ALWAYS_EXCEPT+=" 411"
fi
-# 5 12 8 12 (min)"
-[ "$SLOW" = "no" ] && EXCEPT_SLOW="27m 60i 64b 68 71 115 135 136 300o"
+# 5 12 8 12 15 (min)"
+[ "$SLOW" = "no" ] && EXCEPT_SLOW="27m 60i 64b 68 71 115 135 136 230d 300o"
if [ "$mds1_FSTYPE" = "zfs" ]; then
# bug number for skipped test:
}
run_test 31q "create striped directory on specific MDTs"
+#LU-14949
+test_31r() {
+ touch $DIR/$tfile.target
+ touch $DIR/$tfile.source
+
+ #OBD_FAIL_LLITE_OPEN_DELAY 0x1419
+ $LCTL set_param fail_loc=0x1419 fail_val=3
+ cat $DIR/$tfile.target &
+ CATPID=$!
+
+ # Guarantee open is waiting before we get here
+ sleep 1
+ mv $DIR/$tfile.source $DIR/$tfile.target
+
+ wait $CATPID
+ RC=$?
+ if [[ $RC -ne 0 ]]; then
+ error "open with cat failed, rc=$RC"
+ fi
+}
+run_test 31r "open-rename(replace) race"
+
cleanup_test32_mount() {
local rc=0
trap 0
test_51d() {
[ $PARALLEL == "yes" ] && skip "skip parallel run"
[[ $OSTCOUNT -lt 3 ]] && skip_env "needs >= 3 OSTs"
+ local qos_old
test_mkdir $DIR/$tdir
+ $LFS setstripe -c $OSTCOUNT $DIR/$tdir
+
+ qos_old=$(do_facet mds1 \
+ "$LCTL get_param -n lod.$FSNAME-*.qos_threshold_rr" | head -n 1)
+ do_nodes $(comma_list $(mdts_nodes)) \
+ "$LCTL set_param lod.$FSNAME-*.qos_threshold_rr=100"
+ stack_trap "do_nodes $(comma_list $(mdts_nodes)) \
+ '$LCTL set_param lod.$FSNAME-*.qos_threshold_rr=${qos_old%%%}'"
+
createmany -o $DIR/$tdir/t- 1000
$LFS getstripe $DIR/$tdir > $TMP/$tfile
- for N in $(seq 0 $((OSTCOUNT - 1))); do
- OBJS[$N]=$(awk -vobjs=0 '($1 == '$N') { objs += 1 } \
- END { printf("%0.0f", objs) }' $TMP/$tfile)
- OBJS0[$N]=$(grep -A 1 idx $TMP/$tfile | awk -vobjs=0 \
- '($1 == '$N') { objs += 1 } \
- END { printf("%0.0f", objs) }')
- log "OST$N has ${OBJS[$N]} objects, ${OBJS0[$N]} are index 0"
+ for ((n = 0; n < $OSTCOUNT; n++)); do
+ objs[$n]=$(awk -vobjs=0 '($1 == '$n') { objs += 1 } \
+ END { printf("%0.0f", objs) }' $TMP/$tfile)
+ objs0[$n]=$(grep -A 1 idx $TMP/$tfile | awk -vobjs=0 \
+ '($1 == '$n') { objs += 1 } \
+ END { printf("%0.0f", objs) }')
+ log "OST$n has ${objs[$n]} objects, ${objs0[$n]} are index 0"
done
unlinkmany $DIR/$tdir/t- 1000
- NLAST=0
- for N in $(seq 1 $((OSTCOUNT - 1))); do
- [[ ${OBJS[$N]} -lt $((${OBJS[$NLAST]} - 20)) ]] &&
- error "OST $N has less objects vs OST $NLAST" \
- " (${OBJS[$N]} < ${OBJS[$NLAST]}"
- [[ ${OBJS[$N]} -gt $((${OBJS[$NLAST]} + 20)) ]] &&
- error "OST $N has less objects vs OST $NLAST" \
