X-Git-Url: https://git.whamcloud.com/?p=fs%2Flustre-release.git;a=blobdiff_plain;f=lustre%2Ftests%2Fsanity.sh;h=82d0f76e12de129bad2530928718461e74b23649;hp=34c8753c53e998851ce0e86915e39c67a6a05992;hb=f7392c7c4a16bc1127ee448f937ba81c50dcdfd5;hpb=94441071dbadbde56442e5e2560c642cc2793274 diff --git a/lustre/tests/sanity.sh b/lustre/tests/sanity.sh index 34c8753..82d0f76 100755 --- a/lustre/tests/sanity.sh +++ b/lustre/tests/sanity.sh @@ -41,12 +41,12 @@ init_logging ALWAYS_EXCEPT="$SANITY_EXCEPT " # bug number for skipped test: LU-9693 LU-6493 LU-9693 ALWAYS_EXCEPT+=" 42a 42b 42c " -# bug number: LU-8411 LU-9054 LU-13314 -ALWAYS_EXCEPT+=" 407 312 56ob" +# bug number: LU-8411 LU-9054 +ALWAYS_EXCEPT+=" 407 312" if $SHARED_KEY; then - # bug number: LU-9795 LU-9795 LU-9795 LU-9795 - ALWAYS_EXCEPT+=" 17n 60a 133g 300f" + # bug number: LU-14181 LU-14181 + ALWAYS_EXCEPT+=" 64e 64f" fi selinux_status=$(getenforce) @@ -61,10 +61,12 @@ if [[ $(uname -m) = aarch64 ]]; then ALWAYS_EXCEPT+=" $GRANT_CHECK_LIST" # bug number: LU-11671 LU-11667 ALWAYS_EXCEPT+=" 45 317" + # bug number: LU-14067 LU-14067 + ALWAYS_EXCEPT+=" 400a 400b" fi -# skip nfs tests on kernels >= 4.14.0 until they are fixed -if [ $LINUX_VERSION_CODE -ge $(version_code 4.14.0) ]; then +# skip nfs tests on kernels >= 4.12.0 until they are fixed +if [ $LINUX_VERSION_CODE -ge $(version_code 4.12.0) ]; then # bug number: LU-12661 ALWAYS_EXCEPT+=" 817" fi @@ -80,11 +82,16 @@ fi if [ "$mds1_FSTYPE" = "zfs" ]; then # bug number for skipped test: - ALWAYS_EXCEPT="$ALWAYS_EXCEPT " + ALWAYS_EXCEPT+=" " # 13 (min)" [ "$SLOW" = "no" ] && EXCEPT_SLOW="$EXCEPT_SLOW 51b" fi +if [ "$ost1_FSTYPE" = "zfs" ]; then + # bug number for skipped test: LU-1941 LU-1941 LU-1941 LU-1941 + ALWAYS_EXCEPT+=" 130a 130b 130c 130d 130e 130f 130g" +fi + # Get the SLES distro version # # Returns a version string that should only be used in comparing @@ -144,6 +151,13 @@ check_swap_layouts_support() skip "Does not support layout lock." } +check_swap_layout_no_dom() +{ + local FOLDER=$1 + local SUPP=$(lfs getstripe $FOLDER | grep "pattern: mdt" | wc -l) + [ $SUPP -eq 0 ] || skip "layout swap does not support DOM files so far" +} + check_and_setup_lustre DIR=${DIR:-$MOUNT} assert_DIR @@ -227,12 +241,13 @@ test_0d() { # LU-3397 [ "$exp_val" == "$imp_val" ] || error "export flags '$exp_val' != import flags '$imp_val'" - # Compare the value of client version + # Compare client versions. Only compare top-3 fields for compatibility exp_client_version=$(awk '/target_version:/ { print $2 }' $temp_exp) - exp_val=$(version_code $exp_client_version) - imp_val=$CLIENT_VERSION + exp_val=$(version_code $(cut -d. -f1,2,3 <<<$exp_client_version)) + imp_val=$(version_code $(lustre_build_version client | cut -d. -f1,2,3)) [ "$exp_val" == "$imp_val" ] || - error "export client version '$exp_val' != '$imp_val'" + error "exp version '$exp_client_version'($exp_val) != " \ + "'$(lustre_build_version client)'($imp_val)" } run_test 0d "check export proc =============================" @@ -1436,6 +1451,23 @@ test_24F () { } run_test 24F "hash order vs readdir (LU-11330)" +test_24G () { + [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" + + local ino1 + local ino2 + + $LFS mkdir -i 0 $DIR/$tdir-0 || error "mkdir $tdir-0" + $LFS mkdir -i 1 $DIR/$tdir-1 || error "mkdir $tdir-1" + touch $DIR/$tdir-0/f1 || error "touch f1" + ln -s $DIR/$tdir-0/f1 $DIR/$tdir-0/s1 || error "ln s1" + ino1=$(stat -c%i $DIR/$tdir-0/s1) + mv $DIR/$tdir-0/s1 $DIR/$tdir-1 || error "mv s1" + ino2=$(stat -c%i $DIR/$tdir-1/s1) + [ $ino1 -ne $ino2 ] || error "s1 should be migrated" +} +run_test 24G "migrate symlink in rename" + test_25a() { echo '== symlink sanity =============================================' @@ -1609,7 +1641,7 @@ test_27ce() { [[ $($LCTL get_param mdc.*.import) =~ connect_flags.*overstriping ]] || skip "server does not support overstriping" # We do one more stripe than we have OSTs - [ $OSTCOUNT -ge 159 ] || large_xattr_enabled || + [ $OSTCOUNT -lt 159 ] || large_xattr_enabled || skip_env "ea_inode feature disabled" test_mkdir -p $DIR/$tdir @@ -1876,6 +1908,57 @@ test_27o() { } run_test 27o "create file with all full OSTs (should error)" +function create_and_checktime() { + local fname=$1 + local loops=$2 + local i + + for ((i=0; i < $loops; i++)); do + local start=$SECONDS + multiop $fname-$i Oc + ((SECONDS-start < TIMEOUT)) || + error "creation took " $((SECONDS-$start)) && return 1 + done +} + +test_27oo() { + local mdts=$(comma_list $(mdts_nodes)) + + [ $MDS1_VERSION -lt $(version_code 2.13.57) ] && + skip "Need MDS version at least 2.13.57" + + local f0=$DIR/${tfile}-0 + local f1=$DIR/${tfile}-1 + + wait_delete_completed + + # refill precreated objects + $LFS setstripe -i0 -c1 $f0 + + saved=$(do_facet mds1 $LCTL get_param -n lov.*0000*.qos_threshold_rr) + # force QoS allocation policy + do_nodes $mdts $LCTL set_param lov.*.qos_threshold_rr=0% + stack_trap "do_nodes $mdts $LCTL set_param \ + lov.*.qos_threshold_rr=$saved" EXIT + sleep_maxage + + # one OST is unavailable, but still have few objects preallocated + stop ost1 + stack_trap "start ost1 $(ostdevname 1) $OST_MOUNT_OPTS; \ + rm -rf $f1 $DIR/$tdir*" EXIT + + for ((i=0; i < 7; i++)); do + mkdir $DIR/$tdir$i || error "can't create dir" + $LFS setstripe -c$((OSTCOUNT-1)) $DIR/$tdir$i || + error "can't set striping" + done + for ((i=0; i < 7; i++)); do + create_and_checktime $DIR/$tdir$i/$tfile 100 & + done + wait +} +run_test 27oo "don't let few threads to reserve too many objects" + test_27p() { [[ $OSTCOUNT -lt 2 ]] && skip_env "needs >= 2 OSTs" [ $PARALLEL == "yes" ] && skip "skip parallel run" @@ -2365,7 +2448,7 @@ test_27Cb() { test_mkdir -p $DIR/$tdir local setcount=$(($OSTCOUNT * 2)) - [ $setcount -ge 160 ] || large_xattr_enabled || + [ $setcount -lt 160 ] || large_xattr_enabled || skip_env "ea_inode feature disabled" $LFS setstripe -C $setcount $DIR/$tdir/$tfile || @@ -2391,7 +2474,7 @@ test_27Cc() { test_mkdir -p $DIR/$tdir local setcount=$(($OSTCOUNT - 1)) - [ $setcount -ge 160 ] || large_xattr_enabled || + [ $setcount -lt 160 ] || large_xattr_enabled || skip_env "ea_inode feature disabled" $LFS setstripe -C $setcount $DIR/$tdir/$tfile || @@ -2471,7 +2554,7 @@ test_27Cf() { test_mkdir -p $DIR/$tdir local setcount=$(($OSTCOUNT * 2)) - [ $setcount -ge 160 ] || large_xattr_enabled || + [ $setcount -lt 160 ] || large_xattr_enabled || skip_env "ea_inode feature disabled" $LFS setstripe -C $setcount $DIR/$tdir/ || @@ -2689,6 +2772,17 @@ test_27J() { local uuid2=$(cat /proc/sys/kernel/random/uuid) # create foreign file (raw way) + ! $LFS setstripe --flags 0xda08 $DIR/$tdir/$tfile || + error "creating $tfile w/ hex flags w/o --foreign should fail" + + ! $LFS setstripe --foreign --flags foo \ + --xattr ${uuid1}@${uuid2} $DIR/$tdir/$tfile || + error "creating $tfile with '--flags foo' should fail" + + ! $LFS setstripe --foreign --flags 0xffffffff \ + --xattr ${uuid1}@${uuid2} $DIR/$tdir/$tfile || + error "creating $tfile w/ 0xffffffff flags should fail" + create_foreign_file -f $DIR/$tdir/$tfile -x "${uuid1}@${uuid2}" \ -t 1 -F 0xda08 || error "create_foreign_file failed" @@ -2715,7 +2809,7 @@ test_27J() { error "$DIR/$tdir/$tfile: invalid LOV EA foreign value" # create foreign file (lfs + API) - $LFS setstripe --foreign=daos --flags 0xda08 \ + $LFS setstripe --foreign=none --flags 0xda08 \ -x "${uuid1}@${uuid2}" $DIR/$tdir/${tfile}2 || error "$DIR/$tdir/${tfile}2: create failed" @@ -2725,7 +2819,7 @@ test_27J() { # lfm_length is LOV EA size - sizeof(lfm_magic) - sizeof(lfm_length) $LFS getstripe -v $DIR/$tdir/${tfile}2 | grep "lfm_length:.*73" || error "$DIR/$tdir/${tfile}2: invalid LOV EA foreign size" - $LFS getstripe -v $DIR/$tdir/${tfile}2 | grep "lfm_type:.*daos" || + $LFS getstripe -v $DIR/$tdir/${tfile}2 | grep "lfm_type:.*none" || error "$DIR/$tdir/${tfile}2: invalid LOV EA foreign type" $LFS getstripe -v $DIR/$tdir/${tfile}2 | grep "lfm_flags:.*0x0000DA08" || @@ -2784,6 +2878,17 @@ test_27K() { local uuid2=$(cat /proc/sys/kernel/random/uuid) # create foreign dir (raw way) + ! $LFS setdirstripe --flags 0xda08 $DIR/$tdir/$tdir || + error "creating $tdir w/ hex flags w/o --foreign should fail" + + ! $LFS setdirstripe --foreign --flags foo \ + --xattr ${uuid1}@${uuid2} $DIR/$tdir/$tdir || + error "creating $tdir with '--flags foo' should fail" + + ! $LFS setdirstripe --foreign --flags 0xffffffff \ + --xattr ${uuid1}@${uuid2} $DIR/$tdir/$tdir || + error "creating $tdir w/ 0xffffffff flags should fail" + create_foreign_dir -d $DIR/$tdir/$tdir -x "${uuid1}@${uuid2}" -t 1 || error "create_foreign_dir FAILED" @@ -2795,7 +2900,8 @@ test_27K() { error "$DIR/$tdir/$tdir: invalid LMV EA size" parse_foreign_dir -d $DIR/$tdir/$tdir | grep "lmv_foreign_type: 1$" || error "$DIR/$tdir/$tdir: invalid LMV EA type" - parse_foreign_dir -d $DIR/$tdir/$tdir | grep "lmv_foreign_flags: 0$" || + parse_foreign_dir -d $DIR/$tdir/$tdir | + grep "lmv_foreign_flags: 55813$" || error "$DIR/$tdir/$tdir: invalid LMV EA flags" local lmv=$(parse_foreign_dir -d $DIR/$tdir/$tdir | grep "lmv_foreign_value: 0x" | @@ -2805,7 +2911,7 @@ test_27K() { [[ $lmv == $lmv2 ]] || error "$DIR/$tdir/$tdir: invalid LMV EA value" # create foreign dir (lfs + API) - $LFS mkdir --foreign=daos --xattr="${uuid1}@${uuid2}" --flags=0xda05 \ + $LFS mkdir --foreign=none --xattr="${uuid1}@${uuid2}" --flags=0xda05 \ $DIR/$tdir/${tdir}2 || error "$DIR/$tdir/${tdir}2: create failed" @@ -2816,7 +2922,7 @@ test_27K() { # - sizeof(lfm_type) - sizeof(lfm_flags) $LFS getdirstripe -v $DIR/$tdir/${tdir}2 | grep "lfm_length:.*73" || error "$DIR/$tdir/${tdir}2: invalid LMV EA size" - $LFS getdirstripe -v $DIR/$tdir/${tdir}2 | grep "lfm_type:.*daos" || + $LFS getdirstripe -v $DIR/$tdir/${tdir}2 | grep "lfm_type:.*none" || error "$DIR/$tdir/${tdir}2: invalid LMV EA type" $LFS getdirstripe -v $DIR/$tdir/${tdir}2 | grep "lfm_flags:.*0x0000DA05" || @@ -3015,6 +3121,204 @@ test_27N() { } run_test 27N "lctl pool_list on separate MGS gives correct pool name" +clean_foreign_symlink() { + trap 0 + lctl set_param llite/$FSNAME-*/foreign_symlink_enable=0 + for i in $DIR/$tdir/* ; do + $LFS unlink_foreign $i || true + done +} + +test_27O() { + [[ $(lustre_version_code $SINGLEMDS) -le $(version_code 2.12.51) ]] && + skip "Need MDS version newer than 2.12.51" + + test_mkdir $DIR/$tdir + local uuid1=$(cat /proc/sys/kernel/random/uuid) + local uuid2=$(cat /proc/sys/kernel/random/uuid) + + trap clean_foreign_symlink EXIT + + # enable foreign_symlink behaviour + $LCTL set_param llite/$FSNAME-*/foreign_symlink_enable=1 + + # foreign symlink LOV format is a partial path by default + + # create foreign file (lfs + API) + $LFS setstripe --foreign=symlink --flags 0xda05 \ + -x "${uuid1}/${uuid2}" --mode 0600 $DIR/$tdir/${tfile} || + error "$DIR/$tdir/${tfile}: create failed" + + $LFS getstripe -v $DIR/$tdir/${tfile} | + grep "lfm_magic:.*0x0BD70BD0" || + error "$DIR/$tdir/${tfile}: invalid LOV EA foreign magic" + $LFS getstripe -v $DIR/$tdir/${tfile} | grep "lfm_type:.*symlink" || + error "$DIR/$tdir/${tfile}: invalid LOV EA foreign type" + $LFS getstripe -v $DIR/$tdir/${tfile} | + grep "lfm_flags:.*0x0000DA05" || + error "$DIR/$tdir/${tfile}: invalid LOV EA foreign flags" + $LFS getstripe $DIR/$tdir/${tfile} | + grep "lfm_value:.