X-Git-Url: https://git.whamcloud.com/?p=fs%2Flustre-release.git;a=blobdiff_plain;f=lustre%2Ftests%2Fsanity.sh;h=c3aa3850a3d1b0d811db5747a628884bacd7e74d;hp=2b02c2f1881ca5a0f1680feced9b8eaeab9f7f65;hb=4b0f0164c6ed761897409186376e9edc989323c9;hpb=cedbb25e984ceb85a66bc5a315fbfa05c5bcb423 diff --git a/lustre/tests/sanity.sh b/lustre/tests/sanity.sh index 2b02c2f..c3aa385 100644 --- a/lustre/tests/sanity.sh +++ b/lustre/tests/sanity.sh @@ -46,23 +46,35 @@ ALWAYS_EXCEPT+=" 407 312 " 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-11729 LU-4398 - ALWAYS_EXCEPT+=" 45 317 810 817" + # 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 $(version_code 4.14.0) ];then + # bug number: LU-12661 + ALWAYS_EXCEPT+=" 817" fi # 5 12 (min)" -[ "$SLOW" = "no" ] && EXCEPT_SLOW="27m 64b 68 71 115 300o" +[ "$SLOW" = "no" ] && EXCEPT_SLOW="27m 64b 68 71 115 135 136 300o" if [ "$mds1_FSTYPE" = "zfs" ]; then - # bug number for skipped test: LU-1957 - ALWAYS_EXCEPT="$ALWAYS_EXCEPT 180" + # bug number for skipped test: + ALWAYS_EXCEPT="$ALWAYS_EXCEPT " # 13 (min)" [ "$SLOW" = "no" ] && EXCEPT_SLOW="$EXCEPT_SLOW 51b" fi @@ -2458,10 +2470,6 @@ test_27D() { local ost_list=$(seq $first_ost $ost_step $last_ost) local ost_range="$first_ost $last_ost $ost_step" - if ! combined_mgs_mds ; then - mount_mgs_client - fi - test_mkdir $DIR/$tdir pool_add $POOL || error "pool_add failed" pool_add_targets $POOL $ost_range || error "pool_add_targets failed" @@ -2472,19 +2480,15 @@ test_27D() { [ $MDS1_VERSION -lt $(version_code 2.9.55) ] || [ $CLIENT_VERSION -lt $(version_code 2.9.55) ] && skip27D+=" -s 30,31" - [ $(lustre_version_code $SINGLEMDS) -lt $(version_code $SEL_VER) ] && - skip27D+="-s 32" [[ ! $($LCTL get_param mdc.*.import) =~ connect_flags.*overstriping || $OSTCOUNT -ge $(($LOV_MAX_STRIPE_COUNT / 2)) ]] && skip27D+=" -s 32,33" + [[ $MDS_VERSION -lt $(version_code $SEL_VER) ]] && + skip27D+=" -s 34" llapi_layout_test -d$DIR/$tdir -p$POOL -o$OSTCOUNT $skip27D || error "llapi_layout_test failed" destroy_test_pools || error "destroy test pools failed" - - if ! combined_mgs_mds ; then - umount_mgs_client - fi } run_test 27D "validate llapi_layout API" @@ -2620,6 +2624,8 @@ run_test 27H "Set specific OSTs stripe" test_27I() { [ $PARALLEL == "yes" ] && skip "skip parallel run" [[ $OSTCOUNT -lt 2 ]] && skip_env "needs >= 2 OSTs" + [[ $MDS1_VERSION -gt $(version_code 2.12.52) ]] || + skip "Need MDS version newer than 2.12.52" local pool=$TESTNAME local ostrange="1 1 1" @@ -2815,11 +2821,6 @@ test_27L() { local POOL=${POOL:-$TESTNAME} - if ! combined_mgs_mds ; then - mount_mgs_client - trap umount_mgs_client EXIT - fi - pool_add $POOL || error "pool_add failed" lfs pool_list $MOUNT | grep -Fx "${FSNAME}.${POOL}" || @@ -2828,6 +2829,152 @@ test_27L() { } run_test 27L "lfs pool_list gives correct pool name" +test_27M() { + [[ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.12.57) ]] && + skip "Need MDS version >= than 2.12.57" + remote_mds_nodsh && skip "remote MDS with nodsh" + [[ $OSTCOUNT -lt 2 ]] && skip_env "need > 1 OST" + + test_mkdir $DIR/$tdir + + # Set default striping on directory + $LFS setstripe -C 4 $DIR/$tdir + + echo 1 > $DIR/$tdir/${tfile}.1 + local count=$($LFS getstripe -c $DIR/$tdir/${tfile}.1) + local setcount=4 + [ $count -eq $setcount ] || + error "(1) stripe count $count, should be $setcount" + + # Capture existing append_stripe_count setting for restore + local orig_count=$(do_facet mds1 $LCTL get_param -n mdd.$FSNAME-MDT0000.append_stripe_count) + local mdts=$(comma_list $(mdts_nodes)) + stack_trap "do_nodes $mdts $LCTL set_param mdd.*.append_stripe_count=$orig_count" EXIT + + local appendcount=$orig_count + echo 1 >> $DIR/$tdir/${tfile}.2_append + count=$($LFS getstripe -c $DIR/$tdir/${tfile}.2_append) + [ $count -eq $appendcount ] || + error "(2)stripe count $count, should be $appendcount for append" + + # Disable O_APPEND striping, verify it works + do_nodes $mdts $LCTL set_param mdd.*.append_stripe_count=0 + + # Should now get the default striping, which is 4 + setcount=4 + echo 1 >> $DIR/$tdir/${tfile}.3_append + count=$($LFS getstripe -c $DIR/$tdir/${tfile}.3_append) + [ $count -eq $setcount ] || + error "(3) stripe count $count, should be $setcount" + + # Try changing the stripe count for append files + do_nodes $mdts $LCTL set_param mdd.*.append_stripe_count=2 + + # Append striping is now 2 (directory default is still 4) + appendcount=2 + echo 1 >> $DIR/$tdir/${tfile}.4_append + count=$($LFS getstripe -c $DIR/$tdir/${tfile}.4_append) + [ $count -eq $appendcount ] || + error "(4) stripe count $count, should be $appendcount for append" + + # Test append stripe count of -1 + do_nodes $mdts $LCTL set_param mdd.*.append_stripe_count=-1 + appendcount=$OSTCOUNT + echo 1 >> $DIR/$tdir/${tfile}.5 + count=$($LFS getstripe -c $DIR/$tdir/${tfile}.5) + [ $count -eq $appendcount ] || + error "(5) stripe count $count, should be $appendcount for append" + + # Set append striping back to default of 1 + do_nodes $mdts $LCTL set_param mdd.