- " (${OBJS[$N]} < ${OBJS[$NLAST]}"
-
- [[ ${OBJS0[$N]} -lt $((${OBJS0[$NLAST]} - 20)) ]] &&
- error "OST $N has less #0 objects vs OST $NLAST" \
- " (${OBJS0[$N]} < ${OBJS0[$NLAST]}"
- [[ ${OBJS0[$N]} -gt $((${OBJS0[$NLAST]} + 20)) ]] &&
- error "OST $N has less #0 objects vs OST $NLAST" \
- " (${OBJS0[$N]} < ${OBJS0[$NLAST]}"
- NLAST=$N
+ nlast=0
+ for ((n = 0; n < $OSTCOUNT; n++)); do
+ (( ${objs[$n]} > ${objs[$nlast]} * 4 / 5 )) ||
+ { $LFS df && $LFS df -i &&
+ error "OST $n has fewer objects vs. OST $nlast" \
+ " (${objs[$n]} < ${objs[$nlast]}"; }
+ (( ${objs[$n]} < ${objs[$nlast]} * 5 / 4 )) ||
+ { $LFS df && $LFS df -i &&
+ error "OST $n has fewer objects vs. OST $nlast" \
+ " (${objs[$n]} < ${objs[$nlast]}"; }
+
+ (( ${objs0[$n]} > ${objs0[$nlast]} * 4 / 5 )) ||
+ { $LFS df && $LFS df -i &&
+ error "OST $n has fewer #0 objects vs. OST $nlast" \
+ " (${objs0[$n]} < ${objs0[$nlast]}"; }
+ (( ${objs0[$n]} < ${objs0[$nlast]} * 5 / 4 )) ||
+ { $LFS df && $LFS df -i &&
+ error "OST $n has fewer #0 objects vs. OST $nlast" \
+ " (${objs0[$n]} < ${objs0[$nlast]}"; }
+ nlast=$n
done
- rm -f $TMP/$tfile
}
run_test 51d "check object distribution"
}
run_test 56d "'lfs df -v' prints only configured devices"
+test_56e() {
+ err_enoent=2 # No such file or directory
+ err_eopnotsupp=95 # Operation not supported
+
+ enoent_mnt=/pmt1 # Invalid dentry. Path not present
+ notsup_mnt=/tmp # Valid dentry, but Not a lustreFS
+
+ # Check for handling of path not exists
+ output=$($LFS df $enoent_mnt 2>&1)
+ ret=$?
+
+ fs=$(echo $output | awk -F: '{print $2}' | awk '{print $3}' | tr -d \')
+ [[ $fs = $enoent_mnt && $ret -eq $err_enoent ]] ||
+ error "expect failure $err_enoent, not $ret"
+
+ # Check for handling of non-Lustre FS
+ output=$($LFS df $notsup_mnt)
+ ret=$?
+
+ fs=$(echo $output | awk '{print $1}' | awk -F: '{print $2}')
+ [[ $fs = $notsup_mnt && $ret -eq $err_eopnotsupp ]] ||
+ error "expect success $err_eopnotsupp, not $ret"
+
+ # Check for multiple LustreFS argument
+ output=$($LFS df $MOUNT $MOUNT $MOUNT | grep -c "filesystem_summary:")
+ ret=$?
+
+ [[ $output -eq 3 && $ret -eq 0 ]] ||
+ error "expect success 3, not $output, rc = $ret"
+
+ # Check for correct non-Lustre FS handling among multiple
+ # LustreFS argument
+ output=$($LFS df $MOUNT $notsup_mnt $MOUNT |
+ grep -c "filesystem_summary:"; exit ${PIPESTATUS[0]})
+ ret=$?