*${uuid1}/${uuid2}" || + error "$DIR/$tdir/${tfile}: invalid LOV EA foreign value" + + # modify striping should fail + $LFS setstripe -c 2 $DIR/$tdir/$tfile && + error "$DIR/$tdir/$tfile: setstripe should fail" + + # R/W should fail ("/{foreign_symlink_prefix}/${uuid1}/" missing) + cat $DIR/$tdir/$tfile && error "$DIR/$tdir/$tfile: read should fail" + cat /etc/passwd > $DIR/$tdir/$tfile && + error "$DIR/$tdir/$tfile: write should fail" + + # rename should succeed + mv $DIR/$tdir/$tfile $DIR/$tdir/${tfile}.new || + error "$DIR/$tdir/$tfile: rename has failed" + + #remove foreign_symlink file should fail + rm $DIR/$tdir/${tfile}.new && + error "$DIR/$tdir/${tfile}.new: remove of foreign_symlink file should fail" + + #test fake symlink + mkdir /tmp/${uuid1} || + error "/tmp/${uuid1}: mkdir has failed" + echo FOOFOO > /tmp/${uuid1}/${uuid2} || + error "/tmp/${uuid1}/${uuid2}: echo has failed" + $LCTL set_param llite/$FSNAME-*/foreign_symlink_prefix=/tmp/ + $CHECKSTAT -t link -l /tmp/${uuid1}/${uuid2} $DIR/$tdir/${tfile}.new || + error "$DIR/$tdir/${tfile}.new: not seen as a symlink" + #read should succeed now + cat $DIR/$tdir/${tfile}.new | grep FOOFOO || + error "$DIR/$tdir/${tfile}.new: symlink resolution has failed" + #write should succeed now + cat /etc/passwd > $DIR/$tdir/${tfile}.new || + error "$DIR/$tdir/${tfile}.new: write should succeed" + diff /etc/passwd $DIR/$tdir/${tfile}.new || + error "$DIR/$tdir/${tfile}.new: diff has failed" + diff /etc/passwd /tmp/${uuid1}/${uuid2} || + error "/tmp/${uuid1}/${uuid2}: diff has failed" + + #check that getstripe still works + $LFS getstripe $DIR/$tdir/${tfile}.new || + error "$DIR/$tdir/${tfile}.new: getstripe should still work with foreign_symlink enabled" + + # chmod should still succeed + chmod 644 $DIR/$tdir/${tfile}.new || + error "$DIR/$tdir/${tfile}.new: chmod has failed" + + # chown should still succeed + chown $RUNAS_ID:$RUNAS_GID $DIR/$tdir/${tfile}.new || + error "$DIR/$tdir/${tfile}.new: chown has failed" + + # rename should still succeed + mv $DIR/$tdir/${tfile}.new $DIR/$tdir/${tfile} || + error "$DIR/$tdir/${tfile}.new: rename has failed" + + #remove foreign_symlink file should still fail + rm $DIR/$tdir/${tfile} && + error "$DIR/$tdir/${tfile}: remove of foreign_symlink file should fail" + + #use special ioctl() to unlink foreign_symlink file + $LFS unlink_foreign $DIR/$tdir/${tfile} || + error "$DIR/$tdir/$tfile: unlink/ioctl failed" + +} +run_test 27O "basic ops on foreign file of symlink type" + +test_27P() { + [[ $(lustre_version_code $SINGLEMDS) -le $(version_code 2.12.49) ]] && + skip "Need MDS version newer than 2.12.49" + + test_mkdir $DIR/$tdir + local uuid1=$(cat /proc/sys/kernel/random/uuid) + local uuid2=$(cat /proc/sys/kernel/random/uuid) + + trap clean_foreign_symlink EXIT + + # enable foreign_symlink behaviour + $LCTL set_param llite/$FSNAME-*/foreign_symlink_enable=1 + + # foreign symlink LMV format is a partial path by default + + # create foreign dir (lfs + API) + $LFS mkdir --foreign=symlink --xattr="${uuid1}/${uuid2}" \ + --flags=0xda05 --mode 0750 $DIR/$tdir/${tdir} || + error "$DIR/$tdir/${tdir}: create failed" + + $LFS getdirstripe -v $DIR/$tdir/${tdir} | + grep "lfm_magic:.*0x0CD50CD0" || + error "$DIR/$tdir/${tdir}: invalid LMV EA magic" + $LFS getdirstripe -v $DIR/$tdir/${tdir} | grep "lfm_type:.*symlink" || + error "$DIR/$tdir/${tdir}: invalid LMV EA type" + $LFS getdirstripe -v $DIR/$tdir/${tdir} | + grep "lfm_flags:.*0x0000DA05" || + error "$DIR/$tdir/${tdir}: invalid LMV EA flags" + $LFS getdirstripe $DIR/$tdir/${tdir} | + grep "lfm_value.*${uuid1}/${uuid2}" || + error "$DIR/$tdir/${tdir}: invalid LMV EA value" + + # file create in dir should fail + # ("/{foreign_symlink_prefix}/${uuid1}/${uuid2}/" missing) + touch $DIR/$tdir/$tdir/$tfile && "$DIR/$tdir: file create should fail" + + # rename should succeed + mv $DIR/$tdir/$tdir $DIR/$tdir/${tdir}.new || + error "$DIR/$tdir/$tdir: rename of foreign_symlink dir has failed" + + #remove foreign_symlink dir should fail + rmdir $DIR/$tdir/${tdir}.new && + error "$DIR/$tdir/${tdir}.new: remove of foreign_symlink dir should fail" + + #test fake symlink + mkdir -p /tmp/${uuid1}/${uuid2} || + error "/tmp/${uuid1}/${uuid2}: mkdir has failed" + echo FOOFOO > /tmp/${uuid1}/${uuid2}/foo || + error "/tmp/${uuid1}/${uuid2}/foo: echo has failed" + $LCTL set_param llite/$FSNAME-*/foreign_symlink_prefix=/tmp/ + $CHECKSTAT -t link -l /tmp/${uuid1}/${uuid2} $DIR/$tdir/${tdir}.new || + error "$DIR/$tdir/${tdir}.new: not seen as a symlink" + cat $DIR/$tdir/${tdir}.new/foo | grep FOOFOO || + error "$DIR/$tdir/${tdir}.new: symlink resolution has failed" + + #check that getstripe fails now that foreign_symlink enabled + $LFS getdirstripe $DIR/$tdir/${tdir}.new || + error "$DIR/$tdir/${tdir}.new: getdirstripe should still work with foreign_symlink enabled" + + # file create in dir should work now + cp /etc/passwd $DIR/$tdir/${tdir}.new/$tfile || + error "$DIR/$tdir/${tdir}.new/$tfile: file create should fail" + diff /etc/passwd $DIR/$tdir/${tdir}.new/$tfile || + error "$DIR/$tdir/${tdir}.new/$tfile: diff has failed" + diff /etc/passwd /tmp/${uuid1}/${uuid2}/$tfile || + error "/tmp/${uuid1}/${uuid2}/$tfile: diff has failed" + + # chmod should still succeed + chmod 755 $DIR/$tdir/${tdir}.new || + error "$DIR/$tdir/${tdir}.new: chmod has failed" + + # chown should still succeed + chown $RUNAS_ID:$RUNAS_GID $DIR/$tdir/${tdir}.new || + error "$DIR/$tdir/${tdir}.new: chown has failed" + + # rename should still succeed + mv $DIR/$tdir/${tdir}.new $DIR/$tdir/${tdir} || + error "$DIR/$tdir/${tdir}.new: rename of foreign_symlink dir has failed" + + #remove foreign_symlink dir should still fail + rmdir $DIR/$tdir/${tdir} && + error "$DIR/$tdir/${tdir}: remove of foreign_symlink dir should fail" + + #use special ioctl() to unlink foreign_symlink file + $LFS unlink_foreign $DIR/$tdir/${tdir} || + error "$DIR/$tdir/$tdir: unlink/ioctl failed" + + #created file should still exist + [[ -f /tmp/${uuid1}/${uuid2}/$tfile ]] || + error "/tmp/${uuid1}/${uuid2}/$tfile has been removed" + diff /etc/passwd /tmp/${uuid1}/${uuid2}/$tfile || + error "/tmp/${uuid1}/${uuid2}/$tfile: diff has failed" +} +run_test 27P "basic ops on foreign dir of foreign_symlink type" + # createtest also checks that device nodes are created and # then visible correctly (#2091) test_28() { # bug 2091 @@ -3326,6 +3630,21 @@ test_31p() { } run_test 31p "remove of open striped directory" +test_31q() { + [ $MDSCOUNT -lt 3 ] && skip_env "needs >= 3 MDTs" + + $LFS mkdir -i 3,1 $DIR/$tdir || error "mkdir failed" + index=$($LFS getdirstripe -i $DIR/$tdir) + [ $index -eq 3 ] || error "first stripe index $index != 3" + index=$($LFS getdirstripe $DIR/$tdir | tail -1 | awk '{print $1}') + [ $index -eq 1 ] || error "second stripe index $index != 1" + + # when "-c " is set, the number of MDTs specified after + # "-i" should equal to the stripe count + $LFS mkdir -i 3,1 -c 3 $DIR/$tdir.2 && error "mkdir should fail" || true +} +run_test 31q "create striped directory on specific MDTs" + cleanup_test32_mount() { local rc=0 trap 0 @@ -3827,7 +4146,7 @@ test_33h() { done local failed=0 - for i in {1..50}; do + for i in {1..250}; do for fname in $(mktemp -u $DIR/$tdir/.$tfile.XXXXXX) \ $(mktemp $DIR/$tdir/$tfile.XXXXXXXX); do touch $fname || error "touch $fname failed" @@ -3839,7 +4158,7 @@ test_33h() { done done echo "$failed MDT index mismatches" - (( failed < 4 )) || error "MDT index mismatch $failed times" + (( failed < 20 )) || error "MDT index mismatch $failed times" } run_test 33h "temp file is located on the same MDT as target" @@ -4585,7 +4904,8 @@ test_39r() { # exceed atime_diff and access file sleep 6 - dd if=$DIR/$tfile of=/dev/null || error "can't udpate atime" + dd if=$DIR/$tfile of=/dev/null bs=4k count=1 || + error "can't udpate atime" local atime_cli=$(stat -c %X $DIR/$tfile) echo "client atime: $atime_cli" @@ -4873,6 +5193,8 @@ test_43A() { # was test_43 $DIR/$tdir/$tfile && error "execute $DIR/$tdir/$tfile succeeded" || true kill -USR1 $pid + # Wait for multiop to exit + wait $pid } run_test 43A "execution of file opened for write should return -ETXTBSY" @@ -4920,17 +5242,14 @@ test_44A() { # was test_44 run_test 44A "zero length read from a sparse stripe" test_44a() { - local nstripe=$($LCTL lov_getconfig $DIR | grep default_stripe_count: | - awk '{ print $2 }') + local nstripe=$($LFS getstripe -c -d $DIR) [ -z "$nstripe" ] && skip "can't get stripe info" [[ $nstripe -gt $OSTCOUNT ]] && - skip "Wrong default_stripe_count: $nstripe OSTCOUNT: $OSTCOUNT" + skip "Wrong default stripe_count: $nstripe OSTCOUNT: $OSTCOUNT" - local stride=$($LCTL lov_getconfig $DIR | grep default_stripe_size: | - awk '{ print $2 }') + local stride=$($LFS getstripe -S -d $DIR) if [[ $nstripe -eq 0 || $nstripe -eq -1 ]]; then - nstripe=$($LCTL lov_getconfig $DIR | grep obd_count: | - awk '{ print $2 }') + nstripe=$($LFS df $DIR | grep OST: | wc -l) fi OFFSETS="0 $((stride/2)) $((stride-1))" @@ -5122,6 +5441,27 @@ test_48e() { # bug 4134 } run_test 48e "Access to recreated parent subdir (should return errors)" +test_48f() { + [[ $MDS1_VERSION -ge $(version_code 2.13.55) ]] || + skip "need MDS >= 2.13.55" + [[ $MDSCOUNT -ge 2 ]] || skip "needs >= 2 MDTs" + [[ "$(facet_host mds1)" != "$(facet_host mds2)" ]] || + skip "needs different host for mdt1 mdt2" + [[ $(facet_fstype mds1) == ldiskfs ]] || skip "ldiskfs only" + + $LFS mkdir -i0 $DIR/$tdir + $LFS mkdir -i 1 $DIR/$tdir/sub1 $DIR/$tdir/sub2 $DIR/$tdir/sub3 + + for d in sub1 sub2 sub3; do + #define OBD_FAIL_OSD_REF_DEL 0x19c + do_facet mds1 $LCTL set_param fail_loc=0x8000019c + rm -rf $DIR/$tdir/$d && error "rm $d should fail" + done + + rm -d --interactive=never $DIR/$tdir || error "rm $tdir fail" +} +run_test 48f "non-zero nlink dir unlink won't LBUG()" + test_49() { # LU-1030 [ $PARALLEL == "yes" ] && skip "skip parallel run" remote_ost_nodsh && skip "remote OST with nodsh" @@ -5912,7 +6252,7 @@ test_56ob() { cmd="$LFS find $dir -ctime +1s -type f" nums=$($cmd | wc -l) (( $nums == $count * 2 + 1)) || - error "'$cmd' wrong: found $nums, expected $((expected*2+1))" + error "'$cmd' wrong: found $nums, expected $((count * 2 + 1))" } run_test 56ob "check lfs find -atime -mtime -ctime with units" @@ -5933,6 +6273,8 @@ test_newerXY_base() { ref=$DIR/$tfile.newer.$x$y touch $ref || error "touch $ref failed" fi + + echo "before = $ref" sleep 2 setup_56 $dir $NUMFILES $NUMDIRS "-i0 -c1" "-i0 -c1" sleep 2 @@ -5947,28 +6289,28 @@ test_newerXY_base() { touch $negref || error "touch $negref failed" fi + echo "after = $negref" local cmd="$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" + [ $nums -eq $expected ] || { ls -lauR --full-time $dir ; + error "'$cmd' wrong: found $nums newer, 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" + [ $nums -eq $expected ] || { ls -lauR --full-time $dir ; + error "'$cmd' wrong: found $nums older, 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" + [ $nums -eq $expected ] || { ls -lauR --full-time $dir ; + error "'$cmd' wrong: found $nums between, expected $expected"; } rm -rf $DIR/* } test_56oc() { - test_newerXY_base "b" "t" test_newerXY_base "a" "a" test_newerXY_base "a" "m" test_newerXY_base "a" "c" @@ -5978,10 +6320,19 @@ test_56oc() { test_newerXY_base "c" "a" test_newerXY_base "c" "m" test_newerXY_base "c" "c" - test_newerXY_base "b" "b" + + [[ -n "$sles_version" ]] && + echo "skip timestamp tests on SLES, LU-13665" && return 0 + test_newerXY_base "a" "t" test_newerXY_base "m" "t" test_newerXY_base "c" "t" + + [[ $MDS1_VERSION -lt $(version_code 2.13.54) || + $CLIENT_VERSION -lt $(version_code 2.13.54) ]] && + ! btime_supported && echo "btime unsupported" && return 0 + + test_newerXY_base "b" "b" test_newerXY_base "b" "t" } run_test 56oc "check lfs find -newerXY work" @@ -6227,8 +6578,13 @@ test_56ra() { [[ $MDS1_VERSION -gt $(version_code 2.12.58) ]] || skip "MDS < 2.12.58 doesn't return LSOM data" local dir=$DIR/$tdir + local old_agl=$($LCTL get_param -n llite.