X-Git-Url: https://git.whamcloud.com/?p=fs%2Flustre-release.git;a=blobdiff_plain;f=lustre%2Ftests%2Fsanity.sh;h=2afc8effe6c986a8d47a40e763cb1b4a30a6e6a8;hp=2da2d775f64b623b32d1a49696b666e88587bd0a;hb=3e8fa8a7396cd029cb0d7714a324343eed7f535e;hpb=23cbeb6e31a77ea2face6fdd947aac5ad3c4e099 diff --git a/lustre/tests/sanity.sh b/lustre/tests/sanity.sh index 2da2d77..2afc8ef 100644 --- a/lustre/tests/sanity.sh +++ b/lustre/tests/sanity.sh @@ -45,24 +45,36 @@ ALWAYS_EXCEPT+=" 42a 42b 42c " ALWAYS_EXCEPT+=" 407 312 " if $SHARED_KEY; then - # bug number: LU-9795 LU-9795 LU-9795 LU-9795 - ALWAYS_EXCEPT+=" 17n 60a 133g 300f " + # bug number: LU-9795 LU-9795 LU-9795 LU-9795 LU-12781 + ALWAYS_EXCEPT+=" 17n 60a 133g 300f 272a" +fi + +selinux_status=$(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-11594 LU-11667 LU-11729 - ALWAYS_EXCEPT+=" 45 103a 317 810" + # 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" 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}.${TESTNAME}" || + error "lctl pool_list on MGS failed" +} +run_test 27N "lctl pool_list on separate MGS gives correct pool name" + # createtest also checks that device nodes are created and # then visible correctly (#2091) test_28() { # bug 2091 @@ -3082,10 +3229,10 @@ test_31n() { run_test 31n "check link count of unlinked file" link_one() { - local TEMPNAME=$(mktemp $1_XXXXXX) - mlink $TEMPNAME $1 2> /dev/null && - echo "$BASHPID: link $TEMPNAME to $1 succeeded" - munlink $TEMPNAME + local tempfile=$(mktemp $1_XXXXXX) + mlink $tempfile $1 2> /dev/null && + echo "$BASHPID: link $tempfile to $1 succeeded" + munlink $tempfile } test_31o() { # LU-2901 @@ -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 = $rc" echo "done" @@ -6312,7 +6712,7 @@ check_migrate_links() { # make sure hard links were properly detected, and migration was # performed only once for the entire link set; nonlinked files should # also be migrated - local actual=$(grep -c 'done migrate' <<< "$migrate_out") + local actual=$(grep -c 'done' <<< "$migrate_out") local expected=$(($uniq_count + 1)) [ "$actual" -eq "$expected" ] || @@ -6348,6 +6748,9 @@ test_56xb() { echo "testing rsync mode when all links do not fit within xattrs" LFS_MIGRATE_RSYNC_MODE=true check_migrate_links "$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 @@ -6363,13 +6766,14 @@ test_56xc() { # Test 1: ensure file < 1 GB is always migrated with 1 stripe echo -n "Setting initial stripe for 20MB test file..." - $LFS setstripe -c 2 -i 0 "$dir/20mb" || error "cannot setstripe" + $LFS setstripe -c 2 -i 0 "$dir/20mb" || + error "cannot setstripe 20MB file" echo "done" echo -n "Sizing 20MB test file..." truncate "$dir/20mb" 20971520 || error "cannot create 20MB test file" echo "done" echo -n "Verifying small file autostripe count is 1..." - $LFS_MIGRATE -y -A -C 1 "$dir/20mb" &> /dev/null || + $LFS_MIGRATE -y -A -C 1 "$dir/20mb" || error "cannot migrate 20MB file" local stripe_count=$($LFS getstripe -c "$dir/20mb") || error "cannot get stripe for $dir/20mb" @@ -6382,24 +6786,28 @@ test_56xc() { # sqrt(size_in_gb) + 1 OSTs but is larger than 1GB. The file must # have at least an additional 1KB for each desired stripe for test 3 echo -n "Setting stripe for 1GB test file..." - $LFS setstripe -c 1 -i 0 "$dir/1gb" || error "cannot setstripe" + $LFS setstripe -c 1 -i 0 "$dir/1gb" || error "cannot setstripe 1GB file" echo "done" echo -n "Sizing 1GB test file..." # File size is 1GB + 3KB - truncate "$dir/1gb" 1073744896 &> /dev/null || - error "cannot create 1GB test file" - echo "done" - echo -n "Migrating 1GB file..." - $LFS_MIGRATE -y -A -C 1 "$dir/1gb" &> /dev/null || - error "cannot migrate file" - echo "done" - echo -n "Verifying autostripe count is sqrt(n) + 1..." - stripe_count=$($LFS getstripe -c "$dir/1gb") || - error "cannot get stripe for $dir/1gb" - [ $stripe_count -eq 2 ] || - error "unexpected stripe count $stripe_count (expected 2)" + truncate "$dir/1gb" 1073744896 || error "cannot create 1GB test file" echo "done" + # need at least 512MB per OST for 1GB file to fit in 2 stripes + local avail=$($LCTL get_param -n llite.$FSNAME*.kbytesavail) + if (( avail > 524288 * OSTCOUNT )); then + echo -n "Migrating 1GB file..." + $LFS_MIGRATE -y -A -C 1 "$dir/1gb" || + error "cannot migrate 1GB file" + echo "done" + echo -n "Verifying autostripe count is sqrt(n) + 1..." + stripe_count=$($LFS getstripe -c "$dir/1gb") || + error "cannot getstripe for 1GB file" + [ $stripe_count -eq 2 ] || + error "unexpected stripe count $stripe_count != 2" + echo "done" + fi + # Test 3: File is too large to fit within the available space on # sqrt(n) + 1 OSTs. Simulate limited available space with -X if [ $OSTCOUNT -ge 3 ]; then @@ -6407,15 +6815,15 @@ test_56xc() { # file size (1GB + 3KB) / OST count (3). local kb_per_ost=349526 - echo -n "Migrating 1GB file..." - $LFS_MIGRATE -y -A -C 1 -X $kb_per_ost "$dir/1gb" &>> \ - /dev/null || error "cannot migrate file" + echo -n "Migrating 1GB file with