*.append_stripe_count=1 + + # Try a new default striping, PFL + DOM + $LFS setstripe -L mdt -E 1M -E -1 -c 2 $DIR/$tdir + + # Create normal DOM file, DOM returns stripe count == 0 + setcount=0 + touch $DIR/$tdir/${tfile}.6 + count=$($LFS getstripe -c $DIR/$tdir/${tfile}.6) + [ $count -eq $setcount ] || + error "(6) stripe count $count, should be $setcount" + + # Show + appendcount=1 + echo 1 >> $DIR/$tdir/${tfile}.7_append + count=$($LFS getstripe -c $DIR/$tdir/${tfile}.7_append) + [ $count -eq $appendcount ] || + error "(7) stripe count $count, should be $appendcount for append" + + # Clean up DOM layout + $LFS setstripe -d $DIR/$tdir + + # Now test that append striping works when layout is from root + $LFS setstripe -c 2 $MOUNT + # Make a special directory for this + mkdir $DIR/${tdir}/${tdir}.2 + stack_trap "$LFS setstripe -d $MOUNT" EXIT + + # Verify for normal file + setcount=2 + echo 1 > $DIR/${tdir}/${tdir}.2/${tfile}.8 + count=$($LFS getstripe -c $DIR/$tdir/${tdir}.2/${tfile}.8) + [ $count -eq $setcount ] || + error "(8) stripe count $count, should be $setcount" + + appendcount=1 + echo 1 >> $DIR/${tdir}/${tdir}.2/${tfile}.9_append + count=$($LFS getstripe -c $DIR/${tdir}/${tdir}.2/${tfile}.9_append) + [ $count -eq $appendcount ] || + error "(9) stripe count $count, should be $appendcount for append" + + # Now test O_APPEND striping with pools + do_nodes $mdts $LCTL set_param mdd.*.append_pool="$TESTNAME" + stack_trap "do_nodes $mdts $LCTL set_param mdd.*.append_pool='none'" EXIT + + # Create the pool + pool_add $TESTNAME || error "pool creation failed" + pool_add_targets $TESTNAME 0 1 || error "Pool add targets failed" + + echo 1 >> $DIR/$tdir/${tfile}.10_append + + pool=$($LFS getstripe -p $DIR/$tdir/${tfile}.10_append) + [ "$pool" = "$TESTNAME" ] || error "(10) incorrect pool: $pool" + + # Check that count is still correct + appendcount=1 + echo 1 >> $DIR/$tdir/${tfile}.11_append + count=$($LFS getstripe -c $DIR/$tdir/${tfile}.11_append) + [ $count -eq $appendcount ] || + error "(11) stripe count $count, should be $appendcount for append" + + # Disable O_APPEND stripe count, verify pool works separately + do_nodes $mdts $LCTL set_param mdd.*.append_stripe_count=0 + + echo 1 >> $DIR/$tdir/${tfile}.12_append + + pool=$($LFS getstripe -p $DIR/$tdir/${tfile}.12_append) + [ "$pool" = "$TESTNAME" ] || error "(12) incorrect pool: $pool" + + # Remove pool setting, verify it's not applied + do_nodes $mdts $LCTL set_param mdd.*.append_pool='none' + + echo 1 >> $DIR/$tdir/${tfile}.13_append + + pool=$($LFS getstripe -p $DIR/$tdir/${tfile}.13_append) + [ "$pool" = "" ] || error "(13) pool found: $pool" +} +run_test 27M "test O_APPEND striping" + +test_27N() { + combined_mgs_mds && skip "needs separate MGS/MDT" + + pool_add $TESTNAME || error "pool_add failed" + do_facet mgs "$LCTL pool_list $FSNAME" | + grep -Fx "${FSNAME}.${TESTNAME}" || + error "lctl pool_list on MGS failed" +} +run_test 27N "lctl pool_list on separate MGS gives correct pool name" + # createtest also checks that device nodes are created and # then visible correctly (#2091) test_28() { # bug 2091 @@ -4133,7 +4280,7 @@ test_39j() { error "mtime is lost on close: $mtime2, " \ "should be $mtime1" - cancel_lru_locks $OSC + cancel_lru_locks if [ $i = 0 ] ; then echo "repeat after cancel_lru_locks"; fi done lctl set_param fail_loc=0 @@ -4162,7 +4309,7 @@ test_39k() { [ "$mtime2" = $TEST_39_MTIME ] || \ error "mtime is lost on close: $mtime2, should be $TEST_39_MTIME" - cancel_lru_locks osc + cancel_lru_locks if [ $i = 0 ] ; then echo "repeat after cancel_lru_locks"; fi done } @@ -4840,7 +4987,7 @@ test_49() { # LU-1030 [ $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 @@ -5360,19 +5507,27 @@ test_56c() { skip_env "OST $ost_name is in $old_status status" do_facet ost1 $LCTL set_param -n obdfilter.$ost_name.degraded=1 + [[ $OST1_VERSION -ge $(version_code 2.12.55) ]] && do_facet ost1 \ + $LCTL set_param -n obdfilter.$ost_name.no_precreate=1 sleep_maxage local new_status=$(ost_dev_status $ost_idx) - [[ "$new_status" = "D" ]] || - error "OST $ost_name is in status of '$new_status', not 'D'" + [[ "$new_status" =~ "D" ]] || + error "$ost_name status is '$new_status', missing 'D'" + if [[ $OST1_VERSION -ge $(version_code 2.12.55) ]]; then + [[ "$new_status" =~ "N" ]] || + error "$ost_name status is '$new_status', missing 'N'" + fi do_facet ost1 $LCTL set_param -n obdfilter.$ost_name.degraded=0 + [[ $OST1_VERSION -ge $(version_code 2.12.55) ]] && do_facet ost1 \ + $LCTL set_param -n obdfilter.$ost_name.no_precreate=0 sleep_maxage new_status=$(ost_dev_status $ost_idx) - [[ -z "$new_status" ]] || - error "OST $ost_name is in status of '$new_status', not ''" + [[ ! "$new_status" =~ "D" && ! "$new_status" =~ "N" ]] || + error "$ost_name status is '$new_status', has 'D' and/or 'N'" } run_test 56c "check 'lfs df' showing device status" @@ -5595,6 +5750,65 @@ test_56ob() { } 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" @@ -5649,39 +5863,73 @@ test_56r() { setup_56 $dir $NUMFILES $NUMDIRS local expected=12 - local cmd="$LFS find -size 0 -type f $dir" + local cmd="$LFS find -size 0 -type f -lazy $dir" local nums=$($cmd | wc -l) [ $nums -eq $expected ] || error "'$cmd' wrong: found $nums, expected $expected" + cmd="$LFS find -size 0 -type f $dir" + nums=$($cmd | wc -l) + [ $nums -eq $expected ] || + error "'$cmd' wrong: found $nums, expected $expected" + expected=0 + cmd="$LFS find ! -size 0 -type f -lazy $dir" + nums=$($cmd | wc -l) + [ $nums -eq $expected ] || + error "'$cmd' wrong: found $nums, expected $expected" cmd="$LFS find ! -size 0 -type f $dir" nums=$($cmd | wc -l) [ $nums -eq $expected ] || error "'$cmd' wrong: found $nums, expected $expected" + echo "test" > $dir/$tfile echo "test2" > $dir/$tfile.2 && sync expected=1 + cmd="$LFS find -size 5 -type f -lazy $dir" + nums=$($cmd | wc -l) + [ $nums -eq $expected ] || + error "'$cmd' wrong: found $nums, expected $expected" cmd="$LFS find -size 5 -type f $dir" nums=$($cmd | wc -l) [ $nums -eq $expected ] || error "'$cmd' wrong: found $nums, expected $expected" + expected=1 + cmd="$LFS find -size +5 -type f -lazy $dir" + nums=$($cmd | wc -l) + [ $nums -eq $expected ] || + error "'$cmd' wrong: found $nums, expected $expected" cmd="$LFS find -size +5 -type f $dir" nums=$($cmd | wc -l) [ $nums -eq $expected ] || error "'$cmd' wrong: found $nums, expected $expected" + expected=2 + cmd="$LFS find -size +0 -type f -lazy $dir" + nums=$($cmd | wc -l) + [ $nums -eq $expected ] || + error "'$cmd' wrong: found $nums, expected $expected" cmd="$LFS find -size +0 -type f $dir" nums=$($cmd | wc -l) [ $nums -eq $expected ] || error "'$cmd' wrong: found $nums, expected $expected" + expected=2 + cmd="$LFS find ! -size -5 -type f -lazy $dir" + nums=$($cmd | wc -l) + [ $nums -eq $expected ] || + error "'$cmd' wrong: found $nums, expected $expected" cmd="$LFS find ! -size -5 -type f $dir" nums=$($cmd | wc -l) [ $nums -eq $expected ] || error "'$cmd' wrong: found $nums, expected $expected" + expected=12 + cmd="$LFS find -size -5 -type f -lazy $dir" + nums=$($cmd | wc -l) + [ $nums -eq $expected ] || + error "'$cmd' wrong: found $nums, expected $expected" cmd="$LFS find -size -5 -type f $dir" nums=$($cmd | wc -l) [ $nums -eq $expected ] || @@ -5689,6 +5937,157 @@ test_56r() { } run_test 56r "check lfs find -size works" +test_56ra() { + [[ $MDS1_VERSION -gt $(version_code 2.12.58) ]] || + skip "MDS < 2.12.58 doesn't return LSOM data" + local dir=$DIR/$tdir + + [[ $OSC == "mdc" ]] && skip "DoM files" && return + + setup_56 $dir $NUMFILES $NUMDIRS "-c 1" + + cancel_lru_locks $OSC + + local rpcs_before=$(calc_stats $OSC.*$OSC*.stats ldlm_glimpse_enqueue) + local expected=12 + local cmd="$LFS find -size 0 -type f -lazy $dir" + local nums=$($cmd | wc -l) + + [ $nums -eq $expected ] || + error "'$cmd' wrong: found $nums, expected $expected" + + local rpcs_after=$(calc_stats $OSC.*$OSC*.stats ldlm_glimpse_enqueue) + [ $rpcs_before -eq $rpcs_after ] || + error "'$cmd' should not send glimpse RPCs to OST" + cmd="$LFS find -size 0 -type f $dir" + nums=$($cmd | wc -l) + [ $nums -eq $expected ] || + error "'$cmd' wrong: found $nums, expected $expected" + rpcs_after=$(calc_stats $OSC.*$OSC*.stats ldlm_glimpse_enqueue) + echo "Before: $rpcs_before After: $rpcs_after $NUMFILES" + $LCTL get_param osc.*.stats + [ $rpcs_after -eq $((rpcs_before + 12)) ] || + error "'$cmd' should send 12 glimpse RPCs to OST" + + cancel_lru_locks $OSC + rpcs_before=$(calc_stats $OSC.*$OSC*.stats ldlm_glimpse_enqueue) + expected=0 + cmd="$LFS find ! -size 0 -type f -lazy $dir" + nums=$($cmd | wc -l) + [ $nums -eq $expected ] || + error "'$cmd' wrong: found $nums, expected $expected" + rpcs_after=$(calc_stats $OSC.*$OSC*.stats ldlm_glimpse_enqueue) + $LCTL get_param mdc.*.stats + [ $rpcs_before -eq $rpcs_after ] || + error "'$cmd' should not send glimpse RPCs to OST" + cmd="$LFS find ! -size 0 -type f $dir" + nums=$($cmd | wc -l) + [ $nums -eq $expected ] || + error "'$cmd' wrong: found $nums, expected $expected" + rpcs_after=$(calc_stats $OSC.*$OSC*.stats ldlm_glimpse_enqueue) + echo "Before: $rpcs_before After: $rpcs_after $NUMFILES" + [ $rpcs_after -eq $((rpcs_before + 12)) ] || + error "'$cmd' should send 12 glimpse RPCs to OST" + + echo "test" > $dir/$tfile + echo "test2" > $dir/$tfile.2 && sync + cancel_lru_locks $OSC + rpcs_before=$(calc_stats $OSC.*$OSC*.stats ldlm_glimpse_enqueue) + expected=1 + cmd="$LFS find -size 5 -type f -lazy $dir" + nums=$($cmd | wc -l) + [ $nums -eq $expected ] || + error "'$cmd' wrong: found $nums, expected $expected" + rpcs_after=$(calc_stats $OSC.*$OSC*.stats ldlm_glimpse_enqueue) + [ $rpcs_before -eq $rpcs_after ] || + error "'$cmd' should not send glimpse RPCs to OST" + cmd="$LFS find -size 5 -type f $dir" + nums=$($cmd | wc -l) + [ $nums -eq $expected ] || + error "'$cmd' wrong: found $nums, expected $expected" + rpcs_after=$(calc_stats $OSC.*$OSC*.stats ldlm_glimpse_enqueue) + echo "Before: $rpcs_before After: $rpcs_after $NUMFILES" + [ $rpcs_after -eq $((rpcs_before + 14)) ] || + error "'$cmd' should send 14 glimpse RPCs to OST" + + cancel_lru_locks $OSC + rpcs_before=$(calc_stats $OSC.*$OSC*.stats ldlm_glimpse_enqueue) + expected=1 + cmd="$LFS find -size +5 -type f -lazy $dir" + nums=$($cmd | wc -l) + [ $nums -eq $expected ] || + error "'$cmd' wrong: found $nums, expected $expected" + rpcs_after=$(calc_stats $OSC.*$OSC*.stats ldlm_glimpse_enqueue) + [ $rpcs_before -eq $rpcs_after ] || + error "'$cmd' should not send glimpse RPCs to OST" + cmd="$LFS find -size +5 -type f $dir" + nums=$($cmd | wc -l) + [ $nums -eq $expected ] || + error "'$cmd' wrong: found $nums, expected $expected" + rpcs_after=$(calc_stats $OSC.*$OSC*.stats ldlm_glimpse_enqueue) + echo "Before: $rpcs_before After: $rpcs_after $NUMFILES" + [ $rpcs_after -eq $((rpcs_before + 14)) ] || + error "'$cmd' should send 14 glimpse RPCs to OST" + + cancel_lru_locks $OSC + rpcs_before=$(calc_stats $OSC.*$OSC*.stats ldlm_glimpse_enqueue) + expected=2 + cmd="$LFS find -size +0 -type f -lazy $dir" + nums=$($cmd | wc -l) + [ $nums -eq $expected ] || + error "'$cmd' wrong: found $nums, expected $expected" + rpcs_after=$(calc_stats $OSC.*$OSC*.stats ldlm_glimpse_enqueue) + [ $rpcs_before -eq $rpcs_after ] || + error "'$cmd' should not send glimpse RPCs to OST" + cmd="$LFS find -size +0 -type f $dir" + nums=$($cmd | wc -l) + [ $nums -eq $expected ] || + error "'$cmd' wrong: found $nums, expected $expected" + rpcs_after=$(calc_stats $OSC.*$OSC*.stats ldlm_glimpse_enqueue) + echo "Before: $rpcs_before After: $rpcs_after $NUMFILES" + [ $rpcs_after -eq $((rpcs_before + 14)) ] || + error "'$cmd' should send 14 glimpse RPCs to OST" + + cancel_lru_locks $OSC + rpcs_before=$(calc_stats $OSC.