*.statahead_agl) - [[ $OSC == "mdc" ]] && skip "DoM files" && return + [[ $OSC == "mdc" ]] && skip "statahead not needed for DoM files" + + # statahead_agl may cause extra glimpse which confuses results. LU-13017 + $LCTL set_param -n llite.*.statahead_agl=0 + stack_trap "$LCTL set_param -n llite.*.statahead_agl=$old_agl" setup_56 $dir $NUMFILES $NUMDIRS "-c 1" # open and close all files to ensure LSOM is updated @@ -6268,14 +6624,15 @@ test_56rb() { test_mkdir -p $dir || error "failed to mkdir $dir" $LFS setstripe -c 1 -i 0 $dir/$tfile || error "failed to setstripe $dir/$tfile" + mdt_idx=$($LFS getdirstripe -i $dir) dd if=/dev/zero of=$dir/$tfile bs=1M count=1 stack_trap "rm -f $tmp" EXIT - $LFS find --size +100K --ost 0 $dir 2>&1 | tee $tmp - [ -z "$(cat $tmp | grep "obd_uuid: ")" ] || + $LFS find --size +100K --ost 0 $dir |& tee $tmp + ! grep -q obd_uuid $tmp || error "failed to find --size +100K --ost 0 $dir" - $LFS find --size +100K --mdt $mdt_idx $dir 2>&1 | tee $tmp - [ -z "$(cat $tmp | grep "obd_uuid: ")" ] || + $LFS find --size +100K --mdt $mdt_idx $dir |& tee $tmp + ! grep -q obd_uuid $tmp || error "failed to find --size +100K --mdt $mdt_idx $dir" } run_test 56rb "check lfs find --size --ost/--mdt works" @@ -6992,7 +7349,7 @@ test_56xc() { error "cannot setstripe 20MB file" echo "done" echo -n "Sizing 20MB test file..." - truncate "$dir/20mb" 20971520 || error "cannot create 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" || @@ -7012,7 +7369,7 @@ test_56xc() { echo "done" echo -n "Sizing 1GB test file..." # File size is 1GB + 3KB - truncate "$dir/1gb" 1073744896 || error "cannot create 1GB test file" + $TRUNCATE "$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 @@ -7399,6 +7756,38 @@ test_56ca() { } run_test 56ca "check lfs find --mirror-count|-N and --mirror-state" +test_56da() { # LU-14179 + local path=$DIR/$tdir + + test_mkdir $path + cd $path + + local longdir=$(str_repeat 'a' 255) + + for i in {1..15}; do + path=$path/$longdir + test_mkdir $longdir + cd $longdir + done + + local len=${#path} + local lastdir=$(str_repeat 'a' $((4096 - 1 - $len - 1))) + + test_mkdir $lastdir + cd $lastdir + # PATH_MAX-1 + (( ${#PWD} == 4095 )) || error "bad PWD length ${#PWD}, expect 4095" + + # NAME_MAX + touch $(str_repeat 'f' 255) + + $LFS find $DIR/$tdir --type d |& grep "lfs find: error" && + error "lfs find reported an error" + + rm -rf $DIR/$tdir +} +run_test 56da "test lfs find with long paths" + test_57a() { [ $PARALLEL == "yes" ] && skip "skip parallel run" # note test will not do anything if MDS is not local @@ -7660,6 +8049,25 @@ test_60e() { } run_test 60e "no space while new llog is being created" +test_60f() { + local old_path=$($LCTL get_param -n debug_path) + + stack_trap "$LCTL set_param debug_path=$old_path" + stack_trap "rm -f $TMP/$tfile*" + #define OBD_FAIL_PTLRPC_DUMP_LOG 0x50e + $LCTL set_param debug_path=$TMP/$tfile fail_loc=0x8000050e + test_mkdir $DIR/$tdir + rm -f $TMP/$tfile* 2> /dev/null + # retry in case the open is cached and not released + for (( i = 0; i < 100 && $(ls $TMP/$tfile* | wc -l) == 0; i++ )); do + echo $i > $DIR/$tdir/$tfile.$i && cat $DIR/$tdir/$tfile.$i + sleep 0.1 + done + ls $TMP/$tfile* + (( $(ls $TMP/$tfile* | wc -l) > 0 )) || error "$TMP/$tfile not dumped" +} +run_test 60f "change debug_path works" + test_60g() { local pid local i @@ -7686,7 +8094,7 @@ test_60g() { do_facet mds$index $LCTL set_param fail_loc=0x8000019a \ > /dev/null - usleep 100 + sleep 0.01 done kill -9 $pid @@ -7839,18 +8247,25 @@ test_64c() { } run_test 64c "verify grant shrink" +import_param() { + local tgt=$1 + local param=$2 + + $LCTL get_param osc.$tgt.import | awk "/$param/ { print \$2 }" +} + # this does exactly what osc_request.c:osc_announce_cached() does in # order to calculate max amount of grants to ask from server want_grant() { local tgt=$1 - local nrpages=$($LCTL get_param -n osc.${tgt}.max_pages_per_rpc) - local rpc_in_flight=$($LCTL get_param -n osc.${tgt}.max_rpcs_in_flight) + local nrpages=$($LCTL get_param -n osc.$tgt.max_pages_per_rpc) + local rpc_in_flight=$($LCTL get_param -n osc.$tgt.max_rpcs_in_flight) - ((rpc_in_flight ++)); + ((rpc_in_flight++)); nrpages=$((nrpages * rpc_in_flight)) - local dirty_max_pages=$($LCTL get_param -n osc.${tgt}.max_dirty_mb) + local dirty_max_pages=$($LCTL get_param -n osc.$tgt.max_dirty_mb) dirty_max_pages=$((dirty_max_pages * 1024 * 1024 / PAGE_SIZE)) @@ -7858,13 +8273,11 @@ want_grant() { local undirty=$((nrpages * PAGE_SIZE)) local max_extent_pages - max_extent_pages=$($LCTL get_param osc.${tgt}.import | - grep grant_max_extent_size | awk '{print $2}') + max_extent_pages=$(import_param $tgt grant_max_extent_size) max_extent_pages=$((max_extent_pages / PAGE_SIZE)) local nrextents=$(((nrpages + max_extent_pages - 1) / max_extent_pages)) local grant_extent_tax - grant_extent_tax=$($LCTL get_param osc.${tgt}.import | - grep grant_extent_tax | awk '{print $2}') + grant_extent_tax=$(import_param $tgt grant_extent_tax) undirty=$((undirty + nrextents * grant_extent_tax)) @@ -7878,56 +8291,171 @@ grant_chunk() { local max_brw_size local grant_extent_tax - max_brw_size=$($LCTL get_param osc.${tgt}.import | - grep max_brw_size | awk '{print $2}') + max_brw_size=$(import_param $tgt max_brw_size) - grant_extent_tax=$($LCTL get_param osc.${tgt}.import | - grep grant_extent_tax | awk '{print $2}') + grant_extent_tax=$(import_param $tgt grant_extent_tax) echo $(((max_brw_size + grant_extent_tax) * 2)) } test_64d() { - [ $OST1_VERSION -lt $(version_code 2.10.56) ] && + [ $OST1_VERSION -ge $(version_code 2.10.56) ] || skip "OST < 2.10.55 doesn't limit grants enough" - local tgt=$($LCTL dl | grep "0000-osc-[^mM]" | awk '{print $4}') - local file=$DIR/$tfile + local tgt=$($LCTL dl | awk '/OST0000-osc-[^mM]/ { print $4 }') - [[ $($LCTL get_param osc.${tgt}.import | - grep "connect_flags:.*grant_param") ]] || + [[ "$($LCTL get_param osc.${tgt}.import)" =~ "grant_param" ]] || skip "no grant_param connect flag" - local olddebug=$($LCTL get_param -n debug 2> /dev/null) + local olddebug="$($LCTL get_param -n debug 2> /dev/null)" + + $LCTL set_param -n -n debug="$OLDDEBUG" || true + stack_trap "$LCTL set_param -n debug='$olddebug'" EXIT - $LCTL set_param debug="$OLDDEBUG" 2> /dev/null || true local max_cur_granted=$(($(want_grant $tgt) + $(grant_chunk $tgt))) - stack_trap "rm -f $file" EXIT + stack_trap "rm -f $DIR/$tfile && wait_delete_completed" EXIT - $LFS setstripe $file -i 0 -c 1 - dd if=/dev/zero of=$file bs=1M count=1000 & + $LFS setstripe $DIR/$tfile -i 0 -c 1 + dd if=/dev/zero of=$DIR/$tfile bs=1M count=1000 & ddpid=$! - while true - do - local cur_grant=$($LCTL get_param -n osc.${tgt}.cur_grant_bytes) - if [[ $cur_grant -gt $max_cur_granted ]] - then + while kill -0 $ddpid; do + local cur_grant=$($LCTL get_param -n osc.$tgt.cur_grant_bytes) + + if [[ $cur_grant -gt $max_cur_granted ]]; then kill $ddpid error "cur_grant $cur_grant > $max_cur_granted" fi - kill -0 $ddpid - [[ $? -ne 0 ]] && break; - sleep 2 - done - rm -f $DIR/$tfile - wait_delete_completed - $LCTL set_param debug="$olddebug" 2> /dev/null || true + sleep 1 + done } run_test 64d "check grant limit exceed" +check_grants() { + local tgt=$1 + local expected=$2 + local msg=$3 + local cur_grants=$($LCTL get_param -n osc.$tgt.cur_grant_bytes) + + ((cur_grants == expected)) || + error "$msg: grants mismatch: $cur_grants, expected $expected" +} + +round_up_p2() { + echo $((($1 + $2 - 1) & ~($2 - 1))) +} + +test_64e() { + [ $PARALLEL == "yes" ] && skip "skip parallel run" + [ $OST1_VERSION -ge $(version_code 2.11.56) ] || + skip "Need OSS version at least 2.11.56" + + # Remount client to reset grant + remount_client $MOUNT || error "failed to remount client" + local osc_tgt="$FSNAME-OST0000-osc-$($LFS getname -i $DIR)" + + local init_grants=$(import_param $osc_tgt initial_grant) + + check_grants $osc_tgt $init_grants "init grants" + + local extent_tax=$(import_param $osc_tgt grant_extent_tax) + local max_brw_size=$(import_param $osc_tgt max_brw_size) + local gbs=$(import_param $osc_tgt grant_block_size) + + # write random number of bytes from max_brw_size / 4 to max_brw_size + local write_bytes=$(shuf -i $((max_brw_size / 4))-$max_brw_size -n 1) + # align for direct io + write_bytes=$(round_up_p2 $write_bytes PAGE_SIZE) + # round to grant consumption unit + local wb_round_up=$(round_up_p2 $write_bytes gbs) + + local grants=$((wb_round_up + extent_tax)) + + $LFS setstripe -c 1 -i 0 $DIR/$tfile || error "lfs setstripe failed" + + # define OBD_FAIL_TGT_NO_GRANT 0x725 + # make the server not grant more back + do_facet ost1 $LCTL set_param fail_loc=0x725 + dd if=/dev/zero of=$DIR/$tfile bs=$write_bytes count=1 oflag=direct + + do_facet ost1 $LCTL set_param fail_loc=0 + + check_grants $osc_tgt $((init_grants - grants)) "dio w/o grant alloc" + + rm -f $DIR/$tfile || error "rm failed" + + # Remount client to reset grant + remount_client $MOUNT || error "failed to remount client" + osc_tgt="$FSNAME-OST0000-osc-$($LFS getname -i $DIR)" + + $LFS setstripe -c 1 -i 0 $DIR/$tfile || error "lfs setstripe failed" + + # define OBD_FAIL_TGT_NO_GRANT 0x725 + # make the server not grant more back + do_facet ost1 $LCTL set_param fail_loc=0x725 + $MULTIOP $DIR/$tfile "oO_WRONLY:w${write_bytes}yc" + do_facet ost1 $LCTL set_param fail_loc=0 + + check_grants $osc_tgt $((init_grants - grants)) "buf io w/o grant alloc" +} +run_test 64e "check grant consumption (no grant allocation)" + +test_64f() { + [ $PARALLEL == "yes" ] && skip "skip parallel run" + + # Remount client to reset grant + remount_client $MOUNT || error "failed to remount client" + local osc_tgt="$FSNAME-OST0000-osc-$($LFS getname -i $DIR)" + + local init_grants=$(import_param $osc_tgt initial_grant) + local extent_tax=$(import_param $osc_tgt grant_extent_tax) + local max_brw_size=$(import_param $osc_tgt max_brw_size) + local gbs=$(import_param $osc_tgt grant_block_size) + local chunk=$(grant_chunk $osc_tgt) + + # write random number of bytes from max_brw_size / 4 to max_brw_size + local write_bytes=$(shuf -i $((max_brw_size / 4))-$max_brw_size -n 1) + # align for direct io + write_bytes=$(round_up_p2 $write_bytes PAGE_SIZE) + # round to grant consumption unit + local wb_round_up=$(round_up_p2 $write_bytes gbs) + + local grants=$((wb_round_up + extent_tax)) + + $LFS setstripe -c 1 -i 0 $DIR/$tfile || error "lfs setstripe failed" + dd if=/dev/zero of=$DIR/$tfile bs=$write_bytes count=1 oflag=direct || + error "error writing to $DIR/$tfile" + + check_grants $osc_tgt $((init_grants - grants + chunk)) \ + "direct io with grant allocation" + + rm -f $DIR/$tfile || error "rm failed" + + # Remount client to reset grant + remount_client $MOUNT || error "failed to remount client" + osc_tgt="$FSNAME-OST0000-osc-$($LFS getname -i $DIR)" + + $LFS setstripe -c 1 -i 0 $DIR/$tfile || error "lfs setstripe failed" + + local cmd="oO_WRONLY:w${write_bytes}_yc" + + $MULTIOP $DIR/$tfile $cmd & + MULTIPID=$! + sleep 1 + + check_grants $osc_tgt $((init_grants - grants)) \ + "buffered io, not write rpc" + + kill -USR1 $MULTIPID + wait + + check_grants $osc_tgt $((init_grants - grants + chunk)) \ + "buffered io, one