1_FSTYPE = zfs ]] || skip "ZFS specific test"
- [[ $(lustre_version_code mds1) -lt $(version_code 2.9.55) ]] ||
+ [[ $MDS1_VERSION -lt $(version_code 2.9.55) ]] ||
[[ $OST1_VERSION -lt $(version_code 2.9.55) ]] &&
skip "Need server version at least 2.9.55"
}
test_805() {
- local zfs_version=$(do_node $SINGLEMDS cat /sys/module/zfs/version)
+ local zfs_version=$(do_facet mds1 cat /sys/module/zfs/version)
[ "$mds1_FSTYPE" != "zfs" ] && skip "ZFS specific test"
[ $(version_code $zfs_version) -lt $(version_code 0.7.2) ] &&
skip "netfree not implemented before 0.7"
local usedkb
local old
local quota
- local pref="osd-zfs.lustre-MDT0000."
+ local pref="osd-zfs.$FSNAME-MDT0000."
# limit available space on MDS dataset to meet nospace issue
# quickly. then ZFS 0.7.2 can use reserved space if asked
error "truncate $tdir/trunc failed"
local bs=1048576
- dd if=/dev/zero of=$DIR/$tdir/single_dd bs=$bs count=1 ||
+ dd if=/dev/zero of=$DIR/$tdir/single_dd bs=$bs count=1 conv=fsync ||
error "write $tfile failed"
# multi-client wirtes
wait ${pids[$i]}
done
- sleep 5
+ do_rpc_nodes "$CLIENTS" cancel_lru_locks osc
+ do_nodes "$CLIENTS" "sync ; sleep 5 ; sync"
$LSOM_SYNC -u $cl_user -m $FSNAME-MDT0000 $MOUNT
check_lsom_data $DIR/$tdir/trunc
check_lsom_data $DIR/$tdir/single_dd
}
run_test 810 "partial page writes on ZFS (LU-11663)"
-test_811() {
- [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.11.56) ] &&
- skip "Need MDS version at least 2.11.56"
+test_812a() {
+ [ $OST1_VERSION -lt $(version_code 2.12.51) ] &&
+ skip "OST < 2.12.51 doesn't support this fail_loc"
+ [ "$SHARED_KEY" = true ] &&
+ skip "OSC connections never go IDLE with Shared-Keys enabled"
- #define OBD_FAIL_MDS_ORPHAN_DELETE 0x165
- do_facet mds1 $LCTL set_param fail_loc=0x165
- $MULTIOP $DIR/$tfile Ouc || error "multiop failed"
+ $LFS setstripe -c 1 -i 0 $DIR/$tfile
+ # ensure ost1 is connected
+ stat $DIR/$tfile >/dev/null || error "can't stat"
+ wait_osc_import_state client ost1 FULL
+ # no locks, no reqs to let the connection idle
+ cancel_lru_locks osc
- stop mds1
- start mds1 $(mdsdevname 1) $MDS_MOUNT_OPTS
+ # delay OST_DISCONNECT on OST1 to put OSC into intermediate state
+#define OBD_FAIL_OST_DISCONNECT_DELAY 0x245
+ do_facet ost1 "$LCTL set_param fail_loc=0x245 fail_val=8"
+ wait_osc_import_state client ost1 CONNECTING
+ do_facet ost1 "$LCTL set_param fail_loc=0 fail_val=0"
- sleep 5
- [[ $(do_facet mds1 pgrep orph_.*-MDD | wc -l) -eq 0 ]] ||
- error "MDD orphan cleanup thread not quit"
+ stat $DIR/$tfile >/dev/null || error "can't stat file"
}
-run_test 811 "orphan name stub can be cleaned up in startup"
+run_test 812a "do not drop reqs generated when imp is going to idle (LU-11951)"
-test_812() {
+test_812b() { # LU-12378
[ $OST1_VERSION -lt $(version_code 2.12.51) ] &&
skip "OST < 2.12.51 doesn't support this fail_loc"
[ "$SHARED_KEY" = true ] &&
skip "OSC connections never go IDLE with Shared-Keys enabled"
- $LFS setstripe -c 1 -i 0 $DIR/$tfile
+ $LFS setstripe -c 1 -i 0 $DIR/$tfile || error "setstripe failed"