*$OSC*.stats ldlm_glimpse_enqueue) + expected=2 + cmd="$LFS find ! -size -5 -type f -lazy $dir" + nums=$($cmd | wc -l) + [ $nums -eq $expected ] || + error "'$cmd' wrong: found $nums, expected $expected" + rpcs_after=$(calc_stats $OSC.*$OSC*.stats ldlm_glimpse_enqueue) + [ $rpcs_before -eq $rpcs_after ] || + error "'$cmd' should not send glimpse RPCs to OST" + cmd="$LFS find ! -size -5 -type f $dir" + nums=$($cmd | wc -l) + [ $nums -eq $expected ] || + error "'$cmd' wrong: found $nums, expected $expected" + rpcs_after=$(calc_stats $OSC.*$OSC*.stats ldlm_glimpse_enqueue) + echo "Before: $rpcs_before After: $rpcs_after $NUMFILES" + [ $rpcs_after -eq $((rpcs_before + 14)) ] || + error "'$cmd' should send 14 glimpse RPCs to OST" + + cancel_lru_locks $OSC + rpcs_before=$(calc_stats $OSC.*$OSC*.stats ldlm_glimpse_enqueue) + expected=12 + cmd="$LFS find -size -5 -type f -lazy $dir" + nums=$($cmd | wc -l) + [ $nums -eq $expected ] || + error "'$cmd' wrong: found $nums, expected $expected" + rpcs_after=$(calc_stats $OSC.*$OSC*.stats ldlm_glimpse_enqueue) + [ $rpcs_before -eq $rpcs_after ] || + error "'$cmd' should not send glimpse RPCs to OST" + cmd="$LFS find -size -5 -type f $dir" + nums=$($cmd | wc -l) + [ $nums -eq $expected ] || + error "'$cmd' wrong: found $nums, expected $expected" + rpcs_after=$(calc_stats $OSC.*$OSC*.stats ldlm_glimpse_enqueue) + echo "Before: $rpcs_before After: $rpcs_after $NUMFILES" + [ $rpcs_after -eq $((rpcs_before + 14)) ] || + error "'$cmd' should send 14 glimpse RPCs to OST" +} +run_test 56ra "check lfs find -size -lazy works for data on OSTs" + test_56s() { # LU-611 #LU-9369 [[ $OSTCOUNT -lt 2 ]] && skip_env "need at least 2 OSTs" @@ -5911,7 +6310,7 @@ test_56w() { done # $LFS_MIGRATE will fail if hard link migration is unsupported - if [[ $(lustre_version_code mds1) -gt $(version_code 2.5.55) ]]; then + if [[ $MDS1_VERSION -gt $(version_code 2.5.55) ]]; then createmany -l$dir/dir1/file1 $dir/dir1/link 200 || error "creating links to $dir/dir1/file1 failed" fi @@ -6249,6 +6648,7 @@ check_migrate_links() { local file1="$dir/file1" local begin="$2" local count="$3" + local runas="$4" local total_count=$(($begin + $count - 1)) local symlink_count=10 local uniq_count=10 @@ -6293,7 +6693,7 @@ check_migrate_links() { fi echo -n "migrating files..." - local migrate_out=$($LFS_MIGRATE -y -S '1m' $dir) + local migrate_out=$($runas $LFS_MIGRATE -y -S '1m' $dir) local rc=$? [ $rc -eq 0 ] || error "migrate failed rc = $rc" echo "done" @@ -6348,6 +6748,9 @@ test_56xb() { echo "testing rsync mode when all links do not fit within xattrs" LFS_MIGRATE_RSYNC_MODE=true check_migrate_links "$dir" 101 100 + chown -R $RUNAS_ID $dir + echo "testing non-root lfs migrate mode when not all links are in xattr" + LFS_MIGRATE_RSYNC_MODE=false check_migrate_links "$dir" 101 100 "$RUNAS" # clean up rm -rf $dir @@ -7478,8 +7881,13 @@ test_65n() { local dir3=$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 @@ -7501,8 +7909,14 @@ test_65n() { 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 $dir4 should inherit the pool from # the filesystem default @@ -7521,10 +7935,33 @@ test_65n() { local dir5=$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" + + # though subdir under ROOT doesn't inherit default layout, but + # its sub dir/file should be created with default layout. + [[ $MDSCOUNT -ge 2 ]] || skip_env "needs >= 2 MDTs" + [[ $MDS1_VERSION -ge $(version_code 2.12.59) ]] || + skip "Need MDS version at least 2.12.59" + + local default_lmv_count=$($LFS getdirstripe -D -c $MOUNT) + local default_lmv_index=$($LFS getdirstripe -D -i $MOUNT) + local default_lmv_hash=$($LFS getdirstripe -D -H $MOUNT) + + if [ $default_lmv_hash == "none" ]; then + stack_trap "$LFS setdirstripe -D -d $MOUNT" EXIT + else + stack_trap "$LFS setdirstripe -D -i $default_lmv_index \ + -c $default_lmv_count -H $default_lmv_hash $MOUNT" EXIT + fi + + $LFS setdirstripe -D -c 2 $MOUNT || + error "setdirstripe -D -c 2 failed" + mkdir $MOUNT/$tdir-6 || error "mkdir $tdir-6 failed" + local lmv_count=$($LFS getdirstripe -c $MOUNT/$tdir-6) + [ $lmv_count -eq 2 ] || error "$tdir-6 stripe count $lmv_count" } run_test 65n "don't inherit default layout from root for new subdirectories" @@ -8012,7 +8449,7 @@ test_77k() { # LU-10906 wait_update $HOSTNAME "$get_checksum" $i #remount echo "remount client, checksum should be $i" - remount_client $MOUNT || "failed to remount client" + remount_client $MOUNT || error "failed to remount client" checksum=$(eval $get_checksum) [ $checksum -eq $i ] || error "checksum($checksum) != $i" done @@ -8023,15 +8460,15 @@ test_77k() { # LU-10906 for opt in "checksum" "nochecksum"; do #remount with mount option echo "remount client with option $opt, checksum should be $i" - umount_client $MOUNT || "failed to umount client" + umount_client $MOUNT || error "failed to umount client" mount_client $MOUNT "$MOUNT_OPTS,$opt" || - "failed to mount client with option '$opt'" + error "failed to mount client with option '$opt'" checksum=$(eval $get_checksum) [ $checksum -eq $i ] || error "checksum($checksum) != $i" i=$((i - 1)) done - remount_client $MOUNT || "failed to remount client" + remount_client $MOUNT || error "failed to remount client" } run_test 77k "enable/disable checksum correctly" @@ -8160,31 +8597,35 @@ test_80() { # bug 10718 [ $PARALLEL == "yes" ] && skip "skip parallel run" # relax strong synchronous semantics for slow backends like ZFS - local soc="obdfilter.