RPC" +} +run_test 64f "check grant consumption (with grant allocation)" + # bug 1414 - set/get directories' stripe info test_65a() { [ $PARALLEL == "yes" ] && skip "skip parallel run" @@ -8208,7 +8736,11 @@ test_65n() { local file2_stripe_size=$($LFS getstripe -S $file2) [[ $file2_stripe_size -eq $new_def_stripe_size ]] || + { + echo "file2_stripe_size: '$file2_stripe_size'" + echo "new_def_stripe_size: '$new_def_stripe_size'" error "$file2 didn't inherit stripe size $new_def_stripe_size" + } local dir3=$MOUNT/$tdir-3 mkdir $dir3 || error "mkdir $dir3 failed" @@ -8218,7 +8750,11 @@ test_65n() { local dir3_layout=$(get_layout_param $dir3) local root_dir_layout=$(get_layout_param $MOUNT) [[ "$dir3_layout" = "$root_dir_layout" ]] || + { + echo "dir3_layout: '$dir3_layout'" + echo "root_dir_layout: '$root_dir_layout'" error "$dir3 should show the default layout from $MOUNT" + } # set OST pool on root directory local pool=$TESTNAME @@ -8236,7 +8772,7 @@ test_65n() { local file3_pool=$($LFS getstripe -p $file3) [[ "$file3_pool" = "$pool" ]] || - error "$file3 didn't inherit OST pool $pool" + error "$file3 ('$file3_pool') didn't inherit OST pool '$pool'" local dir4=$MOUNT/$tdir-4 mkdir $dir4 || error "mkdir $dir4 failed" @@ -8247,7 +8783,11 @@ test_65n() { echo "$LFS getstripe -d $MOUNT" $LFS getstripe -d $MOUNT [[ "$dir4_layout" = "$root_dir_layout" ]] || + { + echo "dir4_layout: '$dir4_layout'" + echo "root_dir_layout: '$root_dir_layout'" error "$dir4 should show the default layout from $MOUNT" + } # new file created in $dir4 should inherit the pool from # the filesystem default @@ -8256,7 +8796,7 @@ test_65n() { local file4_pool=$($LFS getstripe -p $file4) [[ "$file4_pool" = "$pool" ]] || - error "$file4 didn't inherit OST pool $pool" + error "$file4 ('$file4_pool') didn't inherit OST pool $pool" # new subdirectory under non-root directory should inherit # the default layout from its parent directory @@ -8269,7 +8809,11 @@ test_65n() { dir4_layout=$(get_layout_param $dir4) local dir5_layout=$(get_layout_param $dir5) [[ "$dir4_layout" = "$dir5_layout" ]] || + { + echo "dir4_layout: '$dir4_layout'" + echo "dir5_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. @@ -8479,39 +9023,78 @@ test_74c() { } run_test 74c "ldlm_lock_create error path, (shouldn't LBUG)" -num_inodes() { - awk '/lustre_inode_cache/ {print $2; exit}' /proc/slabinfo +slab_lic=/sys/kernel/slab/lustre_inode_cache +num_objects() { + [ -f $slab_lic/shrink ] && echo 1 > $slab_lic/shrink + [ -f $slab_lic/objects ] && awk '{ print $1 }' $slab_lic/objects || + awk '/lustre_inode_cache/ { print $2; exit }' /proc/slabinfo } -test_76() { # Now for bug 20433, added originally in bug 1443 +test_76a() { # Now for b=20433, added originally in b=1443 [ $PARALLEL == "yes" ] && skip "skip parallel run" cancel_lru_locks osc + # there may be some slab objects cached per core local cpus=$(getconf _NPROCESSORS_ONLN 2>/dev/null) - local before=$(num_inodes) + local before=$(num_objects) local count=$((512 * cpus)) - [ "$SLOW" = "no" ] && count=$((64 * cpus)) + [ "$SLOW" = "no" ] && count=$((128 * cpus)) + local margin=$((count / 10)) + if [[ -f $slab_lic/aliases ]]; then + local aliases=$(cat $slab_lic/aliases) + (( aliases > 0 )) && margin=$((margin * aliases)) + fi - echo "before inodes: $before" + echo "before slab objects: $before" for i in $(seq $count); do touch $DIR/$tfile rm -f $DIR/$tfile done cancel_lru_locks osc - local after=$(num_inodes) - echo "after inodes: $after" - while (( after > before + 8 * ${cpus:-1} )); do + local after=$(num_objects) + echo "created: $count, after slab objects: $after" + # shared slab counts are not very accurate, allow significant margin + # the main goal is that the cache growth is not permanently > $count + while (( after > before + margin )); do sleep 1 - after=$(num_inodes) + after=$(num_objects) wait=$((wait + 1)) - (( wait % 5 == 0 )) && echo "wait $wait seconds inodes: $after" - if (( wait > 30 )); then - error "inode slab grew from $before to $after" + (( wait % 5 == 0 )) && echo "wait $wait seconds objects: $after" + if (( wait > 60 )); then + error "inode slab grew from $before+$margin to $after" fi done } -run_test 76 "confirm clients recycle inodes properly ====" +run_test 76a "confirm clients recycle inodes properly ====" + +test_76b() { + [ $PARALLEL == "yes" ] && skip "skip parallel run" + [ $CLIENT_VERSION -ge $(version_code 2.13.55) ] || skip "not supported" + + local count=512 + local before=$(num_objects) + + for i in $(seq $count); do + mkdir $DIR/$tdir + rmdir $DIR/$tdir + done + + local after=$(num_objects) + local wait=0 + while (( after > before )); do + sleep 1 + after=$(num_objects) + wait=$((wait + 1)) + (( wait % 5 == 0 )) && echo "wait $wait seconds objects: $after" + if (( wait > 60 )); then + error "inode slab grew from $before to $after" + fi + done + + echo "slab objects before: $before, after: $after" +} +run_test 76b "confirm clients recycle directory inodes properly ====" export ORIG_CSUM="" set_checksums() @@ -8769,9 +9352,8 @@ test_77k() { # LU-10906 local i [ "$ORIG_CSUM" ] || ORIG_CSUM=$(eval $get_checksum) - stack_trap "wait_update $HOSTNAME '$get_checksum' $ORIG_CSUM" EXIT - stack_trap "do_facet mgs $LCTL set_param -P $cksum_param=$ORIG_CSUM" \ - EXIT + stack_trap "wait_update $HOSTNAME '$get_checksum' $ORIG_CSUM || true" + stack_trap "do_facet mgs $LCTL set_param -P $cksum_param=$ORIG_CSUM" for i in 0 1; do do_facet mgs $LCTL set_param -P $cksum_param=$i || @@ -8994,28 +9576,6 @@ test_81b() { # LU-456 } run_test 81b "OST should return -ENOSPC when retry still fails =======" -test_82() { # LU-1031 - dd if=/dev/zero of=$DIR/$tfile bs=1M count=10 - local gid1=14091995 - local gid2=16022000 - - multiop_bg_pause $DIR/$tfile OG${gid1}_g${gid1}c || return 1 - local MULTIPID1=$! - multiop_bg_pause $DIR/$tfile O_G${gid2}r10g${gid2}c || return 2 - local MULTIPID2=$! - kill -USR1 $MULTIPID2 - sleep 2 - if [[ `ps h -o comm -p $MULTIPID2` == "" ]]; then - error "First grouplock does not block second one" - else - echo "Second grouplock blocks first one" - fi - kill -USR1 $MULTIPID1 - wait $MULTIPID1 - wait $MULTIPID2 -} -run_test 82 "Basic grouplock test" - test_99() { [ -z "$(which cvs 2>/dev/null)" ] && skip_env "could not find cvs" @@ -9412,12 +9972,11 @@ test_101g_brw_size_test() { sed -n '/pages per rpc/,/^$/p' | awk '/'$pages':/ { reads += $2; writes += $6 }; \ END { print reads,writes }')) - [ ${rpcs[0]} -ne $count ] && error "${rpcs[0]} != $count read RPCs" && - return 5 - [ ${rpcs[1]} -ne $count ] && error "${rpcs[1]} != $count write RPCs" && - return 6 - - return 0 + # allow one extra full-sized read RPC for async readahead + [[ ${rpcs[0]} == $count || ${rpcs[0]} == $((count + 1)) ]] || + { error "${rpcs[0]} != $count read RPCs"; return 5; } + [[ ${rpcs[1]} == $count ]] || + { error "${rpcs[1]} != $count write RPCs"; return 6; } } test_101g() { @@ -10262,6 +10821,21 @@ test_103c() { } run_test 103c "'cp -rp' won't set empty acl" +test_103e() { + (( $MDS1_VERSION >= $(version_code 2.13.59) )) || + skip "MDS needs to be at least 2.13.59" + + mkdir -p $DIR/$tdir + # one default ACL will be created for the file owner + for U in {2..256}; do + setfacl -m default:user:$U:rwx $DIR/$tdir + numacl=$(getfacl $DIR/$tdir |& grep -c "default:user") + touch $DIR/$tdir/$tfile.$U || + error "failed to create $tfile.$U with $numacl ACLs" + done +} +run_test 103e "inheritance of big amount of default ACLs" + test_104a() { [ $PARALLEL == "yes" ] && skip "skip parallel run" @@ -11675,18 +12249,18 @@ test_124a() { skip "Limit is too small $LIMIT" fi - # Make LVF so higher that sleeping for $SLEEP is enough to _start_ - # killing locks. Some time was spent for creating locks. This means - # that up to the moment of sleep finish we must have killed some of - # them (10-100 locks). This depends on how fast ther were created. - # Many of them were touched in almost the same moment and thus will - # be killed in groups. - local LVF=$(($MAX_HRS * 60 * 60 / $SLEEP * $LIMIT / $LRU_SIZE)) - - # Use $LRU_SIZE_B here to take into account real number of locks - # created in the case of CMD, LRU_SIZE_B != $NR in most of cases - local LRU_SIZE_B=$LRU_SIZE - log "LVF=$LVF" + # Make LVF so higher that sleeping for $SLEEP is enough to _start_ + # killing locks. Some time was spent for creating locks. This means + # that up to the moment of sleep finish we must have killed some of + # them (10-100 locks). This depends on how fast ther were created. + # Many of them were touched in almost the same moment and thus will + # be killed in groups. + local LVF=$(($MAX_HRS * 60 * 60 / $SLEEP * $LIMIT / $LRU_SIZE * 100)) + + # Use $LRU_SIZE_B here to take into account real number of locks + # created in the case of CMD, LRU_SIZE_B != $NR in most of cases + local LRU_SIZE_B=$LRU_SIZE + log "LVF=$LVF" local OLD_LVF=$($LCTL get_param -n $NSDIR.pool.lock_volume_factor) log "OLD_LVF=$OLD_LVF" $LCTL set_param -n $NSDIR.pool.lock_volume_factor $LVF @@ -12421,38 +12995,30 @@ test_130e() { local fm_file=$DIR/$tfile $LFS setstripe -S 131072 -c 2 $fm_file || error "setstripe on $fm_file" - [ "$(facet_fstype ost$(($($LFS getstripe -i $fm_file) + 1)))" = "zfs" ] && - skip_env "ORI-366/LU-1941: FIEMAP unimplemented on ZFS" NUM_BLKS=512 EXPECTED_LEN=$(( (NUM_BLKS / 2) * 64 )) - for ((i = 0; i < $NUM_BLKS; i++)) - do - dd if=/dev/zero of=$fm_file count=1 bs=64k seek=$((2*$i)) conv=notrunc > /dev/null 2>&1 + for ((i = 0; i < $NUM_BLKS; i++)); do + dd if=/dev/zero of=$fm_file count=1 bs=64k seek=$((2*$i)) \ + conv=notrunc > /dev/null 2>&1 done filefrag -ves $fm_file || error "filefrag $fm_file failed" filefrag_op=$(filefrag -ve -k $fm_file | sed -n '/ext:/,/found/{/ext:/d; /found/d; p}') - last_lun=$(echo $filefrag_op | cut -d: -f5 | - sed -e 's/^[ \t]*/0x/' | sed -e 's/0x0x/0x/') + last_lun=$(echo $filefrag_op | cut -d: -f5) IFS=$'\n' tot_len=0 num_luns=1 - for line in $filefrag_op - do - frag_lun=$(echo $line | cut -d: -f5 | - sed -e 's/^[ \t]*/0x/' | sed -e 's/0x0x/0x/') + for line in $filefrag_op; do + frag_lun=$(echo $line | cut -d: -f5) ext_len=$(echo $line | cut -d: -f4) - if (( $frag_lun != $last_lun )); then + if [[ "$frag_lun" != "$last_lun" ]]; then if (( tot_len != $EXPECTED_LEN )); then cleanup_130 - error "FIEMAP on $fm_file failed; returned " \ - "len $tot_len for OST $last_lun instead " \ - "of $EXPECTED_LEN" - return + error "OST$last_lun $tot_len != $EXPECTED_LEN" else (( num_luns += 1 )) tot_len=0 @@ -12463,13 +13029,9 @@ test_130e() { done if (( num_luns != 2 || tot_len != $EXPECTED_LEN )); then cleanup_130 - error "FIEMAP on $fm_file failed; returned wrong number " \ - "of luns or wrong len for OST $last_lun" - return + error "OST$last_lun $num_luns != 2, $tot_len != $EXPECTED_LEN" fi - cleanup_130 - echo "FIEMAP with continuation calls succeeded" } run_test 130e "FIEMAP (test continuation FIEMAP calls)" @@ -12486,14 +13048,38 @@ test_130f() { filefrag_extents=$(filefrag -vek $fm_file | awk '/extents? found/ { print $2 }') if [[ "$filefrag_extents" != "0" ]]; then - error "FIEMAP on $fm_file failed; " \ - "returned $filefrag_extents expected 0" + error "$fm_file: filefrag_extents=$filefrag_extents != 0" fi rm -f $fm_file } run_test 130f "FIEMAP (unstriped file)" +test_130g() { + local file=$DIR/$tfile + local nr=$((OSTCOUNT * 100)) + + $LFS setstripe -C $nr $file || + error "failed to setstripe -C $nr $file" + + dd if=/dev/zero of=$file count=$nr bs=1M + sync + nr=$($LFS getstripe -c $file) + + local