+
+ [[ $output -eq 2 && $ret -eq $err_eopnotsupp ]] ||
+ error "expect success 2, not $output, rc = $ret"
+}
+run_test 56e "'lfs df' Handle non LustreFS & multiple LustreFS"
+
NUMFILES=3
NUMDIRS=3
setup_56() {
}
run_test 56xf "FID is not lost during migration of a composite layout file"
+check_file_ost_range() {
+ local file="$1"
+ shift
+ local range="$*"
+ local -a file_range
+ local idx
+
+ file_range=($($LFS getstripe -y "$file" |
+ awk '/l_ost_idx:/ { print $NF }'))
+
+ if [[ "${#file_range[@]}" = 0 ]]; then
+ echo "No osts found for $file"
+ return 1
+ fi
+
+ for idx in "${file_range[@]}"; do
+ [[ " $range " =~ " $idx " ]] ||
+ return 1
+ done
+
+ return 0
+}
+
+sub_test_56xg() {
+ local stripe_opt="$1"
+ local pool="$2"
+ shift 2
+ local pool_ostidx="$(seq $* | tr '\n' ' ')"
+
+ $LFS migrate $stripe_opt -p $pool $DIR/$tfile ||
+ error "Fail to migrate $tfile on $pool"
+ [[ "$($LFS getstripe -p $DIR/$tfile)" = "$pool" ]] ||
+ error "$tfile is not in pool $pool"
+ check_file_ost_range "$DIR/$tfile" $pool_ostidx ||
+ error "$tfile osts mismatch with pool $pool (osts $pool_ostidx)"
+}
+
+test_56xg() {
+ [[ $PARALLEL != "yes" ]] || skip "skip parallel run"
+ [[ $OSTCOUNT -ge 2 ]] || skip "needs >= 2 OSTs"
+ [[ $MDS1_VERSION -gt $(version_code 2.14.52) ]] ||
+ skip "Need MDS version newer than 2.14.52"
+
+ local -a pool_names=("${TESTNAME}_0" "${TESTNAME}_1" "${TESTNAME}_2")
+ local -a pool_ranges=("0 0" "1 1" "0 1")
+
+ # init pools
+ for i in "${!pool_names[@]}"; do
+ pool_add ${pool_names[$i]} ||
+ error "pool_add failed (pool: ${pool_names[$i]})"
+ pool_add_targets ${pool_names[$i]} ${pool_ranges[$i]} ||
+ error "pool_add_targets failed (pool: ${pool_names[$i]})"
+ done
+
+ # init the file to migrate
+ $LFS setstripe -c1 -i1 $DIR/$tfile ||
+ error "Unable to create $tfile on OST1"
+ dd if=/dev/urandom of=$DIR/$tfile bs=1M count=4 status=none ||
+ error "Unable to write on $tfile"
+
+ echo "1. migrate $tfile on pool ${pool_names[0]}"
+ sub_test_56xg "-c-1" "${pool_names[0]}" ${pool_ranges[0]}
+
+ echo "2. migrate $tfile on pool ${pool_names[2]}"
+ sub_test_56xg "-c-1 -S2M" "${pool_names[2]}" ${pool_ranges[2]}
+
+ echo "3. migrate $tfile on pool ${pool_names[1]}"
+ sub_test_56xg "-n -c-1" "${pool_names[1]}" ${pool_ranges[1]}
+
+ echo "4. migrate $tfile on pool ${pool_names[2]} with default stripe parameters"
+ sub_test_56xg "" "${pool_names[2]}" ${pool_ranges[2]}
+ echo
+
+ # Clean pools
+ destroy_test_pools ||
+ error "pool_destroy failed"
+}
+run_test 56xg "lfs migrate pool support"
+
test_56y() {
[ $MDS1_VERSION -lt $(version_code 2.4.53) ] &&
skip "No HSM $(lustre_build_version $SINGLEMDS) MDS < 2.4.53"
}
run_test 77n "Verify read from a hole inside contiguous blocks with T10PI"
+test_77o() {
+ (( $CLIENT_VERSION >= $(version_code 2.14.54) )) ||
+ skip "Need at least version 2.14.54"
+ local ofd=obdfilter
+ local mdt=mdt
+
+ # print OST checksum_type
+ echo "$ofd.$FSNAME-*.checksum_type:"
+ do_nodes $(comma_list $(osts_nodes)) \
+ $LCTL get_param -n $ofd.$FSNAME-*.checksum_type
+