*.sync_on_lock_cancel" - local soc_old=$(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" @@ -8328,7 +8769,6 @@ cleanup_101a() { test_101a() { [ $PARALLEL == "yes" ] && skip "skip parallel run" - [ $MDSCOUNT -ge 2 ] && skip_env "needs < 2 MDTs" #LU-4322 local s local discard @@ -8353,9 +8793,12 @@ test_101a() { done cleanup_101a - if [[ $(($discard * 10)) -gt $nreads ]]; then - $LCTL get_param osc.*-osc*.rpc_stats - $LCTL get_param llite.*.read_ahead_stats + $LCTL get_param osc.*-osc*.rpc_stats + $LCTL get_param llite.*.read_ahead_stats + + # Discard is generally zero, but sometimes a few random reads line up + # and trigger larger readahead, which is wasted & leads to discards. + if [[ $(($discard)) -gt $nreads ]]; then error "too many ($discard) discarded pages" fi rm -f $DIR/$tfile || true @@ -10998,41 +11441,52 @@ run_test 126 "check that the fsgid provided by the client is taken into account" 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" @@ -11040,6 +11494,8 @@ test_127b() { # bug LU-333 [ $PARALLEL == "yes" ] && skip "skip parallel run" local name count samp unit min max sum sumsq + echo "stats before reset" + $LCTL get_param llite.*.stats $LCTL set_param llite.*.stats=0 # perform 2 reads and writes so MAX is different from SUM. @@ -11050,30 +11506,28 @@ test_127b() { # bug LU-333 dd if=$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 @@ -11082,8 +11536,6 @@ test_127b() { # bug LU-333 [ "$write_bytes" ] || error "Missing write_bytes stats" [ "$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" @@ -12162,6 +12614,73 @@ test_134b() { } 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" + $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" @@ -12335,8 +12854,9 @@ test_151() { 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 @@ -12981,9 +13501,9 @@ test_156() { 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." @@ -12992,7 +13512,7 @@ test_156() { 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 @@ -13007,20 +13527,23 @@ test_156() { 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." @@ -13031,20 +13554,23 @@ test_156() { 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" @@ -13060,7 +13586,7 @@ test_156() { 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 @@ -13078,7 +13604,7 @@ test_156() { 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 @@ -13089,7 +13615,7 @@ test_156() { 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 @@ -13339,14 +13865,17 @@ test_160f() { # generate some changelog records to accumulate on each MDT test_mkdir -c $MDSCOUNT $DIR/$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 @@ -13358,8 +13887,11 @@ test_160f() { 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 $sleep1 # simulate changelog catalog almost full #define OBD_FAIL_CAT_FREE_RECORDS 0x1313 @@ -13395,13 +13927,16 @@ test_160f() { "$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 @@ -13771,6 +14306,77 @@ test_160i() { } run_test 160i "changelog user register/unregister race" +test_160j() { + remote_mds_nodsh && skip "remote MDS with nodsh" + [[ $MDS1_VERSION -lt $(version_code 2.12.56) ]] && + skip "Need MDS version at least 2.12.56" + + mount_client $MOUNT2 || error "mount_client on $MOUNT2 failed" + + changelog_register || error "first changelog_register failed" + + # generate some changelog + test_mkdir -c $MDSCOUNT $DIR/$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 + exec 4/dev/null || error "read changelog failed" + + # clear changelog + local cl_user="${CL_USERS[$SINGLEMDS]%% *}" + changelog_users $SINGLEMDS | grep -q $cl_user || + 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" + +test_160k() { + [ $PARALLEL == "yes" ] && skip "skip parallel run" + remote_mds_nodsh && skip "remote MDS with nodsh" + + mkdir -p $DIR/$tdir/1/1 + + changelog_register || error "changelog_register failed" + local cl_user="${CL_USERS[$SINGLEMDS]%% *}" + + changelog_users $SINGLEMDS | grep -q $cl_user || + error "User '$cl_user' not found in changelog_users" +#define OBD_FAIL_MDS_CHANGELOG_REORDER 0x15d + do_facet mds1 $LCTL set_param fail_loc=0x8000015d fail_val=3 + rmdir $DIR/$tdir/1/1 & sleep 1 + mkdir $DIR/$tdir/2 + touch $DIR/$tdir/2/2 + rm -rf $DIR/$tdir/2 + + wait + sleep 4 + + changelog_dump | grep rmdir || error "rmdir not recorded" + + rm -rf $DIR/$tdir + changelog_deregister +} +run_test 160k "Verify that changelog records are not lost" + test_161a() { [ $PARALLEL == "yes" ] && skip "skip parallel run" @@ -14723,10 +15329,6 @@ test_200() { local subdir=$test_path/subdir local rc=0 - if ! combined_mgs_mds ; then - mount_mgs_client - fi - while : ; do # former test_200a test_200b pool_add $POOL || { rc=$? ; break; } @@ -14760,9 +15362,6 @@ test_200() { destroy_test_pools - if ! combined_mgs_mds ; then - umount_mgs_client - fi return $rc } run_test 200 "OST pools" @@ -15444,10 +16043,6 @@ test_220() { #LU-325 $LFS df -i - if ! combined_mgs_mds ; then - mount_mgs_client - fi - do_facet ost$((OSTIDX + 1)) lctl set_param fail_val=-1 #define OBD_FAIL_OST_ENOINO 0x229 do_facet ost$((OSTIDX + 1)) lctl set_param fail_loc=0x229 @@ -15484,10 +16079,6 @@ test_220() { #LU-325 error "$LCTL pool_destroy $FSNAME.$TESTNAME failed" echo "unlink $MDSOBJS files @$next_id..." unlinkmany $DIR/$tdir/f $MDSOBJS || error "unlinkmany failed" - - if ! combined_mgs_mds ; then - umount_mgs_client - fi } run_test 220 "preallocated MDS objects still used if ENOSPC from OST" @@ -16491,6 +17082,61 @@ test_230l() { } 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_231a() { # For simplicity this test assumes that max_pages_per_rpc @@ -16895,21 +17541,6 @@ test_245() { } 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=$1 @@ -17002,7 +17633,7 @@ test_247e() { } run_test 247e "mount .. as fileset" -test_248() { +test_248a() { local fast_read_sav=$($LCTL get_param -n llite.