limit..." + $LFS_MIGRATE -y -A -C 1 -X $kb_per_ost "$dir/1gb" || + error "cannot migrate 1GB file with limit" echo "done" stripe_count=$($LFS getstripe -c "$dir/1gb") - echo -n "Verifying autostripe count with limited space..." - [ "$stripe_count" -a $stripe_count -eq 3 ] || - error "unexpected stripe count $stripe_count (wanted 3)" + echo -n "Verifying 1GB autostripe count with limited space..." + [ "$stripe_count" -a $stripe_count -ge 3 ] || + error "unexpected stripe count $stripe_count (min 3)" echo "done" fi @@ -6466,18 +6874,20 @@ test_56z() { # LU-4824 test_mkdir $dir for i in d{0..9}; do test_mkdir $dir/$i + touch $dir/$i/$tfile done - touch $dir/d{0..9}/$tfile $LFS find $DIR/non_existent_dir $dir && error "$LFS find did not return an error" # Make a directory unsearchable. This should NOT be the last entry in # directory order. Arbitrarily pick the 6th entry chmod 700 $($LFS find $dir -type d | sed '6!d') + $RUNAS $LFS find $DIR/non_existent $dir local count=$($RUNAS $LFS find $DIR/non_existent $dir | wc -l) # The user should be able to see 10 directories and 9 files - [ $count == 19 ] || error "$LFS find did not continue after error" + (( count == 19 )) || + error "$LFS find found $count != 19 entries after error" } run_test 56z "lfs find should continue after an error" @@ -6800,9 +7210,9 @@ test_60a() { local pass=true #get fid and record list - fid_list=($(awk '/9_sub.*record/ { print $NF }' /$TMP/$tfile | + fid_list=($(awk '/9_sub.*record/ { print $NF }' $TMP/$tfile | tail -n 4)) - rec_list=($(awk '/9_sub.*record/ { print $((NF-3)) }' /$TMP/$tfile | + rec_list=($(awk '/9_sub.*record/ { print $((NF-3)) }' $TMP/$tfile | tail -n 4)) #remount mgs as ldiskfs or zfs type stop mgs || error "stop mgs failed" @@ -6921,13 +7331,16 @@ run_test 60e "no space while new llog is being created" test_60g() { local pid + local i test_mkdir -c $MDSCOUNT $DIR/$tdir - $LFS setdirstripe -D -i -1 -c $MDSCOUNT $DIR/$tdir ( local index=0 while true; do + $LFS setdirstripe -i $(($index % $MDSCOUNT)) \ + -c $MDSCOUNT $DIR/$tdir/subdir$index \ + 2>/dev/null mkdir $DIR/$tdir/subdir$index 2>/dev/null rmdir $DIR/$tdir/subdir$index 2>/dev/null index=$((index + 1)) @@ -6936,16 +7349,34 @@ test_60g() { pid=$! - for i in $(seq 100); do + for i in {0..100}; do # define OBD_FAIL_OSD_TXN_START 0x19a - do_facet mds1 lctl set_param fail_loc=0x8000019a + local index=$((i % MDSCOUNT + 1)) + + do_facet mds$index $LCTL set_param fail_loc=0x8000019a \ + > /dev/null usleep 100 done kill -9 $pid + for i in $(seq $MDSCOUNT); do + do_facet mds$i $LCTL set_param fail_loc=0 > /dev/null + done + mkdir $DIR/$tdir/new || error "mkdir failed" rmdir $DIR/$tdir/new || error "rmdir failed" + + do_facet mds1 $LCTL lfsck_start -M $(facet_svc mds1) -A -C \ + -t namespace + for i in $(seq $MDSCOUNT); do + wait_update_facet mds$i "$LCTL get_param -n \ + mdd.$(facet_svc mds$i).lfsck_namespace | + awk '/^status/ { print \\\$2 }'" "completed" + done + + ls -R $DIR/$tdir || error "ls failed" + rm -rf $DIR/$tdir || error "rmdir failed" } run_test 60g "transaction abort won't cause MDT hung" @@ -7321,7 +7752,7 @@ cleanup_65k() { do_facet $SINGLEMDS lctl --device %$INACTIVE_OSC activate echo $INACTIVE_OSC "is Activate" - wait_osc_import_state mds ost$ostnum FULL + wait_osc_import_state mds ost$(( ostnum + 1 )) FULL } test_65k() { # bug11679 @@ -7381,7 +7812,7 @@ test_65k() { # bug11679 do_facet $SINGLEMDS lctl --device %$INACTIVE_OSC activate echo $INACTIVE_OSC "is Activate" - wait_osc_import_state mds ost$ostnum FULL + wait_osc_import_state mds ost$(( ostnum + 1 )) FULL done } run_test 65k "validate manual striping works properly with deactivated OSCs" @@ -7450,8 +7881,13 @@ test_65n() { local dir3=$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 @@ -7473,8 +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 $dir4 should inherit the pool from # the filesystem default @@ -7493,10 +7935,33 @@ test_65n() { local dir5=$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 $lmv_count" } run_test 65n "don't inherit default layout from root for new subdirectories" @@ -7740,9 +8205,22 @@ CKSUM_TYPES=${CKSUM_TYPES:-$(lctl get_param -n osc.*osc-[^mM]*.checksum_type | set_checksum_type() { lctl set_param -n osc.*osc-[^mM]*.checksum_type $1 - log "set checksum type to $1" - return 0 + rc=$? + log "set checksum type to $1, rc = $rc" + return $rc +} + +get_osc_checksum_type() +{ + # arugment 1: OST name, like OST0000 + ost=$1 + checksum_type=$(lctl get_param -n osc.*${ost}-osc-[^mM]*.checksum_type | + sed 's/.*\[\(.*\)\].