+ # print MDT checksum_type
+ echo "$mdt.$FSNAME-*.checksum_type:"
+ do_nodes $(comma_list $(mdts_nodes)) \
+ $LCTL get_param -n $mdt.$FSNAME-*.checksum_type
+
+ local o_count=$(do_nodes $(comma_list $(osts_nodes)) \
+ $LCTL get_param -n $ofd.$FSNAME-*.checksum_type | wc -l)
+
+ (( $o_count == $OSTCOUNT )) ||
+ error "found $o_count checksums, not \$MDSCOUNT=$OSTCOUNT"
+
+ local m_count=$(do_nodes $(comma_list $(mdts_nodes)) \
+ $LCTL get_param -n $mdt.$FSNAME-*.checksum_type | wc -l)
+
+ (( $m_count == $MDSCOUNT )) ||
+ error "found $m_count checksums, not \$MDSCOUNT=$MDSCOUNT"
+}
+run_test 77o "Verify checksum_type for server (mdt and ofd(obdfilter))"
+
cleanup_test_78() {
trap 0
rm -f $DIR/$tfile
changelog_register --user test_160o -m unlnk+close+open ||
error "changelog_register failed"
- # drop server mask so it doesn't interfere
+
do_facet $SINGLEMDS $LCTL --device $mdt \
changelog_register -u "Tt3_-#" &&
error "bad symbols in name should fail"
}
run_test 160p "Changelog orphan cleanup with no users"
+test_160q() {
+ local mdt="$(facet_svc $SINGLEMDS)"
+ local clu
+
+ [[ $PARALLEL != "yes" ]] || skip "skip parallel run"
+ remote_mds_nodsh && skip "remote MDS with nodsh"
+ [ $MDS1_VERSION -ge $(version_code 2.14.54) ] ||
+ skip "Need MDS version at least 2.14.54"
+
+ # set server mask to minimal value like server init does
+ changelog_chmask "MARK"
+ clu=$(do_facet $SINGLEMDS $LCTL --device $mdt changelog_register -n) ||
+ error "changelog_register failed"
+ # check effective mask again, should be treated as DEFMASK now
+ mask=$(do_facet $SINGLEMDS $LCTL get_param \
+ mdd.$mdt.changelog_current_mask -n)
+ do_facet $SINGLEMDS $LCTL --device $mdt changelog_deregister $clu ||
+ error "changelog_deregister failed"
+ [[ $mask == *"HLINK"* ]] || error "mask is not DEFMASK as expected"
+}
+run_test 160q "changelog effective mask is DEFMASK if not set"
+
test_161a() {
[ $PARALLEL == "yes" ] && skip "skip parallel run"
$MULTIOP $DIR/$tfile oO_CREAT:O_RDWR:eRE+eU || error "get lease error"
echo "==== test 2: verify lease can be broken by upcoming open"
- $MULTIOP $DIR/$tfile oO_RDONLY:eR_E-eUc &
+ $MULTIOP $DIR/$tfile oO_RDWR:eR_E-eUc &
local PID=$!
sleep 1
- $MULTIOP $DIR/$tfile oO_RDONLY:c
+ $MULTIOP $DIR/$tfile oO_RDWR:c
kill -USR1 $PID && wait $PID || error "break lease error"
echo "==== test 3: verify lease can't be granted if an open already exists"
- $MULTIOP $DIR/$tfile oO_RDONLY:_c &
+ $MULTIOP $DIR/$tfile oO_RDWR:_c &
local PID=$!
sleep 1
- $MULTIOP $DIR/$tfile oO_RDONLY:eReUc && error "apply lease should fail"
+ $MULTIOP $DIR/$tfile oO_RDWR:eReUc && error "apply lease should fail"
kill -USR1 $PID && wait $PID || error "open file error"
echo "==== test 4: lease can sustain over recovery"
- $MULTIOP $DIR/$tfile oO_RDONLY:eR_E+eUc &
+ $MULTIOP $DIR/$tfile oO_RDWR:eR_E+eUc &
PID=$!
sleep 1
kill -USR1 $PID && wait $PID || error "lease broken over recovery"
echo "==== test 5: lease broken can't be regained by replay"
- $MULTIOP $DIR/$tfile oO_RDONLY:eR_E-eUc &
+ $MULTIOP $DIR/$tfile oO_RDWR:eR_E-eUc &
PID=$!