*.fast_read 2>/dev/null) [ -z "$fast_read_sav" ] && skip "no fast read support" @@ -17053,7 +17684,89 @@ test_248() { $LCTL set_param -n llite.*.fast_read=$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) ] && @@ -17168,9 +17881,6 @@ test_253() { wait_delete_completed mkdir $DIR/$tdir - if ! combined_mgs_mds ; then - mount_mgs_client - fi pool_add $TESTNAME || error "Pool creation failed" pool_add_targets $TESTNAME 0 || error "Pool add targets failed" @@ -17216,10 +17926,6 @@ test_253() { if (( oa_status != 0 )); then error "Object allocation still disable after rm" fi - - if ! combined_mgs_mds ; then - umount_mgs_client - fi } run_test 253 "Check object allocation limit" @@ -17525,7 +18231,7 @@ run_test 255b "check 'lfs ladvise -a dontneed'" test_255c() { [ $OST1_VERSION -lt $(version_code 2.10.50) ] && - skip "lustre < 2.10.53 does not support lockahead" + skip "lustre < 2.10.50 does not support lockahead" local count local new_count @@ -18358,7 +19064,7 @@ test_272a() { 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" @@ -18385,12 +19091,12 @@ test_272b() { $LFS migrate -c2 $dom || error "failed to migrate to the new composite layout" - [ $($LFS getstripe -L $dom) == 'mdt' ] && + [ $($LFS getstripe -L $dom) != 'mdt' ] || error "MDT stripe was not removed" cancel_lru_locks mdc local new_md5=$(md5sum $dom) - [ "$old_md5" != "$new_md5" ] && + [ "$old_md5" == "$new_md5" ] || error "$old_md5 != $new_md5" # Skip free space checks with ZFS @@ -18430,7 +19136,7 @@ test_272c() { cancel_lru_locks mdc local new_md5=$(md5sum $dom) - [ "$old_md5" != "$new_md5" ] && + [ "$old_md5" == "$new_md5" ] || error "$old_md5 != $new_md5" # Skip free space checks with ZFS @@ -18444,6 +19150,108 @@ test_272c() { } run_test 272c "DoM migration: DOM file to the OST-striped file (composite)" +test_272d() { + [ $MDS1_VERSION -lt $(version_code 2.12.55) ] && + skip "Need MDS version at least 2.12.55" + + local dom=$DIR/$tdir/$tfile + mkdir -p $DIR/$tdir + $LFS setstripe -E 1M -L mdt -E -1 -c1 $dom + + local mdtidx=$($LFS getstripe -m $dom) + local mdtname=MDT$(printf %04x $mdtidx) + local facet=mds$((mdtidx + 1)) + + dd if=/dev/urandom of=$dom bs=2M count=1 oflag=direct || + error "failed to write data into $dom" + local old_md5=$(md5sum $dom) + cancel_lru_locks mdc + local mdtfree1=$(do_facet $facet \ + lctl get_param -n osd*.*$mdtname.kbytesfree) + + $LFS mirror extend -N -E 2M -c1 -E -1 -c2 $dom || + error "failed mirroring to the new composite layout" + $LFS mirror resync $dom || + error "failed mirror resync" + $LFS mirror split --mirror-id 1 -d $dom || + error "failed mirror split" + + [ $($LFS getstripe -L $dom) != 'mdt' ] || + error "MDT stripe was not removed" + + cancel_lru_locks mdc + local new_md5=$(md5sum $dom) + [ "$old_md5" == "$new_md5" ] || + error "$old_md5 != $new_md5" + + # Skip free space checks with ZFS + if [ "$(facet_fstype $facet)" != "zfs" ]; then + local mdtfree2=$(do_facet $facet \ + lctl get_param -n osd*.*$mdtname.kbytesfree) + [ $mdtfree2 -gt $mdtfree1 ] || + error "MDS space is not freed after DOM mirror deletion" + fi + return 0 +} +run_test 272d "DoM mirroring: OST-striped mirror to DOM file" + +test_272e() { + [ $MDS1_VERSION -lt $(version_code 2.12.55) ] && + skip "Need MDS version at least 2.12.55" + + local dom=$DIR/$tdir/$tfile + mkdir -p $DIR/$tdir + $LFS setstripe -c 2 $dom + + dd if=/dev/urandom of=$dom bs=512K count=1 oflag=direct || + error "failed to write data into $dom" + local old_md5=$(md5sum $dom) + cancel_lru_locks mdc + + $LFS mirror extend -N -E 1M -L mdt -E eof -c2 $dom || + error "failed mirroring to the DOM layout" + $LFS mirror resync $dom || + error "failed mirror resync" + $LFS mirror split --mirror-id 1 -d $dom || + error "failed mirror split" + + [ $($LFS getstripe -L $dom) != 'mdt' ] || + error "MDT stripe was not removed" + + cancel_lru_locks mdc + local new_md5=$(md5sum $dom) + [ "$old_md5" == "$new_md5" ] || + error "$old_md5 != $new_md5" + + return 0 +} +run_test 272e "DoM mirroring: DOM mirror to the OST-striped file" + +test_272f() { + [ $MDS1_VERSION -lt $(version_code 2.12.55) ] && + skip "Need MDS version at least 2.12.55" + + local dom=$DIR/$tdir/$tfile + mkdir -p $DIR/$tdir + $LFS setstripe -c 2 $dom + + dd if=/dev/urandom of=$dom bs=512K count=1 oflag=direct || + error "failed to write data into $dom" + local old_md5=$(md5sum $dom) + cancel_lru_locks mdc + + $LFS migrate -E 1M -L mdt -E eof -c2 -v $dom || + error "failed migrating to the DOM file" + + cancel_lru_locks mdc + local new_md5=$(md5sum $dom) + [ "$old_md5" != "$new_md5" ] && + error "$old_md5 != $new_md5" + + return 0 +} +run_test 272f "DoM migration: OST-striped file to DOM file" + test_273a() { [ $MDS1_VERSION -lt $(version_code 2.11.50) ] && skip "Need MDS version at least 2.11.50" @@ -18560,6 +19368,7 @@ test_278() { do_facet mds2 $LCTL set_param fail_loc=0 start mds2 $(mdsdevname 2) $MDS_MOUNT_OPTS + wait_recovery_complete mds2 } run_test 278 "Race starting MDS between MDTs stop/start" @@ -19045,8 +19854,9 @@ test_300k() { # this test needs a huge transaction local kb - kb=$(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 @@ -19728,7 +20538,7 @@ test_fake_rw() { $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 [ $blocks -gt 1000 ] && blocks=1000 # 1G in maximum @@ -20039,9 +20849,6 @@ test_406() { local def_stripe_size=$($LFS getstripe -S $MOUNT) local test_pool=$TESTNAME - if ! combined_mgs_mds ; then - mount_mgs_client - fi pool_add $test_pool || error "pool_add failed" pool_add_targets $test_pool 0 $(($OSTCOUNT - 1)) 1 || error "pool_add_targets failed" @@ -20094,10 +20901,6 @@ test_406() { local f=$DIR/$tdir/$tfile pool_remove_all_targets $test_pool $f pool_remove $test_pool $f - - if ! combined_mgs_mds ; then - umount_mgs_client - fi } run_test 406 "DNE support fs default striping" @@ -20248,174 +21051,247 @@ test_412() { } 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 "mkdir with roundrobin" + echo + echo "Mkdir (stripe_count $stripe_count) 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 - $LCTL set_param lmv.