*/\1/g') + rc=$? + [ $rc -ne 0 ] && error "failed to get checksum type of $ost, rc = $rc, output = $checksum_type" + echo $checksum_type } + F77_TMP=$TMP/f77-temp F77SZ=8 setup_f77() { @@ -7971,7 +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 @@ -7982,18 +8460,50 @@ 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" +test_77l() { + [ $PARALLEL == "yes" ] && skip "skip parallel run" + $GSS && skip_env "could not run with gss" + + set_checksums 1 + stack_trap "set_checksums $ORIG_CSUM" EXIT + stack_trap "set_checksum_type $ORIG_CSUM_TYPE" EXIT + + set_checksum_type invalid && error "unexpected success of invalid checksum type" + + $LFS setstripe -c 1 -i 0 $DIR/$tfile + for algo in $CKSUM_TYPES; do + set_checksum_type $algo || error "fail to set checksum type $algo" + osc_algo=$(get_osc_checksum_type OST0000) + [ "$osc_algo" != "$algo" ] && error "checksum type is $osc_algo after setting it to $algo" + + # no locks, no reqs to let the connection idle + cancel_lru_locks osc + lru_resize_disable osc + wait_osc_import_state client ost1 IDLE + + # ensure ost1 is connected + stat $DIR/$tfile >/dev/null || error "can't stat" + wait_osc_import_state client ost1 FULL + + osc_algo=$(get_osc_checksum_type OST0000) + [ "$osc_algo" != "$algo" ] && error "checksum type changed from $algo to $osc_algo after reconnection" + done + return 0 +} +run_test 77l "preferred checksum type is remembered after reconnected" + [ "$ORIG_CSUM" ] && set_checksums $ORIG_CSUM || true rm -f $F77_TMP unset F77_TMP @@ -8087,31 +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" @@ -8255,7 +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 @@ -8280,9 +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 @@ -8392,12 +8908,14 @@ test_101c() { cancel_lru_locks osc $LCTL set_param osc.*.rpc_stats 0 $READS -f $DIR/$tfile -s$FILE_LENGTH -b$rsize -n$nreads -t 180 + $LCTL get_param osc.*.rpc_stats for osc_rpc_stats in $($LCTL get_param -N osc.*.rpc_stats); do local stats=$($LCTL get_param -n $osc_rpc_stats) local lines=$(echo "$stats" | awk 'END {print NR;}') local size if [ $lines -le 20 ]; then + echo "continue debug" continue fi for size in 1 2 4 8; do @@ -8617,12 +9135,32 @@ test_101g() { } run_test 101g "Big bulk(4/16 MiB) readahead" -setup_test102() { - test_mkdir $DIR/$tdir - chown $RUNAS_ID $DIR/$tdir - STRIPE_SIZE=65536 - STRIPE_OFFSET=1 - STRIPE_COUNT=$OSTCOUNT +test_101h() { + $LFS setstripe -i 0 -c 1 $DIR/$tfile + + dd if=/dev/zero of=$DIR/$tfile bs=1M count=70 || + error "dd 70M file failed" + echo Cancel LRU locks on lustre client to flush the client cache + cancel_lru_locks osc + + echo "Reset readahead stats" + $LCTL set_param -n llite.*.read_ahead_stats 0 + + echo "Read 10M of data but cross 64M bundary" + dd if=$DIR/$tfile of=/dev/null bs=10M skip=6 count=1 + local miss=$($LCTL get_param -n llite.*.read_ahead_stats | + get_named_value 'misses' | cut -d" " -f1 | calc_total) + [ $miss -eq 1 ] || error "expected miss 1 but got $miss" + rm -f $p $DIR/$tfile +} +run_test 101h "Readahead should cover current read window" + +setup_test102() { + test_mkdir $DIR/$tdir + chown $RUNAS_ID $DIR/$tdir + STRIPE_SIZE=65536 + STRIPE_OFFSET=1 + STRIPE_COUNT=$OSTCOUNT [[ $OSTCOUNT -gt 4 ]] && STRIPE_COUNT=4 trap cleanup_test102 EXIT @@ -10833,6 +11371,47 @@ test_124c() { } run_test 124c "LRUR cancel very aged locks" +test_124d() { + [ $PARALLEL == "yes" ] && skip "skip parallel run" + $LCTL get_param -n mdc.*.connect_flags | grep -q lru_resize || + skip_env "no lru resize on server" + + # cache ununsed locks on client + local nr=100 + + lru_resize_disable mdc + stack_trap "lru_resize_enable mdc" EXIT + + cancel_lru_locks mdc + + # asynchronous object destroy at MDT could cause bl ast to client + test_mkdir $DIR/$tdir + createmany -o $DIR/$tdir/f $nr || + error "failed to create $nr files in $DIR/$tdir" + stack_trap "unlinkmany $DIR/$tdir/f $nr" EXIT + + ls -l $DIR/$tdir > /dev/null + + local nsdir="ldlm.namespaces.*-MDT0000-mdc-*" + local unused=$($LCTL get_param -n $nsdir.lock_unused_count) + local max_age=$($LCTL get_param -n $nsdir.lru_max_age) + local recalc_p=$($LCTL get_param -n $nsdir.pool.recalc_period) + + echo "unused=$unused, max_age=$max_age, recalc_p=$recalc_p" + + # set lru_max_age to 1 sec + $LCTL set_param $nsdir.lru_max_age=1000 # milliseconds + stack_trap "$LCTL set_param -n $nsdir.lru_max_age $max_age" EXIT + + echo "sleep $((recalc_p * 2)) seconds..." + sleep $((recalc_p * 2)) + + local remaining=$($LCTL get_param -n $nsdir.lock_unused_count) + + [ $remaining -eq 0 ] || error "$remaining locks are not canceled" +} +run_test 124d "cancel very aged locks if lru-resize diasbaled" + test_125() { # 13358 $LCTL get_param -n llite.*.client_type | grep -q local || skip "must run as local client" @@ -10862,41 +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" @@ -10904,6 +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 $LCTL set_param llite.*.stats=0 # perform 2 reads and writes so MAX is different from SUM. @@ -10914,30 +11506,28 @@ test_127b() { # bug LU-333 dd if=$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 @@ -10946,11 +11536,78 @@ 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" +test_127c() { # LU-12394 + [ "$OSTCOUNT" -lt "2" ] && skip_env "needs >= 2 OSTs" + local size + local bsize + local reads + local writes + local count + + $LCTL set_param llite.