extents=$(filefrag -v $file | + sed -n '/ext:/,/found/{/ext:/d; /found/d; p}' | wc -l) + + echo "filefrag list $extents extents in file with stripecount $nr" + if (( extents < nr )); then + $LFS getstripe $file + filefrag -v $file + error "filefrag printed $extents < $nr extents" + fi + + rm -f $file +} +run_test 130g "FIEMAP (overstripe file)" + # Test for writev/readv test_131a() { rwv -f $DIR/$tfile -w -n 3 524288 1048576 1572864 || @@ -13156,6 +13742,7 @@ test_150a() { local TF="$TMP/$tfile" + stack_trap "rm -f $DIR/$tfile; wait_delete_completed" dd if=/dev/urandom of=$TF bs=6096 count=1 || error "dd failed" cp $TF $DIR/$tfile cancel_lru_locks $OSC @@ -13178,58 +13765,87 @@ test_150a() { echo "12345" >>$DIR/$tfile cancel_lru_locks $OSC cmp $TF $DIR/$tfile || error "$TF $DIR/$tfile differ (append2)" - - rm -f $TF - true } run_test 150a "truncate/append tests" test_150b() { - [ "$ost1_FSTYPE" != ldiskfs ] && skip "non-ldiskfs backend" - [ $OST1_VERSION -lt $(version_code 2.13.50) ] && - skip "Need OST version at least 2.13.53" + check_set_fallocate_or_skip + touch $DIR/$tfile + stack_trap "rm -f $DIR/$tfile; wait_delete_completed" check_fallocate $DIR/$tfile || error "fallocate failed" } run_test 150b "Verify fallocate (prealloc) functionality" -test_150c() { - local bytes - local want +test_150bb() { + check_set_fallocate_or_skip - [ "$ost1_FSTYPE" != ldiskfs ] && skip "non-ldiskfs backend" - [ $OST1_VERSION -lt $(version_code 2.13.50) ] && - skip "Need OST version at least 2.13.53" + touch $DIR/$tfile + stack_trap "rm -f $DIR/$tfile; wait_delete_completed" + dd if=/dev/urandom of=$DIR/$tfile bs=1M count=20 || error "dd failed" + > $DIR/$tfile + fallocate -l $((1048576 * 20)) $DIR/$tfile || error "fallocate failed" + # precomputed md5sum for 20MB of zeroes + local expect="8f4e33f3dc3e414ff94e5fb6905cba8c" + local sum=($(md5sum $DIR/$tfile)) - $LFS setstripe -c $OSTCOUNT -S1M $DIR/$tdir || error "setstripe failed" - fallocate -l ${OSTCOUNT}m $DIR/$tdir || error "fallocate failed" + [[ "${sum[0]}" == "$expect" ]] || error "fallocate unwritten is not zero" + + check_set_fallocate 1 + + > $DIR/$tfile + fallocate -l $((1048576 * 20)) $DIR/$tfile || error "fallocate failed" + sum=($(md5sum $DIR/$tfile)) + + [[ "${sum[0]}" == "$expect" ]] || error "fallocate zero is not zero" +} +run_test 150bb "Verify fallocate modes both zero space" + +test_150c() { + check_set_fallocate_or_skip + + stack_trap "rm -f $DIR/$tfile; wait_delete_completed" + $LFS setstripe -c $OSTCOUNT -S1M $DIR/$tfile || error "setstripe failed" + fallocate -l ${OSTCOUNT}m $DIR/$tfile || error "fallocate failed" sync; sync_all_data cancel_lru_locks $OSC sleep 5 - bytes=$(($(stat -c '%b * %B' $DIR/$tdir))) + bytes=$(($(stat -c '%b * %B' $DIR/$tfile))) want=$((OSTCOUNT * 1048576)) # Must allocate all requested space, not more than 5% extra (( $bytes >= $want && $bytes < $want * 105 / 100 )) || error "bytes $bytes is not $want" + + rm -f $DIR/$tfile + # verify fallocate on PFL file + $LFS setstripe -E1M -c1 -E16M -c3 -Eeof -c 4 $DIR/$tfile || + error "Create $DIR/$tfile failed" + fallocate -l $((1048576 * 1024)) $DIR/$tfile || + error "fallocate failed" + sync; sync_all_data + cancel_lru_locks $OSC + sleep 5 + local bytes=$(($(stat -c '%b * %B' $DIR/$tfile))) + local want=$((1024 * 1048576)) + + # Must allocate all requested space, not more than 5% extra + (( $bytes >= $want && $bytes < $want * 105 / 100 )) || + error "bytes $bytes is not $want" } run_test 150c "Verify fallocate Size and Blocks" test_150d() { - local bytes - local want - - [ "$ost1_FSTYPE" != ldiskfs ] && skip "non-ldiskfs backend" - [ $OST1_VERSION -lt $(version_code 2.13.50) ] && - skip "Need OST version at least 2.13.53" + check_set_fallocate_or_skip + stack_trap "rm -f $DIR/$tfile; wait_delete_completed" $LFS setstripe -c $OSTCOUNT -S1M $DIR/$tdir || error "setstripe failed" fallocate -o 1G -l ${OSTCOUNT}m $DIR/$tdir || error "fallocate failed" sync; sync_all_data cancel_lru_locks $OSC sleep 5 - bytes=$(($(stat -c '%b * %B' $DIR/$tdir))) - want=$((OSTCOUNT * 1048576)) + local bytes=$(($(stat -c '%b * %B' $DIR/$tdir))) + local want=$((OSTCOUNT * 1048576)) # Must allocate all requested space, not more than 5% extra (( $bytes >= $want && $bytes < $want * 105 / 100 )) || @@ -13237,6 +13853,55 @@ test_150d() { } run_test 150d "Verify fallocate Size and Blocks - Non zero start" +test_150e() { + check_set_fallocate_or_skip + + echo "df before:" + $LFS df + stack_trap "rm -f $DIR/$tfile; wait_delete_completed" + $LFS setstripe -c${OSTCOUNT} $DIR/$tfile || + error "$LFS setstripe -c${OSTCOUNT} $DIR/$tfile failed" + + # Find OST with Minimum Size + min_size_ost=$($LFS df | awk "/$FSNAME-OST/ { print \$4 }" | + sort -un | head -1) + + # Get 100MB per OST of the available space to reduce run time + # else 60% of the available space if we are running SLOW tests + if [ $SLOW == "no" ]; then + local space=$((1024 * 100 * OSTCOUNT)) + else + local space=$(((min_size_ost * 60)/100 * OSTCOUNT)) + fi + + fallocate -l${space}k $DIR/$tfile || + error "fallocate ${space}k $DIR/$tfile failed" + echo "'fallocate -l ${space}k $DIR/$tfile' succeeded" + + # get size immediately after fallocate. This should be correctly + # updated + local size=$(stat -c '%s' $DIR/$tfile) + local used=$(( $(stat -c '%b * %B' $DIR/$tfile) / 1024)) + + # Sleep for a while for statfs to get updated. And not pull from cache. + sleep 2 + + echo "df after fallocate:" + $LFS df + + (( size / 1024 == space )) || error "size $size != requested $space" + [ "$ost1_FSTYPE" != ldiskfs ] || (( used >= space )) || + error "used $used < space $space" + + rm $DIR/$tfile || error "rm failed" + sync + wait_delete_completed + + echo "df after unlink:" + $LFS df +} +run_test 150e "Verify 60% of available OST space consumed by fallocate" + #LU-2902 roc_hit was not able to read all values from lproc function roc_hit_init() { local list=$(comma_list $(osts_nodes)) @@ -13522,10 +14187,19 @@ test_154A() { [ -z "$fid" ] && error "path2fid unable to get $tf FID" # check that we get the same pathname back - local found=$($LFS fid2path $MOUNT "$fid") - [ -z "$found" ] && error "fid2path unable to get '$fid' path" - [ "$found" == "$tf" ] || - error "fid2path($fid=path2fid($tf)) = $found != $tf" + local rootpath + local found + for rootpath in "$MOUNT" "$MOUNT///" "$MOUNT/$tfile"; do + echo "$rootpath $fid" + found=$($LFS fid2path $rootpath "$fid") + [ -z "$found" ] && error "fid2path unable to get '$fid' path" + [ "$found" == "$tf" ] || error "fid2path $found != $tf" + done + + # check wrong root path format + rootpath=$MOUNT"_wrong" + found=$($LFS fid2path $rootpath "$fid") + [ -z "$found" ] || error "should fail ($rootpath != $MOUNT)" } run_test 154A "lfs path2fid and fid2path basic checks" @@ -14347,12 +15021,14 @@ test_160f() { local i # generate some changelog records to accumulate on each MDT - # use fnv1a because created files should be evenly distributed - test_mkdir -c $MDSCOUNT -H fnv_1a_64 $DIR/$tdir || + # use all_char because created files should be evenly distributed + test_mkdir -c $MDSCOUNT -H all_char $DIR/$tdir || error "test_mkdir $tdir failed" log "$(date +%s): creating first files" - createmany -m $DIR/$tdir/$tfile $((MDSCOUNT * 2)) || - error "create $DIR/$tdir/$tfile failed" + for ((i = 0; i < MDSCOUNT * 2; i++)); do + $LFS mkdir -i $((i%MDSCOUNT)) $DIR/$tdir/d$i.$((i/MDSCOUNT)) || + error "create $DIR/$tdir/d$i.$((i/MDSCOUNT)) failed" + done # check changelogs have been generated local start=$SECONDS @@ -14380,7 +15056,7 @@ test_160f() { # simulate changelog catalog almost full #define OBD_FAIL_CAT_FREE_RECORDS 0x1313 - do_nodes $mdts $LCTL set_param fail_loc=0x1313 fail_val=3 + do_nodes $mdts "$LCTL set_param fail_loc=0x1313 fail_val=3" for i in $(seq $MDSCOUNT); do cl_users=(${CL_USERS[mds$i]}) @@ -14420,18 +15096,19 @@ test_160f() { # Generate one more changelog to trigger GC at fail_loc for cl_user2. # cl_user1 should be OK because it recently processed records. echo "$(date +%s): creating $((MDSCOUNT * 2)) files" - createmany -m $DIR/$tdir/${tfile}b $((MDSCOUNT * 2)) || - error "create $DIR/$tdir/${tfile}b failed" + for ((i = 0; i < MDSCOUNT * 2; i++)); do + $LFS mkdir -i $((i%MDSCOUNT)) $DIR/$tdir/d$i.$((i/MDSCOUNT+2))|| + error "create $DIR/$tdir/d$i.$((i/MDSCOUNT+2)) failed" + done # ensure gc thread is done for i in $(mdts_nodes); do - wait_update $i \ - "ps -e -o comm= | grep chlg_gc_thread" "" 20 || + wait_update $i "ps -e -o comm= | grep chlg_gc_thread" "" 20 || error "$i: GC-thread not done" done local first_rec - for i in $(seq $MDSCOUNT); do + for (( i = 1; i <= MDSCOUNT; i++ )); do # check cl_user1 still registered changelog_users mds$i | grep -q "${cl_user1[mds$i]}" || error "mds$i: User ${cl_user1[mds$i]} not registered" @@ -14446,10 +15123,9 @@ test_160f() { first_rec=$($LFS changelog $(facet_svc mds$i) | awk '{ print $1; exit; }') - echo "mds$i: verifying first index $user_rec1 + 1 == $first_rec" + echo "mds$i: $(date +%s) verify rec $user_rec1+1 == $first_rec" [ $((user_rec1 + 1)) == $first_rec ] || - error "mds$i: first index should be $user_rec1 + 1, " \ - "but is $first_rec" + error "mds$i: rec $first_rec != $user_rec1 + 1" done } run_test 160f "changelog garbage collect (timestamped users)" @@ -14475,20 +15151,20 @@ test_160g() { local i # generate some changelog records to accumulate on each MDT - # use fnv1a because created files should be evenly distributed - test_mkdir -c $MDSCOUNT -H fnv_1a_64 $DIR/$tdir || - error "mkdir $tdir failed" - createmany -m $DIR/$tdir/$tfile $((MDSCOUNT * 2)) || - error "create $DIR/$tdir/$tfile failed" + # use all_char because created files should be evenly distributed + test_mkdir -c $MDSCOUNT -H all_char $DIR/$tdir || + error "test_mkdir $tdir failed" + for ((i = 0; i < MDSCOUNT; i++)); do + $LFS mkdir -i $i $DIR/$tdir/d$i.1 $DIR/$tdir/d$i.2 || + error "create $DIR/$tdir/d$i.1 failed" + done # check changelogs have been generated local nbcl=$(changelog_dump | wc -l) - [[ $nbcl -eq 0 ]] && error "no changelogs found" + (( $nbcl > 0 )) || error "no changelogs found" # reduce the max_idle_indexes value to make sure we exceed it - max_ndx=$((nbcl / 2 - 1)) - - for param in "changelog_max_idle_indexes=$max_ndx" \ + for param in "changelog_max_idle_indexes=1" \ "changelog_gc=1" \ "changelog_min_gc_interval=2" \ "changelog_min_free_cat_entries=3"; do @@ -14503,8 +15179,9 @@ test_160g() { # simulate changelog catalog almost full #define OBD_FAIL_CAT_FREE_RECORDS 0x1313 - do_nodes $mdts $LCTL set_param fail_loc=0x1313 fail_val=3 + do_nodes $mdts "$LCTL set_param fail_loc=0x1313 fail_val=3" + local start=$SECONDS for i in $(seq $MDSCOUNT); do cl_users=(${CL_USERS[mds$i]}) cl_user1[mds$i]="${cl_users[0]}" @@ -14536,21 +15213,24 @@ test_160g() { done # ensure we are past the previous changelog_min_gc_interval set above - sleep 2 + local sleep2=$((start + 2 - SECONDS)) + (( sleep2 > 0 )) && echo "sleep $sleep2 for interval" && sleep $sleep2 - # generate one more changelog to trigger fail_loc - createmany -m $DIR/$tdir/${tfile}bis $((MDSCOUNT * 2)) || - error "create $DIR/$tdir/${tfile}bis failed" + # Generate one more changelog to trigger GC at fail_loc for cl_user2. + # cl_user1 should be OK because it recently processed records. + for ((i = 0; i < MDSCOUNT; i++)); do + $LFS mkdir -i $i $DIR/$tdir/d$i.3 $DIR/$tdir/d$i.4 || + error "create $DIR/$tdir/d$i.3 failed" + done # ensure gc thread is done for i in $(mdts_nodes); do - wait_update $i \ - "ps -e -o