# ensure ost1 is connected
stat $DIR/$tfile >/dev/null || error "can't stat"
wait_osc_import_state client ost1 FULL
wait_osc_import_state client ost1 CONNECTING
do_facet ost1 "$LCTL set_param fail_loc=0 fail_val=0"
- stat $DIR/$tfile >/dev/null || error "can't stat file"
+ $LFS quota -u 0 $DIR/ || error "lfs quota should succeed"
+ wait_osc_import_state client ost1 IDLE
}
-run_test 812 "do not drop reqs generated when imp is going to idle (LU-11951)"
+run_test 812b "do not drop no resend request for idle connect"
test_813() {
local file_heat_sav=$($LCTL get_param -n llite.*.file_heat 2>/dev/null)
}
run_test 819b "too big niobuf in write"
+
+function test_820_start_ost() {
+ sleep 5
+
+ for num in $(seq $OSTCOUNT); do
+ start ost$num $(ostdevname $num) $OST_MOUNT_OPTS
+ done
+}
+
+test_820() {
+ [[ $MDSCOUNT -lt 2 ]] && skip_env "needs >= 2 MDTs"
+
+ mkdir $DIR/$tdir
+ umount_client $MOUNT || error "umount failed"
+ for num in $(seq $OSTCOUNT); do
+ stop ost$num
+ done
+
+ # mount client with no active OSTs
+ # so that the client can't initialize max LOV EA size
+ # from OSC notifications
+ mount_client $MOUNT || error "mount failed"
+ # delay OST starting to keep this 0 max EA size for a while
+ test_820_start_ost &
+
+ # create a directory on MDS2
+ test_mkdir -i 1 -c1 $DIR/$tdir/mds2 ||
+ error "Failed to create directory"
+ # open intent should update default EA size
+ # see mdc_update_max_ea_from_body()
+ # notice this is the very first RPC to MDS2
+ cp /etc/services $DIR/$tdir/mds2 ||
+ error "Failed to copy files to mds$n"
+}
+run_test 820 "update max EA from open intent"
+
#
# tests that do cleanup/setup should be run at the end
#
}
run_test 900 "umount should not race with any mgc requeue thread"
+# LUS-6253/LU-11185
+test_901() {
+ local oldc
+ local newc
+ local olds
+ local news
+ [ $PARALLEL == "yes" ] && skip "skip parallel run"
+
+ # some get_param have a bug to handle dot in param name
+ cancel_lru_locks MGC
+ oldc=$($LCTL get_param -n 'ldlm.namespaces.MGC*.lock_count')
+ olds=$(do_facet mgs $LCTL get_param -n 'ldlm.namespaces.MGS*.lock_count')
+ umount_client $MOUNT || error "umount failed"
+ mount_client $MOUNT || error "mount failed"
+ cancel_lru_locks MGC
+ newc=$($LCTL get_param -n 'ldlm.namespaces.MGC*.lock_count')
+ news=$(do_facet mgs $LCTL get_param -n 'ldlm.namespaces.MGS*.lock_count')
+
+ [ $oldc -lt $newc ] && error "mgc lock leak ($oldc != $newc)"
+ [ $olds -lt $news ] && error "mgs lock leak ($olds != $news)"
+
+ return 0
+}
+run_test 901 "don't leak a mgc lock on client umount"
+
+# LU-13377
+test_902() {
+ [ $CLIENT_VERSION -lt $(version_code 2.13.52) ] &&
+ skip "client does not have LU-13377 fix"
+ #define OBD_FAIL_LLITE_SHORT_COMMIT 0x1415
+ $LCTL set_param fail_loc=0x1415
+ dd if=/dev/zero of=$DIR/$tfile bs=1M count=1
+ cancel_lru_locks osc
+ rm -f $DIR/$tfile
+}
+run_test 902 "test short write doesn't hang lustre"
+
complete $SECONDS
[ -f $EXT2_DEV ] && rm $EXT2_DEV || true
check_and_cleanup_lustre
git://git.whamcloud.com / fs / lustre-release.git / blobdiff