*.extents_stats=1 + stack_trap "$LCTL set_param llite.*.extents_stats=0" EXIT + + # Use two stripes so there is enough space in default config + $LFS setstripe -c 2 $DIR/$tfile + + # Extent stats start at 0-4K and go in power of two buckets + # LL_HIST_START = 12 --> 2^12 = 4K + # We do 3K*2^i, so 3K, 6K, 12K, 24K... hitting each bucket. + # We do not do buckets larger than 64 MiB to avoid ENOSPC issues on + # small configs + for size in 3K 6K 12K 24K 48K 96K 192K 384K 768K 1536K 3M 6M 12M 24M 48M; + do + # Write and read, 2x each, second time at a non-zero offset + dd if=/dev/zero of=$DIR/$tfile bs=$size count=1 + dd if=/dev/zero of=$DIR/$tfile bs=$size count=1 seek=10 + dd if=$DIR/$tfile of=/dev/null bs=$size count=1 + dd if=$DIR/$tfile of=/dev/null bs=$size count=1 seek=10 + rm -f $DIR/$tfile + done + + $LCTL get_param llite.*.extents_stats + + count=2 + for bsize in 4K 8K 16K 32K 64K 128K 256K 512K 1M 2M 4M 8M 16M 32M 64M; + do + local bucket=$($LCTL get_param -n llite.*.extents_stats | + grep -m 1 $bsize) + reads=$(echo $bucket | awk '{print $5}') + writes=$(echo $bucket | awk '{print $9}') + [ "$reads" -eq $count ] || + error "$reads reads in < $bsize bucket, expect $count" + [ "$writes" -eq $count ] || + error "$writes writes in < $bsize bucket, expect $count" + done + + # Test mmap write and read + $LCTL set_param llite.*.extents_stats=c + size=512 + dd if=/dev/zero of=$DIR/$tfile bs=${size}K count=1 + $MULTIOP $DIR/$tfile OSMRUc || error "$MULTIOP $DIR/$tfile failed" + $MULTIOP $DIR/$tfile OSMWUc || error "$MULTIOP $DIR/$tfile failed" + + $LCTL get_param llite.*.extents_stats + + count=$(((size*1024) / PAGE_SIZE)) + + bsize=$((2 * PAGE_SIZE / 1024))K + + bucket=$($LCTL get_param -n llite.*.extents_stats | + grep -m 1 $bsize) + reads=$(echo $bucket | awk '{print $5}') + writes=$(echo $bucket | awk '{print $9}') + # mmap writes fault in the page first, creating an additonal read + [ "$reads" -eq $((2 * count)) ] || + error "$reads reads in < $bsize bucket, expect $count" + [ "$writes" -eq $count ] || + error "$writes writes in < $bsize bucket, expect $count" +} +run_test 127c "test llite extent stats with regular & mmap i/o" + test_128() { # bug 15212 touch $DIR/$tfile $LFS 2>&1 <<-EOF | tee $TMP/$tfile.log @@ -12130,8 +12787,9 @@ 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 @@ -12776,9 +13434,9 @@ 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." @@ -12787,7 +13445,7 @@ 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 @@ -12802,20 +13460,23 @@ 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." @@ -12826,20 +13487,23 @@ 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" @@ -12855,7 +13519,7 @@ 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 @@ -12873,7 +13537,7 @@ 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 @@ -12884,7 +13548,7 @@ 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 @@ -13566,6 +14230,77 @@ 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" @@ -13762,7 +14497,8 @@ test_161d() { ps -p $pid [[ $? -eq 0 ]] || error "create should be blocked" - local tempfile=$(mktemp) + local tempfile="$(mktemp --tmpdir $tfile.XXXXXX)" + stack_trap "rm -f $tempfile" fid=$(changelog_extract_field "CREAT" "$tfile" "t=") cat $MOUNT/.lustre/fid/$fid 2>/dev/null >$tempfile || error "cat failed" # some delay may occur during ChangeLog publishing and file read just @@ -14424,6 +15160,41 @@ test_185() { # LU-2441 } run_test 185 "Volatile file support" +function create_check_volatile() { + local idx=$1 + local tgt + + $MULTIOP $MOUNT/.lustre/fid V${idx}Fw4096_c >&/tmp/${tfile}.fid & + local PID=$! + sleep 1 + local FID=$(cat /tmp/${tfile}.fid) + [ "$FID" == "" ] && error "can't get FID for volatile" + $CHECKSTAT -t file $MOUNT/.lustre/fid/$FID || error "can't stat $FID" + tgt=$($LFS getstripe -m $MOUNT/.lustre/fid/$FID) + [ "$tgt" != "$idx" ] && error "wrong MDS $tgt, expected $idx" + kill -USR1 $PID + wait + sleep 1 + cancel_lru_locks mdc # flush opencache + $CHECKSTAT -t file $MOUNT/.lustre/fid/$FID && error "can stat $FID" + return 0 +} + +test_185a(){ + # LU-12516 - volatile creation via .lustre + [[ $MDS1_VERSION -ge $(version_code 2.12.55) ]] || + skip "Need MDS version at least 2.3.55" + + create_check_volatile 0 + [ $MDSCOUNT -lt 2 ] && return 0 + + # DNE case + create_check_volatile 1 + + return 0 +} +run_test 185a "Volatile file creation in .lustre/fid/" + test_187a() { remote_mds_nodsh && skip "remote MDS with nodsh" [ $MDS1_VERSION -lt $(version_code 2.3.0) ] && @@ -14482,10 +15253,6 @@ test_200() { local subdir=$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; } @@ -14519,9 +15286,6 @@ test_200() { destroy_test_pools - if ! combined_mgs_mds ; then - umount_mgs_client - fi return $rc } run_test 200 "OST pools" @@ -15203,10 +15967,6 @@ 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$((OSTIDX + 1)) lctl set_param fail_loc=0x229 @@ -15243,10 +16003,6 @@ test_220() { #LU-325 error "$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" @@ -16250,6 +17006,61 @@ 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 @@ -16600,7 +17411,7 @@ test_243() } run_test 243 "various group lock tests" -test_244() +test_244a() { test_mkdir $DIR/$tdir dd if=/dev/zero of=$DIR/$tdir/$tfile bs=1M count=35 @@ -16608,7 +17419,26 @@ test_244() error "sendfile+grouplock failed" rm -rf $DIR/$tdir } -run_test 244 "sendfile with group lock tests" +run_test 244a "sendfile with group lock tests" + +test_244b() +{ + [ $PARALLEL == "yes" ] && skip "skip parallel run" && return + + local threads=50 + local size=$((1024*1024)) + + test_mkdir $DIR/$tdir + for i in $(seq 1 $threads); do + local file=$DIR/$tdir/file_$((i / 10)) + $MULTIOP $file OG1234w$size_$((i % 3))w$size_$((i % 4))g1234c & + local pids[$i]=$! + done + for i in $(seq 1 $threads); do + wait ${pids[$i]} + done +} +run_test 244b "multi-threaded write with group lock" test_245() { local flagname="multi_mod_rpcs" @@ -16635,21 +17465,6 @@ 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=$1 @@ -16742,7 +17557,7 @@ test_247e() { } run_test 247e "mount .. as fileset" -test_248() { +test_248a() { local fast_read_sav=$($LCTL get_param -n llite.*.fast_read 2>/dev/null) [ -z "$fast_read_sav" ] && skip "no fast read support" @@ -16793,7 +17608,89 @@ 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) ] && @@ -16908,9 +17805,6 @@ 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" @@ -16956,10 +17850,6 @@ 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" @@ -17265,7 +18155,7 @@ 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 @@ -17380,7 +18270,8 @@ test_256() { #after mount new plainllog is used touch $DIR/$tdir/{11..19} - local tmpfile=$(mktemp -u $tfile.XXXXXX) + local tmpfile="$(mktemp --tmpdir -u $tfile.XXXXXX)" + stack_trap "rm -f $tmpfile" cat_sl=$(do_facet $SINGLEMDS "sync; \ $DEBUGFS -c -R 'dump changelog_catalog $tmpfile' $mdt_dev; \ llog_reader $tmpfile | grep -c type=1064553b") @@ -17392,7 +18283,7 @@ test_256() { cat_sl=$(do_facet $SINGLEMDS "sync; \ $DEBUGFS -c -R 'dump changelog_catalog $tmpfile' $mdt_dev; \ - llog_reader $tmpfile | grep -c type=1064553b; rm -f $tmpfile") + llog_reader $tmpfile | grep -c type=1064553b") if (( cat_sl == 2 )); then error "Empty plain llog was not deleted from changelog catalog" @@ -18016,8 +18907,8 @@ test_271f() { local mdtidx=$($LFS getstripe --mdt-index $DIR/$tdir) cancel_lru_locks mdc - dd if=/dev/urandom of=$tmp bs=200000 count=1 - dd if=$tmp of=$dom bs=200000 count=1 + dd if=/dev/urandom of=$tmp bs=265000 count=1 + dd if=$tmp of=$dom bs=265000 count=1 cancel_lru_locks mdc cat /etc/hosts >> $tmp lctl set_param -n mdc.*.stats=clear @@ -18044,6 +18935,7 @@ test_271f() { local ra=$(get_mdc_stats $mdtidx req_active) local rw=$(get_mdc_stats $mdtidx req_waittime) + [ -z $num ] && num=0 [ $num -eq 1 ] || error "expect 1 READ RPC, $num occured" [ $ra == $rw ] || error "$((ra - rw)) resend occured" echo "... DONE" @@ -18123,12 +19015,12 @@ 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 != $new_md5" # Skip free space checks with ZFS @@ -18168,7 +19060,7 @@ test_272c() { cancel_lru_locks mdc local new_md5=$(md5sum $dom) - [ "$old_md5" != "$new_md5" ] && + [ "$old_md5" == "$new_md5" ] || error "$old_md5 != $new_md5" # Skip free space checks with ZFS @@ -18182,31 +19074,133 @@ test_272c() { } run_test 272c "DoM migration: DOM file to the OST-striped file (composite)" -test_273a() { - [ $MDS1_VERSION -lt $(version_code 2.11.50) ] && - skip "Need MDS version at least 2.11.50" - - # Layout swap cannot be done if either file has DOM component, - # this will never be supported, migration should be used instead +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 - $LFS setstripe -c2 ${dom}_plain - $LFS setstripe -E 1M -L mdt -E -1 -c2 ${dom}_dom - $LFS swap_layouts ${dom}_plain ${dom}_dom && - error "can swap layout with DoM component" - $LFS swap_layouts ${dom}_dom ${dom}_plain && - error "can swap layout with DoM component" + local mdtidx=$($LFS getstripe -m $dom) + local mdtname=MDT$(printf %04x $mdtidx) + local facet=mds$((mdtidx + 1)) - $LFS setstripe -E 1M -c1 -E -1 -c2 ${dom}_comp - $LFS swap_layouts ${dom}_comp ${dom}_dom && - error "can swap layout with DoM component" - $LFS swap_layouts ${dom}_dom ${dom}_comp && - error "can swap layout with DoM component" - return 0 -} -run_test 273a "DoM: layout swapping should fail with DOM" + dd if=/dev/urandom of=$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 $(version_code 2.11.50) ] && + skip "Need MDS version at least 2.11.50" + + # Layout swap cannot be done if either file has DOM component, + # this will never be supported, migration should be used instead + + local dom=$DIR/$tdir/$tfile + mkdir -p $DIR/$tdir + + $LFS setstripe -c2 ${dom}_plain + $LFS setstripe -E 1M -L mdt -E -1 -c2 ${dom}_dom + $LFS swap_layouts ${dom}_plain ${dom}_dom && + error "can swap layout with DoM component" + $LFS swap_layouts ${dom}_dom ${dom}_plain && + error "can swap layout with DoM component" + + $LFS setstripe -E 1M -c1 -E -1 -c2 ${dom}_comp + $LFS swap_layouts ${dom}_comp ${dom}_dom && + error "can swap layout with DoM component" + $LFS swap_layouts ${dom}_dom ${dom}_comp && + error "can swap layout with DoM component" + return 0 +} +run_test 273a "DoM: layout swapping should fail with DOM" test_275() { remote_ost_nodsh && skip "remote OST with nodsh" @@ -18260,6 +19254,48 @@ test_276() { } run_test 276 "Race between mount and obd_statfs" +test_277() { + $LCTL set_param ldlm.namespaces.