comm= | grep chlg_gc_thread" "" 20 || + wait_update $i "ps -e -o comm= | grep chlg_gc_thread" "" 20 || error "$i: GC-thread not done" done local first_rec - for i in $(seq $MDSCOUNT); do + for (( i = 1; i <= MDSCOUNT; i++ )); do # check cl_user1 still registered changelog_users mds$i | grep -q "${cl_user1[mds$i]}" || error "mds$i: User ${cl_user1[mds$i]} not registered" @@ -14565,10 +15245,9 @@ test_160g() { first_rec=$($LFS changelog $(facet_svc mds$i) | awk '{ print $1; exit; }') - echo "mds$i: verifying first index $user_rec1 + 1 == $first_rec" + echo "mds$i: $(date +%s) verify rec $user_rec1+1 == $first_rec" [ $((user_rec1 + 1)) == $first_rec ] || - error "mds$i: first index should be $user_rec1 + 1, " \ - "but is $first_rec" + error "mds$i: rec $first_rec != $user_rec1 + 1" done } run_test 160g "changelog garbage collect (old users)" @@ -14591,11 +15270,13 @@ test_160h() { local i # generate some changelog records to accumulate on each MDT - # use fnv1a because created files should be evenly distributed - test_mkdir -c $MDSCOUNT -H fnv_1a_64 $DIR/$tdir || + # use all_char because created files should be evenly distributed + test_mkdir -c $MDSCOUNT -H all_char $DIR/$tdir || error "test_mkdir $tdir failed" - createmany -m $DIR/$tdir/$tfile $((MDSCOUNT * 2)) || - error "create $DIR/$tdir/$tfile failed" + for ((i = 0; i < MDSCOUNT; i++)); do + $LFS mkdir -i $i $DIR/$tdir/d$i.1 $DIR/$tdir/d$i.2 || + error "create $DIR/$tdir/d$i.1 failed" + done # check changelogs have been generated local nbcl=$(changelog_dump | wc -l) @@ -14742,11 +15423,13 @@ test_160i() { changelog_register || error "first changelog_register failed" # generate some changelog records to accumulate on each MDT - # use fnv1a because created files should be evenly distributed - test_mkdir -c $MDSCOUNT -H fnv_1a_64 $DIR/$tdir || - error "mkdir $tdir failed" - createmany -m $DIR/$tdir/$tfile $((MDSCOUNT * 2)) || - error "create $DIR/$tdir/$tfile failed" + # use all_char because created files should be evenly distributed + test_mkdir -c $MDSCOUNT -H all_char $DIR/$tdir || + error "test_mkdir $tdir failed" + for ((i = 0; i < MDSCOUNT; i++)); do + $LFS mkdir -i $i $DIR/$tdir/d$i.1 $DIR/$tdir/d$i.2 || + error "create $DIR/$tdir/d$i.1 failed" + done # check changelogs have been generated local nbcl=$(changelog_dump | wc -l) @@ -14809,11 +15492,13 @@ test_160j() { stack_trap "changelog_deregister" EXIT # generate some changelog - # use fnv1a because created files should be evenly distributed - test_mkdir -c $MDSCOUNT -H fnv_1a_64 $DIR/$tdir || + # use all_char because created files should be evenly distributed + test_mkdir -c $MDSCOUNT -H all_char $DIR/$tdir || error "mkdir $tdir failed" - createmany -m $DIR/$tdir/${tfile}bis $((MDSCOUNT * 2)) || - error "create $DIR/$tdir/${tfile}bis failed" + for ((i = 0; i < MDSCOUNT; i++)); do + $LFS mkdir -i $i $DIR/$tdir/d$i.1 $DIR/$tdir/d$i.2 || + error "create $DIR/$tdir/d$i.1 failed" + done # open the changelog device exec 3>/dev/changelog-$FSNAME-MDT0000 @@ -14825,8 +15510,8 @@ test_160j() { umount $MOUNT stack_trap "mount_client $MOUNT" EXIT - # read changelog - cat <&4 >/dev/null || error "read changelog failed" + # read changelog, which may or may not fail, but should not crash + cat <&4 >/dev/null # clear changelog local cl_user="${CL_USERS[$SINGLEMDS]%% *}" @@ -14835,7 +15520,7 @@ test_160j() { printf 'clear:'$cl_user':0' >&3 } -run_test 160j "client can be umounted while its chanangelog is being used" +run_test 160j "client can be umounted while its chanangelog is being used" test_160k() { [ $PARALLEL == "yes" ] && skip "skip parallel run" @@ -14859,12 +15544,73 @@ test_160k() { 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" +# Verifies that a file passed as a parameter has recently had an operation +# performed on it that has generated an MTIME changelog which contains the +# correct parent FID. As files might reside on a different MDT from the +# parent directory in DNE configurations, the FIDs are translated to paths +# before being compared, which should be identical +compare_mtime_changelog() { + local file="${1}" + local mdtidx + local mtime + local cl_fid + local pdir + local dir + + mdtidx=$($LFS getstripe --mdt-index $file) + mdtidx=$(printf "%04x" $mdtidx) + + # Obtain the parent FID from the MTIME changelog + mtime=$($LFS changelog $FSNAME-MDT$mdtidx | tail -n 1 | grep MTIME) + [ -z "$mtime" ] && error "MTIME changelog not recorded" + + cl_fid=$(sed -e 's/.* p=//' -e 's/ .*//' <<<$mtime) + [ -z "$cl_fid" ] && error "parent FID not present" + + # Verify that the path for the parent FID is the same as the path for + # the test directory + pdir=$($LFS fid2path $MOUNT "$cl_fid") + + dir=$(dirname $1) + + [[ "${pdir%/}" == "$dir" ]] || + error "MTIME changelog parent FID is wrong, expected $dir, got $pdir" +} + +test_160l() { + [ $PARALLEL == "yes" ] && skip "skip parallel run" + + remote_mds_nodsh && skip "remote MDS with nodsh" + [[ $MDS1_VERSION -ge $(version_code 2.13.55) ]] || + skip "Need MDS version at least 2.13.55" + + local cl_user + + changelog_register || error "changelog_register failed" + cl_user="${CL_USERS[$SINGLEMDS]%% *}" + + changelog_users $SINGLEMDS | grep -q $cl_user || + error "User '$cl_user' not found in changelog_users" + + # Clear some types so that MTIME changelogs are generated + changelog_chmask "-CREAT" + changelog_chmask "-CLOSE" + + test_mkdir $DIR/$tdir || error "failed to mkdir $DIR/$tdir" + + # Test CL_MTIME during setattr + touch $DIR/$tdir/$tfile + compare_mtime_changelog $DIR/$tdir/$tfile + + # Test CL_MTIME during close + $MULTIOP $DIR/$tdir/${tfile}_2 O_2w4096c || error "multiop failed" + compare_mtime_changelog $DIR/$tdir/${tfile}_2 +} +run_test 160l "Verify that MTIME changelog records contain the parent FID" + test_161a() { [ $PARALLEL == "yes" ] && skip "skip parallel run" @@ -15236,8 +15982,11 @@ test_165a() { local rc local count - do_facet ost1 ofd_access_log_reader --debug=- --trace=- > "${trace}" & + (( $OST1_VERSION >= $(version_code 2.13.54) )) || + skip "OFD access log unsupported" + setup_165 + do_facet ost1 ofd_access_log_reader --debug=- --trace=- > "${trace}" & sleep 5 do_facet ost1 ofd_access_log_reader --list @@ -15269,13 +16018,19 @@ test_165b() { local size local flags + (( $OST1_VERSION >= $(version_code 2.13.54) )) || + skip "OFD access log unsupported" + setup_165 + do_facet ost1 ofd_access_log_reader --debug=- --trace=- > "${trace}" & + sleep 5 - lfs setstripe -c 1 -i 0 "${file}" - $MULTIOP "${file}" oO_CREAT:O_DIRECT:O_WRONLY:w1048576c || error "cannot create '${file}'" do_facet ost1 ofd_access_log_reader --list - do_facet ost1 ofd_access_log_reader --debug=- --trace=- > "${trace}" & + lfs setstripe -c 1 -i 0 "${file}" + $MULTIOP "${file}" oO_CREAT:O_DIRECT:O_WRONLY:w1048576c || + error "cannot create '${file}'" + sleep 5 do_facet ost1 killall -TERM ofd_access_log_reader wait @@ -15311,8 +16066,12 @@ test_165b() { fi do_facet ost1 ofd_access_log_reader --debug=- --trace=- > "${trace}" & - $MULTIOP "${file}" oO_CREAT:O_DIRECT:O_RDONLY:r524288c || error "cannot read '${file}'" sleep 5 + + $MULTIOP "${file}" oO_CREAT:O_DIRECT:O_RDONLY:r524288c || + error "cannot read '${file}'" + sleep 5 + do_facet ost1 killall -TERM ofd_access_log_reader wait rc=$? @@ -15344,89 +16103,202 @@ test_165b() { run_test 165b "ofd access log entries are produced and consumed" test_165c() { + local trace="/tmp/${tfile}.trace" local file="${DIR}/${tdir}/${tfile}" + + (( $OST1_VERSION >= $(version_code 2.13.54) )) || + skip "OFD access log unsupported" + test_mkdir "${DIR}/${tdir}" setup_165 + do_facet ost1 ofd_access_log_reader --debug=- --trace=- > "${trace}" & + sleep 5 lfs setstripe -c 1 -i 0 "${DIR}/${tdir}" # 4096 / 64 = 64. Create twice as many entries. for ((i = 0; i < 128; i++)); do - $MULTIOP "${file}-${i}" oO_CREAT:O_WRONLY:w512c || error "cannot create file" + $MULTIOP "${file}-${i}" oO_CREAT:O_WRONLY:w512c || + error "cannot create file" done sync - do_facet ost1 ofd_access_log_reader --list + + do_facet ost1 killall -TERM ofd_access_log_reader + wait + rc=$? + if ((rc != 0)); then + error "ofd_access_log_reader exited with rc = '${rc}'" + fi + unlinkmany "${file}-%d" 128 } run_test 165c "full ofd access logs do not block IOs" -oal_peek_entry_count() { - do_facet ost1 ofd_access_log_reader --list | awk '$1 == "_entry_count:" { print $2; }' +oal_get_read_count() { + local stats="$1" + + # STATS lustre-OST0001 alr_read_count 1 + + do_facet ost1 cat "${stats}" | + awk '$1 == "STATS" && $3 == "alr_read_count" { count = $4; } + END { print count; }' } -oal_expect_entry_count() { - local entry_count=$(oal_peek_entry_count) - local expect="$1" +oal_expect_read_count() { + local stats="$1" + local count + local expect="$2" + + # Ask ofd_access_log_reader to write stats. + do_facet ost1 killall -USR1 ofd_access_log_reader + + # Allow some time for things to happen. + sleep 1 - if ((entry_count == expect)); then + count=$(oal_get_read_count "${stats}") + if ((count == expect)); then return 0 fi - error_noexit "bad entry count, got ${entry_count}, expected ${expect}" - do_facet ost1 ofd_access_log_reader --list >&2 + error_noexit "bad read count, got ${count}, expected ${expect}" + do_facet ost1 cat "${stats}" >&2 exit 1 } test_165d() { - local trace="/tmp/${tfile}.trace" + local stats="/tmp/${tfile}.stats" local file="${DIR}/${tdir}/${tfile}" local param="obdfilter.${FSNAME}-OST0000.access_log_mask" - local entry_count + + (( $OST1_VERSION >= $(version_code 2.13.54) )) || + skip "OFD access log unsupported" + test_mkdir "${DIR}/${tdir}" setup_165 + do_facet ost1 ofd_access_log_reader --stats="${stats}" & + sleep 5 + lfs setstripe -c 1 -i 0 "${file}" do_facet ost1 lctl set_param "${param}=rw" - $MULTIOP "${file}" oO_CREAT:O_DIRECT:O_WRONLY:w1048576c || error "cannot create '${file}'" - oal_expect_entry_count 1 + $MULTIOP "${file}" oO_CREAT:O_DIRECT:O_WRONLY:w1048576c || + error "cannot create '${file}'" + oal_expect_read_count "${stats}" 1 - $MULTIOP "${file}" oO_CREAT:O_DIRECT:O_RDONLY:r1048576c || error "cannot read '${file}'" - oal_expect_entry_count 2 + $MULTIOP "${file}" oO_CREAT:O_DIRECT:O_RDONLY:r1048576c || + error "cannot read '${file}'" + oal_expect_read_count "${stats}" 2 do_facet ost1 lctl set_param "${param}=r" - $MULTIOP "${file}" oO_CREAT:O_DIRECT:O_WRONLY:w1048576c || error "cannot create '${file}'" - oal_expect_entry_count 2 + $MULTIOP "${file}" oO_CREAT:O_DIRECT:O_WRONLY:w1048576c || + error "cannot create '${file}'" + oal_expect_read_count "${stats}" 2 - $MULTIOP "${file}" oO_CREAT:O_DIRECT:O_RDONLY:r1048576c || error "cannot read '${file}'" - oal_expect_entry_count 3 + $MULTIOP "${file}" oO_CREAT:O_DIRECT:O_RDONLY:r1048576c || + error "cannot read '${file}'" + oal_expect_read_count "${stats}" 3 do_facet ost1 lctl set_param "${param}=w" - $MULTIOP "${file}" oO_CREAT:O_DIRECT:O_WRONLY:w1048576c || error "cannot create '${file}'" - oal_expect_entry_count 4 + $MULTIOP "${file}" oO_CREAT:O_DIRECT:O_WRONLY:w1048576c || + error "cannot create '${file}'" + oal_expect_read_count "${stats}" 4 - $MULTIOP "${file}" oO_CREAT:O_DIRECT:O_RDONLY:r1048576c || error "cannot read '${file}'" - oal_expect_entry_count 4 + $MULTIOP "${file}" oO_CREAT:O_DIRECT:O_RDONLY:r1048576c || + error "cannot read '${file}'" + oal_expect_read_count "${stats}" 4 do_facet ost1 lctl set_param "${param}=0" - $MULTIOP "${file}" oO_CREAT:O_DIRECT:O_WRONLY:w1048576c || error "cannot create '${file}'" - oal_expect_entry_count 4 + $MULTIOP "${file}" oO_CREAT:O_DIRECT:O_WRONLY:w1048576c || + error "cannot create '${file}'" + oal_expect_read_count "${stats}" 4 - $MULTIOP "${file}" oO_CREAT:O_DIRECT:O_RDONLY:r1048576c || error "cannot read '${file}'" - oal_expect_entry_count 4 + $MULTIOP "${file}" oO_CREAT:O_DIRECT:O_RDONLY:r1048576c || + error "cannot read '${file}'" + oal_expect_read_count "${stats}" 4 + + do_facet ost1 killall -TERM ofd_access_log_reader + wait + rc=$? + if ((rc != 0)); then + error "ofd_access_log_reader exited with rc = '${rc}'" + fi } run_test 165d "ofd_access_log mask works" +test_165e() { + local stats="/tmp/${tfile}.stats" + local file0="${DIR}/${tdir}-0/${tfile}" + local file1="${DIR}/${tdir}-1/${tfile}" + + (( $OST1_VERSION >= $(version_code 2.13.54) )) || + skip "OFD access log unsupported" + + [[ $MDSCOUNT -lt 2 ]] && skip_env "needs >= 2 MDTs" + + test_mkdir -c 1 -i 0 "${DIR}/${tdir}-0" + test_mkdir -c 1 -i 1 "${DIR}/${tdir}-1" + + lfs setstripe -c 1 -i 0 "${file0}" + lfs setstripe -c 1 -i 0 "${file1}" + + setup_165 + do_facet ost1 ofd_access_log_reader -I 1 --stats="${stats}" & + sleep 5 + + $MULTIOP "${file0}" oO_CREAT:O_WRONLY:w512c || + error "cannot create '${file0}'" + sync + oal_expect_read_count "${stats}" 0 + + $MULTIOP "${file1}" oO_CREAT:O_WRONLY:w512c || + error "cannot create '${file1}'" + sync + oal_expect_read_count "${stats}" 1 + + do_facet ost1 killall -TERM ofd_access_log_reader + wait + rc=$? + if ((rc != 0)); then + error "ofd_access_log_reader exited with rc = '${rc}'" + fi +} +run_test 165e "ofd_access_log MDT index filter works" + +test_165f() { + local trace="/tmp/${tfile}.trace" + local rc + local count + + setup_165 + do_facet ost1 timeout 60 ofd_access_log_reader \ + --exit-on-close --debug=- --trace=- > "${trace}" & + sleep 5 + stop ost1 + + wait + rc=$? + + if ((rc != 0)); then + error_noexit "ofd_access_log_reader exited with rc = '${rc}'" + cat "${trace}" + exit 1 + fi +} +run_test 165f "ofd_access_log_reader --exit-on-close works" + test_169() { # do directio so as not to populate the page cache log "creating a 10 Mb file" - $MULTIOP $DIR/$tfile oO_CREAT:O_DIRECT:O_RDWR:w$((10*1048576))c || error "multiop failed while creating a file" + $MULTIOP $DIR/$tfile oO_CREAT:O_DIRECT:O_RDWR:w$((10*1048576))c || + error "multiop failed while creating a file" log "starting reads" dd if=$DIR/$tfile of=/dev/null bs=4096 & log "truncating the file" - $MULTIOP $DIR/$tfile oO_TRUNC:c || error "multiop failed while truncating the file" + $MULTIOP $DIR/$tfile oO_TRUNC:c || + error "multiop failed while truncating the file" log "killing dd" kill %+ || true # reads might have finished echo "wait until dd is finished" @@ -15574,7 +16446,8 @@ obdecho_test() { test_180a() { [ $PARALLEL == "yes" ] && skip "skip parallel run" - if ! module_loaded obdecho; then + if ! [ -d /sys/fs/lustre/echo_client ] && + ! module_loaded obdecho; then load_module obdecho/obdecho && stack_trap "rmmod obdecho" EXIT || error "unable to load obdecho on client" @@ -15785,6 +16658,7 @@ test_184c() { local cmpn_arg=$(cmp -n 2>&1 | grep "invalid option") [ -n "$cmpn_arg" ] && skip_env "cmp does not support -n" check_swap_layouts_support + check_swap_layout_no_dom $DIR local dir0=$DIR/$tdir/$testnum mkdir -p $dir0 || error "creating dir $dir0" @@ -15803,8 +16677,9 @@ test_184c() { dd if=$ref1 of=$file1 bs=16k & local DD_PID=$! - # Make sure dd starts to copy file - while [ ! -f $file1 ]; do sleep 0.1; done + # Make sure dd starts to copy file, but wait at most 5 seconds + local loops=0 + while [ ! -s $file1 -a $((loops++)) -lt 50 ]; do sleep 0.1; done $LFS swap_layouts $file1 $file2 local rc=$? @@ -15832,6 +16707,7 @@ run_test 184c "Concurrent write and layout swap" test_184d() { check_swap_layouts_support + check_swap_layout_no_dom $DIR [ -z "$(which getfattr 2>/dev/null)" ] && skip_env "no getfattr command" @@ -15868,6 +16744,7 @@ test_184e() { [[ $MDS1_VERSION -ge $(version_code 2.6.94) ]] || skip "Need MDS version at least 2.6.94" check_swap_layouts_support + check_swap_layout_no_dom $DIR [ -z "$(which getfattr 2>/dev/null)" ] && skip_env "no getfattr command" @@ -16350,11 +17227,32 @@ test_205a() { # Job stats error "Unexpected jobids when jobid_var=$JOBENV" fi - lctl set_param jobid_var=USER jobid_name="S.%j.%e.%u.%h.E" - JOBENV="JOBCOMPLEX" - JOBCOMPLEX="S.$USER.touch.$(id -u).$(hostname).E" + # test '%j' access to environment variable - if supported + if lctl set_param jobid_var=USER jobid_name="S.%j.%e.%u.%h.E"; then + JOBENV="JOBCOMPLEX" + JOBCOMPLEX="S.$USER.touch.$(id -u).$(hostname).E" + + verify_jobstats "touch $DIR/$tfile" $SINGLEMDS + fi - verify_jobstats "touch $DIR/$tfile" $SINGLEMDS + if lctl set_param jobid_var=USER jobid_name="S.%j.%e.%u.%H.E"; then + JOBENV="JOBCOMPLEX" + JOBCOMPLEX="S.$USER.touch.$(id -u).$(hostname -s).E" + + verify_jobstats "touch $DIR/$tfile" $SINGLEMDS + fi + + # test '%j' access to per-session jobid - if supported + if lctl list_param jobid_this_session > /dev/null 2>&1 + then + lctl set_param jobid_var=session jobid_name="S.%j.%e.%u.%h.E" + lctl set_param jobid_this_session=$USER + + JOBENV="JOBCOMPLEX" + JOBCOMPLEX="S.$USER.touch.$(id -u).$(hostname).E" + + verify_jobstats "touch $DIR/$tfile" $SINGLEMDS + fi } run_test 205a "Verify job stats" @@ -16362,7 +17260,9 @@ run_test 205a "Verify job stats" test_205b() { job_stats="mdt.*.job_stats" $LCTL set_param $job_stats=clear - $LCTL set_param jobid_var=USER jobid_name="%e.%u" + # Setting jobid_var to USER might not be supported + $LCTL set_param jobid_var=USER || true + $LCTL set_param jobid_name="%e.%u" env -i USERTESTJOBSTATS=foolish touch $DIR/$tfile.1 do_facet $SINGLEMDS $LCTL get_param $job_stats | grep "job_id:.*foolish" && @@ -16373,6 +17273,17 @@ test_205b() { } run_test 205b "Verify job stats jobid and output format" +# LU-13733 +test_205c() { + $LCTL set_param llite.*.stats=0 + dd if=/dev/zero of=$DIR/$tfile.1 bs=4k count=1 + $LCTL get_param llite.*.stats + $LCTL get_param llite.*.stats | grep \ + "write_bytes *1 samples \[bytes\] 4096 4096 4096 16777216" || + error "wrong client stats format found" +} +run_test 205c "Verify client stats format" + # LU-1480, LU-1773 and LU-1657 test_206() { mkdir -p $DIR/$tdir @@ -16479,6 +17390,8 @@ test_209() { sync; sleep 5; sync; echo 3 > /proc/sys/vm/drop_caches + [ -f /sys/kernel/slab/ptlrpc_cache/shrink ] && + echo 1 > /sys/kernel/slab/ptlrpc_cache/shrink req_before=$(awk '/ptlrpc_cache / { print $2 }' /proc/slabinfo) # open/close 500 times @@ -16487,6 +17400,8 @@ test_209() { done echo 3 > /proc/sys/vm/drop_caches + [ -f /sys/kernel/slab/ptlrpc_cache/shrink ] && + echo 1 > /sys/kernel/slab/ptlrpc_cache/shrink req_after=$(awk '/ptlrpc_cache / { print $2 }' /proc/slabinfo) echo "before: $req_before, after: $req_after" @@ -17080,6 +17995,34 @@ test_226b () { } run_test 226b "call path2fid and fid2path on files of all type under remote dir" +test_226c () { + [ $MDSCOUNT -lt 2 ] && skip_env "needs >= 2 MDTs" + [[ $MDS1_VERSION -ge $(version_code 2.13.55) ]] || + skip "Need MDS version at least 2.13.55" + + local submnt=/mnt/submnt + local srcfile=/etc/passwd + local dstfile=$submnt/passwd + local path + local fid + + rm -rf $DIR/$tdir + rm -rf $submnt + $LFS setdirstripe -c -1 -i 1 $DIR/$tdir || + error "create remote directory failed" + mkdir -p $submnt || error "create $submnt failed" + $MOUNT_CMD $MGSNID:/$FSNAME/$tdir $submnt || + error "mount $submnt failed" + stack_trap "umount $submnt" EXIT + + cp $srcfile $dstfile + fid=$($LFS path2fid $dstfile) + path=$($LFS fid2path $submnt "$fid") + [ "$path" = "$dstfile" ] || + error "fid2path $submnt $fid failed ($path != $dstfile)" +} +run_test 226c "call path2fid and fid2path under remote dir with subdir mount" + # LU-1299 Executing or running ldd on a truncated executable does not # cause an out-of-memory condition. test_227() { @@ -17930,11 +18873,12 @@ test_230o() { skip "Need MDS version at least 2.13.52" local mdts=$(comma_list $(mdts_nodes)) - + local timeout=100 local restripe_status local delta local i - local j + + [[ $mds1_FSTYPE == zfs ]] && timeout=300 # in case "crush" hash type is not set do_nodes $mdts "$LCTL set_param lod.*.mdt_hash=crush" @@ -17952,18 +18896,18 @@ test_230o() { error "create dirs under remote dir failed $i" for i in $(seq 2 $MDSCOUNT); do - do_nodes $mdts "$LCTL set_param mdt.*.md_stats=clear > /dev/null" + do_nodes $mdts "$LCTL set_param mdt.*.md_stats=clear >/dev/null" $LFS setdirstripe -c $i $DIR/$tdir || error "split -c $i $tdir failed" wait_update $HOSTNAME \ - "$LFS getdirstripe -H $DIR/$tdir" "crush" 100 || + "$LFS getdirstripe -H $DIR/$tdir" "crush" $timeout || error "dir split not finished" delta=$(do_nodes $mdts "lctl get_param -n mdt.*MDT*.md_stats" | awk '/migrate/ {sum += $2} END { print sum }') - echo "$delta files migrated when dir split from $((i - 1)) to $i stripes" + echo "$delta migrated when dir split $((i - 1)) to $i stripes" # delta is around total_files/stripe_count - [ $delta -lt $((200 /(i - 1))) ] || - error "$delta files migrated" + (( $delta < 200 / (i - 1) + 4 )) || + error "$delta files migrated >= $((200 / (i - 1) + 4))" done } run_test 230o "dir split" @@ -17974,11 +18918,12 @@ test_230p() { skip "Need MDS version at least 2.13.52" local mdts=$(comma_list $(mdts_nodes)) - + local timeout=100 local restripe_status local delta local i - local j + + [[ $mds1_FSTYPE == zfs ]] && timeout=300 do_nodes $mdts "$LCTL set_param lod.*.mdt_hash=crush" @@ -17997,19 +18942,19 @@ test_230p() { for i in $(seq $((MDSCOUNT - 1)) -1 1); do local mdt_hash="crush" - do_nodes $mdts "$LCTL set_param mdt.*.md_stats=clear > /dev/null" + do_nodes $mdts "$LCTL set_param mdt.*.md_stats=clear >/dev/null" $LFS setdirstripe -c $i $DIR/$tdir || error "split -c $i $tdir failed" [ $i -eq 1 ] && mdt_hash="none" wait_update $HOSTNAME \ - "$LFS getdirstripe -H $DIR/$tdir" $mdt_hash 100 || + "$LFS getdirstripe -H $DIR/$tdir" $mdt_hash $timeout || error "dir merge not finished" delta=$(do_nodes $mdts "lctl get_param -n mdt.*MDT*.md_stats" | awk '/migrate/ {sum += $2} END { print sum }') - echo "$delta files migrated when dir merge from $((i + 1)) to $i stripes" + echo "$delta migrated when dir merge $((i + 1)) to $i stripes" # delta is around total_files/stripe_count - [ $delta -lt $((200 / i)) ] || - error "$delta files migrated" + (( $delta < 200 / i + 4 )) || + error "$delta files migrated >= $((200 / i + 4))" done } run_test 230p "dir merge" @@ -18030,6 +18975,10 @@ test_230q() { local stripe_count=0 local stripe_index local nr_files + local create + + # test with fewer files on ZFS + [ "$mds1_FSTYPE" == "zfs" ] && threshold=40 stack_trap "do_nodes $mdts $LCTL set_param \ mdt.