sleep 1
test_224a() { # LU-1039, MRP-303
[ $PARALLEL == "yes" ] && skip "skip parallel run"
-
#define OBD_FAIL_PTLRPC_CLIENT_BULK_CB 0x508
$LCTL set_param fail_loc=0x508
- dd if=/dev/zero of=$DIR/$tfile bs=4096 count=1 conv=fsync
+ dd if=/dev/zero of=$DIR/$tfile bs=1M count=1 conv=fsync
$LCTL set_param fail_loc=0
df $DIR
}
run_test 224a "Don't panic on bulk IO failure"
-test_224b() { # LU-1039, MRP-303
+test_224bd_sub() { # LU-1039, MRP-303
[ $PARALLEL == "yes" ] && skip "skip parallel run"
+ local timeout=$1
- dd if=/dev/zero of=$DIR/$tfile bs=4096 count=1
+ shift
+ dd if=/dev/urandom of=$TMP/$tfile bs=1M count=1
+
+ $LFS setstripe -c 1 -i 0 $DIR/$tfile
+
+ dd if=$TMP/$tfile of=$DIR/$tfile bs=1M count=1
cancel_lru_locks osc
+ set_checksums 0
+ stack_trap "set_checksums $ORIG_CSUM" EXIT
+ local at_max_saved=0
+
+ # adaptive timeouts may prevent seeing the issue
+ if at_is_enabled; then
+ at_max_saved=$(at_max_get mds)
+ at_max_set 0 mds client
+ stack_trap "at_max_set $at_max_saved mds client" EXIT
+ fi
+
#define OBD_FAIL_PTLRPC_CLIENT_BULK_CB2 0x515
- $LCTL set_param fail_loc=0x515
- dd of=/dev/null if=$DIR/$tfile bs=4096 count=1
- $LCTL set_param fail_loc=0
+ do_facet ost1 $LCTL set_param fail_val=$timeout fail_loc=0x80000515
+ dd of=$TMP/$tfile.new if=$DIR/$tfile bs=1M count=1 || "$@"
+
+ do_facet ost1 $LCTL set_param fail_loc=0
+ cmp $TMP/$tfile $TMP/$tfile.new || error "file contents wrong"
df $DIR
}
+
+test_224b() {
+ test_224bd_sub 3 error "dd failed"
+}
run_test 224b "Don't panic on bulk IO failure"
test_224c() { # LU-6441
}
run_test 224c "Don't hang if one of md lost during large bulk RPC"
+test_224d() { # LU-11169
+ test_224bd_sub $((TIMEOUT + 2)) error "dd failed"
+}
+run_test 224d "Don't corrupt data on bulk IO timeout"
+
MDSSURVEY=${MDSSURVEY:-$(which mds-survey 2>/dev/null || true)}
test_225a () {
[ $PARALLEL == "yes" ] && skip "skip parallel run"
[ "$1" == "blocks" ] && inodes= || inodes="-i"
for count in {1..100}; do
- cancel_lru_locks
+ do_rpc_nodes "$CLIENTS" cancel_lru_locks
sync; sleep 0.2
# read the lines of interest
}
run_test 902 "test short write doesn't hang lustre"
+# LU-14711
+test_903() {
+ $LFS setstripe -i 0 -c 1 $DIR/$tfile $DIR/${tfile}-2
+ echo "blah" > $DIR/${tfile}-2
+ dd if=/dev/zero of=$DIR/$tfile bs=1M count=6 conv=fsync
+ #define OBD_FAIL_OSC_SLOW_PAGE_EVICT 0x417
+ $LCTL set_param fail_loc=0x417 fail_val=20
+
+ mv $DIR/${tfile}-2 $DIR/$tfile # Destroys the big object
+ sleep 1 # To start the destroy
+ wait_destroy_complete 150 || error "Destroy taking too long"
+ cat $DIR/$tfile > /dev/null || error "Evicted"
+}
+run_test 903 "Test long page discard does not cause evictions"
+
complete $SECONDS
[ -f $EXT2_DEV ] && rm $EXT2_DEV || true
check_and_cleanup_lustre
git://git.whamcloud.com / fs / lustre-release.git / blobdiff