*.lru_size=0 + dd if=/dev/zero of=$DIR/$tfile bs=1M count=1 + local cached_mb=$($LCTL get_param llite.*.max_cached_mb | + grep ^used_mb | awk '{print $2}') + [ $cached_mb -eq 1 ] || error "expected mb 1 got $cached_mb" + dd if=/dev/zero of=$DIR/$tfile bs=1M count=1 \ + oflag=direct conv=notrunc + cached_mb=$($LCTL get_param llite.*.max_cached_mb | + grep ^used_mb | awk '{print $2}') + [ $cached_mb -eq 0 ] || error "expected mb 0 got $cached_mb" +} +run_test 277 "Direct IO shall drop page cache" + +test_278() { + [ $PARALLEL == "yes" ] && skip "skip parallel run" && return + [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return + [[ "$(facet_host mds1)" != "$(facet_host mds2)" ]] && + skip "needs the same host for mdt1 mdt2" && return + + local pid1 + local pid2 + +#define OBD_FAIL_OBD_STOP_MDS_RACE 0x60b + do_facet mds2 $LCTL set_param fail_loc=0x8000060c + stop mds2 & + pid2=$! + + stop mds1 + + echo "Starting MDTs" + start mds1 $(mdsdevname 1) $MDS_MOUNT_OPTS + wait $pid2 +#For the error assertion will happen. lu_env_get_key(..., &mdt_thread_key) +#will return NULL + 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" + cleanup_test_300() { trap 0 umask $SAVE_UMASK @@ -18742,8 +19778,9 @@ test_300k() { # this test needs a huge transaction local kb - kb=$(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 @@ -19273,14 +20310,13 @@ run_test 316 "lfs mv" test_317() { [ $MDS1_VERSION -lt $(version_code 2.11.53) ] && skip "Need MDS version at least 2.11.53" + if [ "$ost1_FSTYPE" == "zfs" ]; then + skip "LU-10370: no implementation for ZFS" + fi + local trunc_sz local grant_blk_size - if [ "$(facet_fstype $facet)" == "zfs" ]; then - skip "LU-10370: no implementation for ZFS" && return - fi - - stack_trap "rm -f $DIR/$tfile" EXIT grant_blk_size=$($LCTL get_param osc.$FSNAME*.import | awk '/grant_block_size:/ { print $2; exit; }') # @@ -19288,7 +20324,8 @@ test_317() { # blocks count. # dd if=/dev/zero of=$DIR/$tfile bs=5M count=1 conv=fsync || - error "Create file : $DIR/$tfile" + error "Create file $DIR/$tfile failed" + stack_trap "rm -f $DIR/$tfile" EXIT for trunc_sz in 2097152 4097 4000 509 0; do $TRUNCATE $DIR/$tfile $trunc_sz || @@ -19425,7 +20462,7 @@ 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 [ $blocks -gt 1000 ] && blocks=1000 # 1G in maximum @@ -19736,9 +20773,6 @@ test_406() { local def_stripe_size=$($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" @@ -19791,10 +20825,6 @@ 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" @@ -19945,174 +20975,247 @@ 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 + 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 - $LCTL set_param lmv.*.qos_maxage=1 - sleep_maxage + echo + echo "Check for uneven MDTs: " local ffree + local bavail local max local min local max_index local min_index + local tmp - ffree=($(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 @@ -20348,6 +21451,278 @@ test_420() } run_test 420 "clear SGID bit on non-directories for non-members" +test_421a() { + local cnt + local fid1 + local fid2 + + [ $MDS1_VERSION -lt $(version_code 2.12.54) ] && + skip "Need MDS version at least 2.12.54" + + test_mkdir $DIR/$tdir + createmany -o $DIR/$tdir/f 3 + cnt=$(ls -1 $DIR/$tdir | wc -l) + [ $cnt != 3 ] && error "unexpected #files: $cnt" + + fid1=$(lfs path2fid $DIR/$tdir/f1) + fid2=$(lfs path2fid $DIR/$tdir/f2) + $LFS rmfid $DIR $fid1 $fid2 || error "rmfid failed" + + stat $DIR/$tdir/f1 && error "f1 still visible on the client" + stat $DIR/$tdir/f2 && error "f2 still visible on the client" + + cnt=$(ls -1 $DIR/$tdir | wc -l) + [ $cnt == 1 ] || error "unexpected #files after: $cnt" + + rm -f $DIR/$tdir/f3 || error "can't remove f3" + createmany -o $DIR/$tdir/f 3 + cnt=$(ls -1 $DIR/$tdir | wc -l) + [ $cnt != 3 ] && error "unexpected #files: $cnt" + + fid1=$(lfs path2fid $DIR/$tdir/f1) + fid2=$(lfs path2fid $DIR/$tdir/f2) + echo "remove using fsname $FSNAME" + $LFS rmfid $FSNAME $fid1 $fid2 || error "rmfid with fsname failed" + + cnt=$(ls -1 $DIR/$tdir | wc -l) + [ $cnt == 1 ] || error "unexpected #files after: $cnt" +} +run_test 421a "simple rm by fid" + +test_421b() { + local cnt + local FID1 + local FID2 + + [ $MDS1_VERSION -lt $(version_code 2.12.54) ] && + skip "Need MDS version at least 2.12.54" + + test_mkdir $DIR/$tdir + createmany -o $DIR/$tdir/f 3 + multiop_bg_pause $DIR/$tdir/f1 o_c || error "multiop failed to