*.qos_maxage=1 - sleep_maxage + echo + echo "Check for uneven MDTs: " local ffree + local bavail local max local min local max_index local min_index + local tmp - ffree=($(lctl get_param -n mdc.*[mM][dD][cC]-[^M]*.filesfree | uniq)) - echo "MDT filesfree available: ${ffree[@]}" - max=${ffree[0]} - min=${ffree[0]} + ffree=($(lctl get_param -n mdc.*[mM][dD][cC]-[^M]*.filesfree)) + bavail=($(lctl get_param -n mdc.*[mM][dD][cC]-[^M]*.kbytesavail)) + bsize=$(lctl get_param -n mdc.*MDT0000*.blocksize) + + max=$(((${ffree[0]} >> 8) * (${bavail[0]} * bsize >> 16))) + min=$(((${ffree[0]} >> 8) * (${bavail[0]} * bsize >> 16))) max_index=0 min_index=0 - for ((i = 0; i < ${#ffree[@]}; i++)); do - if [[ ${ffree[i]} -gt $max ]]; then - max=${ffree[i]} + for ((i = 1; i < ${#ffree[@]}; i++)); do + tmp=$(((${ffree[i]} >> 8) * (${bavail[i]} * bsize >> 16))) + if [ $tmp -gt $max ]; then + max=$tmp max_index=$i fi - if [[ ${ffree[i]} -lt $min ]]; then - min=${ffree[i]} + if [ $tmp -lt $min ]; then + min=$tmp min_index=$i fi done - echo "Min free files: MDT$min_index: $min" - echo "Max free files: MDT$max_index: $max" - [ $min -eq 0 ] && skip "no free files in MDT$min_index" - [ $min -gt 10000000 ] && skip "too much free files in MDT$min_index" + [ ${ffree[min_index]} -eq 0 ] && + skip "no free files in MDT$min_index" + [ ${ffree[min_index]} -gt 100000000 ] && + skip "too much free files in MDT$min_index" # Check if we need to generate uneven MDTs - test_mkdir -i $min_index -c 1 -p $DIR/$tdir-MDT$min_index - local threshold=10 - local diff=$((max - min)) - local diff2=$((diff * 100 / min)) - - echo -n "Check for uneven MDTs: " - echo -n "diff=$diff files ($diff2%) must be > $threshold% ..." - - if [ $diff2 -gt $threshold ]; then - echo "ok" - echo "Don't need to fill MDT$min_index" - else - # generate uneven MDTs, create till 25% diff - echo "no" - diff2=$((threshold - diff2)) - diff=$((min * diff2 / 100)) + local threshold=50 + local diff=$(((max - min) * 100 / min)) + local value="$(generate_string 1024)" + + while [ $diff -lt $threshold ]; do + # generate uneven MDTs, create till $threshold% diff + echo -n "weight diff=$diff% must be > $threshold% ..." + count=$((${ffree[min_index]} / 10)) # 50 sec per 10000 files in vm - [ $diff -gt 40000 ] && [ "$SLOW" = "no" ] && - skip "$diff files to create" - echo "Fill $diff2% diff in MDT$min_index with $diff files" - local i - local value="$(generate_string 1024)" - for i in $(seq $diff); do + [ $count -gt 40000 ] && [ "$SLOW" = "no" ] && + skip "$count files to create" + echo "Fill MDT$min_index with $count files" + [ -d $DIR/$tdir-MDT$min_index ] || + $LFS mkdir -i $min_index $DIR/$tdir-MDT$min_index || + 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_$i failed" done - fi - min=$((100 *MDSCOUNT)) - max=0 + ffree=($(lctl get_param -n mdc.*[mM][dD][cC]-*.filesfree)) + bavail=($(lctl get_param -n mdc.*[mM][dD][cC]-*.kbytesavail)) + max=$(((${ffree[max_index]} >> 8) * \ + (${bavail[max_index]} * bsize >> 16))) + min=$(((${ffree[min_index]} >> 8) * \ + (${bavail[min_index]} * bsize >> 16))) + diff=$(((max - min) * 100 / min)) + done + + echo "MDT filesfree available: ${ffree[@]}" + echo "MDT blocks available: ${bavail[@]}" + echo "weight diff=$diff%" + + 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 - echo "mkdir with balanced space usage" - $LCTL set_param lmv.*.qos_prio_free=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))" - [ $min -gt $count ] && min=$count - [ $max -lt $count ] && max=$count + + 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 - min)) -gt $MDSCOUNT ] || + + max=$($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 +} - $LFS setdirstripe -D -d $DIR/$tdir || error "setdirstripe -d failed" - getfattr -n trusted.dmv $DIR/$tdir && error "default dir layout exists" - true +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 413b "mkdir with balanced space usage" +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" + + 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" $stripe_count + done +} +run_test 413b "QoS mkdir under dir whose default LMV starting MDT offset is -1" test_414() { #define OBD_FAIL_PTLRPC_BULK_ATTACH 0x521 @@ -20824,9 +21700,9 @@ test_421f() { cnt=$(ls -1 $DIR/$tdir | wc -l) [ $cnt == 1 ] || error "unexpected #files after (5): $cnt" - umount_client $MOUNT || "failed to umount client" + umount_client $MOUNT || error "failed to umount client" mount_client $MOUNT "$MOUNT_OPTS,user_fid2path" || - "failed to mount client'" + error "failed to mount client'" $RUNAS $LFS rmfid $DIR $FID || error "rmfid failed" # rmfid should succeed @@ -20840,9 +21716,9 @@ test_421f() { FID=$(lfs path2fid $DIR/$tdir/f) $RUNAS $LFS rmfid $DIR $FID && error "rmfid didn't fail" - umount_client $MOUNT || "failed to umount client" + umount_client $MOUNT || error "failed to umount client" mount_client $MOUNT "$MOUNT_OPTS" || - "failed to mount client'" + error "failed to mount client'" } run_test 421f "rmfid checks permissions" @@ -20879,6 +21755,50 @@ test_421g() { } run_test 421g "rmfid to return errors properly" +test_422() { + test_mkdir -i 0 -c 1 -p $DIR/$tdir/d1 + test_mkdir -i 0 -c 1 -p $DIR/$tdir/d2 + test_mkdir -i 0 -c 1 -p $DIR/$tdir/d3 + dd if=/dev/zero of=$DIR/$tdir/d1/file1 bs=1k count=1 + dd if=/dev/zero of=$DIR/$tdir/d2/file1 bs=1k count=1 + + local amc=$(at_max_get client) + local amo=$(at_max_get mds1) + local timeout=`lctl get_param -n timeout` + + at_max_set 0 client + at_max_set 0 mds1 + +#define OBD_FAIL_PTLRPC_PAUSE_REQ 0x50a + do_facet mds1 $LCTL set_param fail_loc=0x8000050a \ + fail_val=$(((2*timeout + 10)*1000)) + touch $DIR/$tdir/d3/file & + sleep 2 +#define OBD_FAIL_TGT_REPLY_DATA_RACE 0x722 + do_facet mds1 $LCTL set_param fail_loc=0x80000722 \ + fail_val=$((2*timeout + 5)) + mv $DIR/$tdir/d1/file1 $DIR/$tdir/d1/file2 & + local pid=$! + sleep 1 + kill -9 $pid + sleep $((2 * timeout)) + echo kill $pid + kill -9 $pid + lctl mark touch + touch $DIR/$tdir/d2/file3 + touch $DIR/$tdir/d2/file4 + touch $DIR/$tdir/d2/file5 + + wait + at_max_set $amc client + at_max_set $amo mds1 + + # LU-12838 - verify the ptlrpc thread watchdog is not always throttled + do_facet mds1 "dmesg | grep 'Dumping the stack trace for debugging'" || + error "Watchdog is always throttled" +} +run_test 422 "kill a process with RPC in progress" + prep_801() { [[ $(lustre_version_code mds1) -lt $(version_code 2.9.55) ]] || [[ $OST1_VERSION -lt $(version_code 2.9.55) ]] && @@ -21110,7 +22030,7 @@ cleanup_802a() { } test_802a() { - + [[ $mds1_FSTYPE = zfs ]] || skip "ZFS specific test" [[ $(lustre_version_code mds1) -lt $(version_code 2.9.55) ]] || [[ $OST1_VERSION -lt $(version_code 2.9.55) ]] && skip "Need server version at least 2.9.55" @@ -21326,7 +22246,7 @@ cleanup_805() { } test_805() { - local zfs_version=$(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 $(version_code 0.7.2) ] && skip "netfree not implemented before 0.7" @@ -21338,7 +22258,7 @@ test_805() { local usedkb local old local quota - local pref="osd-zfs.lustre-MDT0000." + local pref="osd-zfs.$FSNAME-MDT0000." # limit available space on MDS dataset to meet nospace issue # quickly. then ZFS 0.7.2 can use reserved space if asked @@ -21436,7 +22356,7 @@ test_806() { done check_lsom_size $DIR/$tfile $size - # multi-client wirtes + # multi-client writes num=$(get_node_count ${CLIENTS//,/ }) size=$(($num * $bs)) offset=0 @@ -21601,20 +22521,31 @@ test_809() { run_test 809 "Verify no SOM xattr store for DoM-only files" test_810() { - local ORIG - local CSUM - - # t10 seem to dislike partial pages - lctl set_param osc.*.checksum_type=adler - lctl set_param fail_loc=0x411 - dd if=/dev/urandom of=$DIR/$tfile bs=10240 count=2 - ORIG=$(md5sum $DIR/$tfile) - lctl set_param ldlm.namespaces.*osc*.lru_size=clear - CSUM=$(md5sum $DIR/$tfile) - set_checksum_type adler - if [ "$ORIG" != "$CSUM" ]; then - error "$ORIG != $CSUM" - fi + [ $PARALLEL == "yes" ] && skip "skip parallel run" + $GSS && skip_env "could not run with gss" + [[ $OST1_VERSION -gt $(version_code 2.12.58) ]] || + skip "OST < 2.12.58 doesn't align checksum" + + set_checksums 1 + stack_trap "set_checksums $ORIG_CSUM" EXIT + stack_trap "set_checksum_type $ORIG_CSUM_TYPE" EXIT + + local csum + local before + local after + for csum in $CKSUM_TYPES; do + #define OBD_FAIL_OSC_NO_GRANT 0x411 + $LCTL set_param osc.*.checksum_type=$csum fail_loc=0x411 + for i in "10240 0" "10000 0" "4000 1" "500 1"; do + eval set -- $i + dd if=/dev/urandom of=$DIR/$tfile bs=$1 count=2 seek=$2 + before=$(md5sum $DIR/$tfile) + $LCTL set_param ldlm.namespaces.*osc*.lru_size=clear + after=$(md5sum $DIR/$tfile) + [ "$before" == "$after" ] || + error "$csum: $before != $after bs=$1 seek=$2" + done + done } run_test 810 "partial page writes on ZFS (LU-11663)" @@ -21635,7 +22566,7 @@ test_811() { } run_test 811 "orphan name stub can be cleaned up in startup" -test_812() { +test_812a() { [ $OST1_VERSION -lt $(version_code 2.12.51) ] && skip "OST < 2.12.51 doesn't support this fail_loc" [ "$SHARED_KEY" = true ] && @@ -21656,7 +22587,31 @@ test_812() { stat $DIR/$tfile >/dev/null || error "can't stat file" } -run_test 812 "do not drop reqs generated when imp is going to idle (LU-11951)" +run_test 812a "do not drop reqs generated when imp is going to idle (LU-11951)" + +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 || error "setstripe failed" + # 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 + + # 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" + + $LFS quota -u 0 $DIR/ || error "lfs quota should succeed" + wait_osc_import_state client ost1 IDLE +} +run_test 812b "do not drop no resend request for idle connect" test_813() { local file_heat_sav=$($LCTL get_param -n llite.*.file_heat 2>/dev/null) @@ -21906,6 +22861,26 @@ test_818() { } run_test 818 "unlink with failed llog" +test_819a() { + dd if=/dev/zero of=$DIR/$tfile bs=1M count=1 + cancel_lru_locks osc + #define OBD_FAIL_OST_2BIG_NIOBUF 0x248 + do_facet $SINGLEMDS lctl set_param fail_loc=0x80000248 + dd if=$DIR/$tfile of=/dev/null bs=1M count=1 + rm -f $TDIR/$tfile +} +run_test 819a "too big niobuf in read" + +test_819b() { + #define OBD_FAIL_OST_2BIG_NIOBUF 0x248 + do_facet $SINGLEMDS lctl set_param fail_loc=0x80000248 + dd if=/dev/zero of=$DIR/$tfile bs=1M count=1 + cancel_lru_locks osc + sleep 1 + rm -f $TDIR/$tfile +} +run_test 819b "too big niobuf in write" + # # tests that do cleanup/setup should be run at the end # @@ -21924,6 +22899,31 @@ test_900() { } 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" + complete $SECONDS [ -f $EXT2_DEV ] && rm $EXT2_DEV || true check_and_cleanup_lustre