start" + MULTIPID=$! + + FID1=$(lfs path2fid $DIR/$tdir/f1) + FID2=$(lfs path2fid $DIR/$tdir/f2) + $LFS rmfid $DIR $FID1 $FID2 && error "rmfid didn't fail" + + kill -USR1 $MULTIPID + wait + + cnt=$(ls $DIR/$tdir | wc -l) + [ $cnt == 2 ] || error "unexpected #files after: $cnt" +} +run_test 421b "rm by fid on open file" + +test_421c() { + local cnt + local FIDS + + [ $MDS1_VERSION -lt $(version_code 2.12.54) ] && + skip "Need MDS version at least 2.12.54" + + test_mkdir $DIR/$tdir + createmany -o $DIR/$tdir/f 3 + touch $DIR/$tdir/$tfile + createmany -l$DIR/$tdir/$tfile $DIR/$tdir/h 180 + cnt=$(ls -1 $DIR/$tdir | wc -l) + [ $cnt != 184 ] && error "unexpected #files: $cnt" + + FID1=$(lfs path2fid $DIR/$tdir/$tfile) + $LFS rmfid $DIR $FID1 || error "rmfid failed" + + cnt=$(ls $DIR/$tdir | wc -l) + [ $cnt == 3 ] || error "unexpected #files after: $cnt" +} +run_test 421c "rm by fid against hardlinked files" + +test_421d() { + local cnt + local FIDS + + [ $MDS1_VERSION -lt $(version_code 2.12.54) ] && + skip "Need MDS version at least 2.12.54" + + test_mkdir $DIR/$tdir + createmany -o $DIR/$tdir/f 4097 + cnt=$(ls -1 $DIR/$tdir | wc -l) + [ $cnt != 4097 ] && error "unexpected #files: $cnt" + + FIDS=$(lfs path2fid $DIR/$tdir/f* | sed "s/[/][^:]*://g") + $LFS rmfid $DIR $FIDS || error "rmfid failed" + + cnt=$(ls $DIR/$tdir | wc -l) + rm -rf $DIR/$tdir + [ $cnt == 0 ] || error "unexpected #files after: $cnt" +} +run_test 421d "rmfid en masse" + +test_421e() { + local cnt + local FID + + [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" + [ $MDS1_VERSION -lt $(version_code 2.12.54) ] && + skip "Need MDS version at least 2.12.54" + + mkdir -p $DIR/$tdir + $LFS setdirstripe -c$MDSCOUNT $DIR/$tdir/striped_dir + createmany -o $DIR/$tdir/striped_dir/f 512 + cnt=$(ls -1 $DIR/$tdir/striped_dir | wc -l) + [ $cnt != 512 ] && error "unexpected #files: $cnt" + + FIDS=$(lfs path2fid $DIR/$tdir/striped_dir/f* | + sed "s/[/][^:]*://g") + $LFS rmfid $DIR $FIDS || error "rmfid failed" + + cnt=$(ls $DIR/$tdir/striped_dir | wc -l) + rm -rf $DIR/$tdir + [ $cnt == 0 ] || error "unexpected #files after: $cnt" +} +run_test 421e "rmfid in DNE" + +test_421f() { + local cnt + local FID + + [ $MDS1_VERSION -lt $(version_code 2.12.54) ] && + skip "Need MDS version at least 2.12.54" + + test_mkdir $DIR/$tdir + touch $DIR/$tdir/f + cnt=$(ls -1 $DIR/$tdir | wc -l) + [ $cnt != 1 ] && error "unexpected #files: $cnt" + + FID=$(lfs path2fid $DIR/$tdir/f) + $RUNAS $LFS rmfid $DIR $FID && error "rmfid didn't fail (1)" + # rmfid should fail + cnt=$(ls -1 $DIR/$tdir | wc -l) + [ $cnt != 1 ] && error "unexpected #files after (2): $cnt" + + chmod a+rw $DIR/$tdir + ls -la $DIR/$tdir + $RUNAS $LFS rmfid $DIR $FID && error "rmfid didn't fail (2)" + # rmfid should fail + cnt=$(ls -1 $DIR/$tdir | wc -l) + [ $cnt != 1 ] && error "unexpected #files after (3): $cnt" + + rm -f $DIR/$tdir/f + $RUNAS touch $DIR/$tdir/f + FID=$(lfs path2fid $DIR/$tdir/f) + echo "rmfid as root" + $LFS rmfid $DIR $FID || error "rmfid as root failed" + cnt=$(ls -1 $DIR/$tdir | wc -l) + [ $cnt == 0 ] || error "unexpected #files after (4): $cnt" + + rm -f $DIR/$tdir/f + $RUNAS touch $DIR/$tdir/f + cnt=$(ls -1 $DIR/$tdir | wc -l) + [ $cnt != 1 ] && error "unexpected #files (4): $cnt" + FID=$(lfs path2fid $DIR/$tdir/f) + # rmfid w/o user_fid2path mount option should fail + $RUNAS $LFS rmfid $DIR $FID && error "rmfid didn't fail(3)" + cnt=$(ls -1 $DIR/$tdir | wc -l) + [ $cnt == 1 ] || error "unexpected #files after (5): $cnt" + + umount_client $MOUNT || error "failed to umount client" + mount_client $MOUNT "$MOUNT_OPTS,user_fid2path" || + error "failed to mount client'" + + $RUNAS $LFS rmfid $DIR $FID || error "rmfid failed" + # rmfid should succeed + cnt=$(ls -1 $DIR/$tdir | wc -l) + [ $cnt == 0 ] || error "unexpected #files after (6): $cnt" + + # rmfid shouldn't allow to remove files due to dir's permission + chmod a+rwx $DIR/$tdir + touch $DIR/$tdir/f + ls -la $DIR/$tdir + FID=$(lfs path2fid $DIR/$tdir/f) + $RUNAS $LFS rmfid $DIR $FID && error "rmfid didn't fail" + + umount_client $MOUNT || error "failed to umount client" + mount_client $MOUNT "$MOUNT_OPTS" || + error "failed to mount client'" + +} +run_test 421f "rmfid checks permissions" + +test_421g() { + local cnt + local FIDS + + [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" + [ $MDS1_VERSION -lt $(version_code 2.12.54) ] && + skip "Need MDS version at least 2.12.54" + + mkdir -p $DIR/$tdir + $LFS setdirstripe -c$MDSCOUNT $DIR/$tdir/striped_dir + createmany -o $DIR/$tdir/striped_dir/f 512 + cnt=$(ls -1 $DIR/$tdir/striped_dir | wc -l) + [ $cnt != 512 ] && error "unexpected #files: $cnt" + + FIDS=$(lfs path2fid $DIR/$tdir/striped_dir/f* | + sed "s/[/][^:]*://g") + + rm -f $DIR/$tdir/striped_dir/f1* + cnt=$(ls -1 $DIR/$tdir/striped_dir | wc -l) + removed=$((512 - cnt)) + + # few files have been just removed, so we expect + # rmfid to fail on their fids + errors=$($LFS rmfid $DIR $FIDS 2>&1 | wc -l) + [ $removed != $errors ] && error "$errors != $removed" + + cnt=$(ls $DIR/$tdir/striped_dir | wc -l) + rm -rf $DIR/$tdir + [ $cnt == 0 ] || error "unexpected #files after: $cnt" +} +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) ]] && @@ -20390,7 +21765,8 @@ test_801a() { echo "Start barrier_freeze at: $(date)" #define OBD_FAIL_BARRIER_DELAY 0x2202 do_facet mgs $LCTL set_param fail_val=5 fail_loc=0x2202 - do_facet mgs $LCTL barrier_freeze $FSNAME 10 & + # Do not reduce barrier time - See LU-11873 + do_facet mgs $LCTL barrier_freeze $FSNAME 20 & sleep 2 local b_status=$(barrier_stat) @@ -20412,7 +21788,8 @@ test_801a() { [ "$b_status" = "'expired'" ] || error "(3) unexpected barrier status $b_status" - do_facet mgs $LCTL barrier_freeze $FSNAME 10 || + # Do not reduce barrier time - See LU-11873 + do_facet mgs $LCTL barrier_freeze $FSNAME 20 || error "(4) fail to freeze barrier" b_status=$(barrier_stat) @@ -20537,7 +21914,8 @@ test_801c() { do_facet mgs $LCTL barrier_rescan $FSNAME || error "(3) Fail to rescan barrier bitmap" - do_facet mgs $LCTL barrier_freeze $FSNAME 10 + # Do not reduce barrier time - See LU-11873 + do_facet mgs $LCTL barrier_freeze $FSNAME 20 b_status=$(barrier_stat) [ "$b_status" = "'frozen'" ] || @@ -20562,6 +21940,7 @@ run_test 801c "rescan barrier bitmap" saved_MGS_MOUNT_OPTS=$MGS_MOUNT_OPTS saved_MDS_MOUNT_OPTS=$MDS_MOUNT_OPTS saved_OST_MOUNT_OPTS=$OST_MOUNT_OPTS +saved_MOUNT_OPTS=$MOUNT_OPTS cleanup_802a() { trap 0 @@ -20570,11 +21949,12 @@ cleanup_802a() { MGS_MOUNT_OPTS=$saved_MGS_MOUNT_OPTS MDS_MOUNT_OPTS=$saved_MDS_MOUNT_OPTS OST_MOUNT_OPTS=$saved_OST_MOUNT_OPTS + MOUNT_OPTS=$saved_MOUNT_OPTS setupall } 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" @@ -20605,7 +21985,8 @@ test_802a() { error "(4) Mount client without 'ro' should fail" echo "Mount client with ro should succeed" - mount_client $MOUNT ro || + MOUNT_OPTS=$(csa_add "$MOUNT_OPTS" -o ro) + mount_client $MOUNT || error "(5) Mount client with 'ro' should succeed" echo "Modify should be refused" @@ -20789,7 +22170,7 @@ cleanup_805() { } test_805() { - local zfs_version=$(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 $(version_code 0.7.2) ] && skip "netfree not implemented before 0.7" @@ -20801,7 +22182,7 @@ test_805() { local usedkb local old local quota - local pref="osd-zfs.lustre-MDT0000." + local pref="osd-zfs.$FSNAME-MDT0000." # limit available space on MDS dataset to meet nospace issue # quickly. then ZFS 0.7.2 can use reserved space if asked @@ -20899,7 +22280,7 @@ test_806() { done check_lsom_size $DIR/$tfile $size - # multi-client wirtes + # multi-client writes num=$(get_node_count ${CLIENTS//,/ }) size=$(($num * $bs)) offset=0 @@ -21064,20 +22445,31 @@ test_809() { run_test 809 "Verify no SOM xattr store for DoM-only files" test_810() { - local ORIG - local CSUM - - # t10 seem to dislike partial pages - lctl set_param osc.*.checksum_type=adler - lctl set_param fail_loc=0x411 - dd if=/dev/urandom of=$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)" @@ -21098,7 +22490,7 @@ 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 ] && @@ -21119,7 +22511,31 @@ 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 $DIR/ || error "lfs quota should succeed" + wait_osc_import_state client ost1 IDLE +} +run_test 812b "do not drop no resend request for idle connect" test_813() { local file_heat_sav=$($LCTL get_param -n llite.*.file_heat 2>/dev/null) @@ -21309,6 +22725,9 @@ test_815() run_test 815 "zero byte tiny write doesn't hang (LU-12382)" test_816() { + [ "$SHARED_KEY" = true ] && + skip "OSC connections never go IDLE with Shared-Keys enabled" + $LFS setstripe -c 1 -i 0 $DIR/$tfile # ensure ost1 is connected stat $DIR/$tfile >/dev/null || error "can't stat" @@ -21329,6 +22748,63 @@ test_816() { } run_test 816 "do not reset lru_resize on idle reconnect" +cleanup_817() { + umount $tmpdir + exportfs -u localhost:$DIR/nfsexp + rm -rf $DIR/nfsexp +} + +test_817() { + systemctl restart nfs-server.service || skip "failed to restart nfsd" + + mkdir -p $DIR/nfsexp + exportfs -orw,no_root_squash localhost:$DIR/nfsexp || + error "failed to export nfs" + + tmpdir=$(mktemp -d /tmp/nfs-XXXXXX) + stack_trap cleanup_817 EXIT + + mount -t nfs -orw localhost:$DIR/nfsexp $tmpdir || + error "failed to mount nfs to $tmpdir" + + cp /bin/true $tmpdir + $DIR/nfsexp/true || error "failed to execute 'true' command" +} +run_test 817 "nfsd won't cache write lock for exec file" + +test_818() { + mkdir $DIR/$tdir + $LFS setstripe -c1 -i0 $DIR/$tfile + $LFS setstripe -c1 -i1 $DIR/$tfile + stop $SINGLEMDS + #define OBD_FAIL_OSP_CANT_PROCESS_LLOG 0x2105 + do_facet $SINGLEMDS lctl set_param fail_loc=0x80002105 + start $SINGLEMDS $(mdsdevname ${SINGLEMDS//mds/}) $MDS_MOUNT_OPTS || + error "start $SINGLEMDS failed" + rm -rf $DIR/$tdir +} +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 # @@ -21347,6 +22823,31 @@ 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