*.dir_split_count=$saved_threshold" @@ -18045,11 +18994,12 @@ test_230q() { $LFS mkdir -i -1 -c 1 $DIR/$tdir || error "mkdir $tdir failed" stripe_index=$($LFS getdirstripe -i $DIR/$tdir) + create=$((threshold * 3 / 2)) while [ $stripe_count -lt $MDSCOUNT ]; do - createmany -m $DIR/$tdir/f $total $((threshold * 3 / 2)) || + createmany -m $DIR/$tdir/f $total $create || error "create sub files failed" stat $DIR/$tdir > /dev/null - total=$((total + threshold * 3 / 2)) + total=$((total + create)) stripe_count=$((stripe_count + delta)) [ $stripe_count -gt $MDSCOUNT ] && stripe_count=$MDSCOUNT @@ -18061,19 +19011,54 @@ test_230q() { "$LFS getdirstripe -H $DIR/$tdir" "crush" 200 || error "stripe hash $($LFS getdirstripe -H $DIR/$tdir) != crush" - nr_files=$($LFS getstripe -m $DIR/$tdir/* | - grep -w $stripe_index | wc -l) - echo "$nr_files files on MDT$stripe_index after split" - [ $nr_files -lt $((total / (stripe_count - 1))) ] || + nr_files=$($LFS find -m 1 $DIR/$tdir | grep -c -w $stripe_index) + echo "$nr_files/$total files on MDT$stripe_index after split" + # allow 10% margin of imbalance with crush hash + (( $nr_files <= $total / $stripe_count + $create / 10)) || error "$nr_files files on MDT$stripe_index after split" - nr_files=$(ls $DIR/$tdir | wc -w) + nr_files=$($LFS find -type f $DIR/$tdir | wc -l) [ $nr_files -eq $total ] || error "total sub files $nr_files != $total" done } run_test 230q "dir auto split" +test_230r() { + [[ $PARALLEL != "yes" ]] || skip "skip parallel run" + [[ $MDSCOUNT -ge 2 ]] || skip_env "needs >= 2 MDTs" + [[ $MDS1_VERSION -ge $(version_code 2.13.54) ]] || + skip "Need MDS version at least 2.13.54" + + # maximum amount of local locks: + # parent striped dir - 2 locks + # new stripe in parent to migrate to - 1 lock + # source and target - 2 locks + # Total 5 locks for regular file + mkdir -p $DIR/$tdir + $LFS mkdir -i1 -c2 $DIR/$tdir/dir1 + touch $DIR/$tdir/dir1/eee + + # create 4 hardlink for 4 more locks + # Total: 9 locks > RS_MAX_LOCKS (8) + $LFS mkdir -i1 -c1 $DIR/$tdir/dir2 + $LFS mkdir -i1 -c1 $DIR/$tdir/dir3 + $LFS mkdir -i1 -c1 $DIR/$tdir/dir4 + $LFS mkdir -i1 -c1 $DIR/$tdir/dir5 + ln $DIR/$tdir/dir1/eee $DIR/$tdir/dir2/eee + ln $DIR/$tdir/dir1/eee $DIR/$tdir/dir3/eee + ln $DIR/$tdir/dir1/eee $DIR/$tdir/dir4/eee + ln $DIR/$tdir/dir1/eee $DIR/$tdir/dir5/eee + + cancel_lru_locks mdc + + $LFS migrate -m1 -c1 $DIR/$tdir/dir1 || + error "migrate dir fails" + + rm -rf $DIR/$tdir || error "rm dir failed after migration" +} +run_test 230r "migrate with too many local locks" + test_231a() { # For simplicity this test assumes that max_pages_per_rpc @@ -18535,7 +19520,7 @@ test_247c() { $LFS fid2path $submount $fid && error "fid2path should fail" cleanup_247 $submount } -run_test 247c "running fid2path outside root" +run_test 247c "running fid2path outside subdirectory root" test_247d() { lctl get_param -n mdc.$FSNAME-MDT0000*.import | grep -q subtree || @@ -18548,11 +19533,28 @@ test_247d() { FILESET="$FILESET/$tdir" mount_client $submount || error "mount $submount failed" trap "cleanup_247 $submount" EXIT - local fid=$($LFS path2fid $submount/dir1) - $LFS fid2path $submount $fid || error "fid2path should succeed" + + local td=$submount/dir1 + local fid=$($LFS path2fid $td) + [ -z "$fid" ] && error "path2fid unable to get $td FID" + + # check that we get the same pathname back + local rootpath + local found + for rootpath in "$submount" "$submount///" "$submount/dir1"; do + echo "$rootpath $fid" + found=$($LFS fid2path $rootpath "$fid") + [ -n "found" ] || error "fid2path should succeed" + [ "$found" == "$td" ] || error "fid2path $found != $td" + done + # check wrong root path format + rootpath=$submount"_wrong" + found=$($LFS fid2path $rootpath "$fid") + [ -z "$found" ] || error "fid2path should fail ($rootpath != $submount)" + cleanup_247 $submount } -run_test 247d "running fid2path inside root" +run_test 247d "running fid2path inside subdirectory root" # LU-8037 test_247e() { @@ -18574,6 +19576,8 @@ test_247f() { [ $MDSCOUNT -lt 2 ] && skip_env "needs >= 2 MDTs" [ $MDS1_VERSION -lt $(version_code 2.13.52) ] && skip "Need at least version 2.13.52" + [ $CLIENT_VERSION -lt $(version_code 2.14.50) ] && + skip "Need at least version 2.14.50" lctl get_param -n mdc.$FSNAME-MDT0000*.import | grep -q subtree || skip "Fileset feature is not supported" @@ -18582,26 +19586,78 @@ test_247f() { $LFS mkdir -i $((MDSCOUNT - 1)) $DIR/$tdir/remote || error "mkdir remote failed" mkdir $DIR/$tdir/remote/subdir || error "mkdir remote/subdir failed" - $LFS mkdir -c $MDSCOUNT $DIR/$tdir/striped || + $LFS mkdir -i 0 -c $MDSCOUNT $DIR/$tdir/striped || error "mkdir striped failed" mkdir $DIR/$tdir/striped/subdir || error "mkdir striped/subdir failed" local submount=${MOUNT}_$tdir mkdir -p $submount || error "mkdir $submount failed" + stack_trap "rmdir $submount" local dir + local stat local fileset=$FILESET + local mdts=$(comma_list $(mdts_nodes)) - for dir in $tdir/remote $tdir/remote/subdir \ - $tdir/striped $tdir/striped/subdir $tdir/striped/. ; do + stat=$(do_facet mds1 $LCTL get_param -n \ + mdt.*MDT0000.enable_remote_subdir_mount) + stack_trap "do_nodes $mdts $LCTL set_param \ + mdt.*.enable_remote_subdir_mount=$stat" + + do_nodes $mdts "$LCTL set_param mdt.*.enable_remote_subdir_mount=0" + stack_trap "umount_client $submount" + FILESET="$fileset/$tdir/remote" mount_client $submount && + error "mount remote dir $dir should fail" + + for dir in $tdir/remote/subdir $tdir/striped $tdir/striped/subdir \ + $tdir/striped/. ; do FILESET="$fileset/$dir" mount_client $submount || error "mount $dir failed" umount_client $submount done + + do_nodes $mdts "$LCTL set_param mdt.*.enable_remote_subdir_mount=1" + FILESET="$fileset/$tdir/remote" mount_client $submount || + error "mount $tdir/remote failed" } run_test 247f "mount striped or remote directory as fileset" +test_247g() { + [ $MDSCOUNT -lt 4 ] && skip_env "needs >= 4 MDTs" + [ $CLIENT_VERSION -lt $(version_code 2.14.50) ] && + skip "Need at least version 2.14.50" + + $LFS mkdir -i 0 -c 4 -H fnv_1a_64 $DIR/$tdir || + error "mkdir $tdir failed" + touch $DIR/$tdir/$tfile || error "touch $tfile failed" + + local submount=${MOUNT}_$tdir + + mkdir -p $submount || error "mkdir $submount failed" + stack_trap "rmdir $submount" + + FILESET="$fileset/$tdir" mount_client $submount || + error "mount $dir failed" + stack_trap "umount $submount" + + local mdts=$(comma_list $(mdts_nodes)) + + local nrpcs + + stat $submount > /dev/null + cancel_lru_locks $MDC + stat $submount > /dev/null + stat $submount/$tfile > /dev/null + do_nodes $mdts "$LCTL set_param mdt.*.md_stats=clear > /dev/null" + stat $submount/$tfile > /dev/null + nrpcs=$(do_nodes $mdts "lctl get_param -n mdt.*.md_stats" | + awk '/getattr/ {sum += $2} END {print sum}') + + [ -z "$nrpcs" ] || error "$nrpcs extra getattr sent" +} +run_test 247g "mount striped directory as fileset caches ROOT lookup lock" + test_248a() { local fast_read_sav=$($LCTL get_param -n llite.*.fast_read 2>/dev/null) [ -z "$fast_read_sav" ] && skip "no fast read support" @@ -19043,7 +20099,7 @@ ladvise_willread_performance() return 0 lowest_speedup=$(bc <<<"scale=2; $average_cache / 2") - [ ${average_ladvise%.*} -gt $lowest_speedup ] || + [[ ${average_ladvise%.*} > $lowest_speedup ]] || error_not_in_vm "Speedup with willread is less than " \ "$lowest_speedup%, got $average_ladvise%" } @@ -19202,6 +20258,9 @@ test_255c() { [ $OST1_VERSION -lt $(version_code 2.10.50) ] && skip "lustre < 2.10.50 does not support lockahead" + local ost1_imp=$(get_osc_import_name client ost1) + local imp_name=$($LCTL list_param osc.$ost1_imp | head -n1 | + cut -d'.' -f2) local count local new_count local difference @@ -19249,7 +20308,7 @@ test_255c() { cancel_lru_locks osc count=$($LCTL get_param -n \ - ldlm.namespaces.$FSNAME-OST0000*osc-[-0-9a-f]*.lock_unused_count) + ldlm.namespaces.$imp_name.lock_unused_count) lockahead_test -d $DIR/$tdir -t $i -f $tfile rc=$? @@ -19258,7 +20317,7 @@ test_255c() { fi new_count=$($LCTL get_param -n \ - ldlm.namespaces.$FSNAME-OST0000*osc-[-0-9a-f]*.lock_unused_count) + ldlm.namespaces.$imp_name.lock_unused_count) difference="$((new_count - count))" # Test 15 output is divided by 100 to map down to valid return @@ -20342,6 +21401,17 @@ test_273a() { } run_test 273a "DoM: layout swapping should fail with DOM" +test_273b() { + mkdir -p $DIR/$tdir + $LFS setstripe -E 1M -L mdt -E -1 -c -1 $DIR/$tdir + +#define OBD_FAIL_MDS_COMMITRW_DELAY 0x16b + do_facet mds1 $LCTL set_param fail_loc=0x8000016b fail_val=2 + + $MULTIOP $DIR/$tdir/$tfile Ouw2097152c +} +run_test 273b "DoM: race writeback and object destroy" + test_275() { remote_ost_nodsh && skip "remote OST with nodsh" [ $OST1_VERSION -lt $(version_code 2.10.57) ] && @@ -20451,6 +21521,9 @@ test_280() { stop mgs || error "stop mgs failed" #for a race mgs would crash start mgs $(mgsdevname) $MGS_MOUNT_OPTS || error "start mgs failed" + # make sure we unmount client before remounting + wait + umount_client $MOUNT mount_client $MOUNT || error "mount client failed" } run_test 280 "Race between MGS umount and client llog processing" @@ -20737,7 +21810,7 @@ test_300_check_default_striped_dir() echo "checking $dirname $default_count $default_index" $LFS setdirstripe -D -c $default_count -i $default_index \ - -t all_char $DIR/$tdir/$dirname || + -H all_char $DIR/$tdir/$dirname || error "set default stripe on striped dir error" stripe_count=$($LFS getdirstripe -D -c $DIR/$tdir/$dirname) [ $stripe_count -eq $default_count ] || @@ -20754,8 +21827,9 @@ test_300_check_default_striped_dir() unlinkmany $DIR/$tdir/$dirname/f- 10 || error "unlink files failed" for dir in $(find $DIR/$tdir/$dirname/*); do stripe_count=$($LFS getdirstripe -c $dir) - [ $stripe_count -eq $default_count ] || - [ $stripe_count -eq 0 ] || [ $default_count -eq 1 ] || + (( $stripe_count == $default_count )) || + (( $stripe_count == $MDSCOUNT && $default_count == -1 )) || + (( $stripe_count == 0 )) || (( $default_count == 1 )) || error "stripe count $default_count != $stripe_count for $dir" stripe_index=$($LFS getdirstripe -i $dir) @@ -20807,7 +21881,7 @@ test_300g() { #check default stripe count/stripe index test_300_check_default_striped_dir normal_dir $MDSCOUNT 1 test_300_check_default_striped_dir normal_dir 1 0 - test_300_check_default_striped_dir normal_dir 2 1 + test_300_check_default_striped_dir normal_dir -1 1 test_300_check_default_striped_dir normal_dir 2 -1 #delete default stripe information @@ -20841,7 +21915,7 @@ test_300h() { test_300_check_default_striped_dir striped_dir $MDSCOUNT 1 test_300_check_default_striped_dir striped_dir 1 0 - test_300_check_default_striped_dir striped_dir 2 1 + test_300_check_default_striped_dir striped_dir -1 1 test_300_check_default_striped_dir striped_dir 2 -1 #delete default stripe information @@ -20881,15 +21955,17 @@ test_300i() { error "create dir0 under hash dir failed" $LFS setdirstripe -i0 -c$MDSCOUNT -H fnv_1a_64 $DIR/$tdir/hashdir/d1 || error "create dir1 under hash dir failed" + $LFS setdirstripe -i0 -c$MDSCOUNT -H crush $DIR/$tdir/hashdir/d2 || + error "create dir2 under hash dir failed" # unfortunately, we need to umount to clear dir layout cache for now # once we fully implement dir layout, we can drop this umount_client $MOUNT || error "umount failed" mount_client $MOUNT || error "mount failed" - $LFS find -H fnv_1a_64 $DIR/$tdir/hashdir - local dircnt=$($LFS find -H fnv_1a_64 $DIR/$tdir/hashdir | wc -l) - [ $dircnt -eq 1 ] || error "lfs find striped dir got:$dircnt,except:1" + $LFS find -H fnv_1a_64,crush $DIR/$tdir/hashdir + local dircnt=$($LFS find -H fnv_1a_64,crush $DIR/$tdir/hashdir | wc -l) + [ $dircnt -eq 2 ] || error "lfs find striped dir got:$dircnt,except:1" #set the stripe to be unknown hash type #define OBD_FAIL_UNKNOWN_LMV_STRIPE 0x1901 @@ -21181,6 +22257,56 @@ test_300r() { } run_test 300r "test -1 striped directory" +test_300s_helper() { + local count=$1 + + local stripe_dir=$DIR/$tdir/striped_dir.$count + + $LFS mkdir -c $count $stripe_dir || + error "lfs mkdir -c error" + + $LFS getdirstripe $stripe_dir || + error "lfs getdirstripe fails" + + local stripe_count + stripe_count=$($LFS getdirstripe $stripe_dir | + awk '/lmv_stripe_count:/ { print $2 }') + + [ $count -ne $stripe_count ] && + error_noexit "bad stripe count $stripe_count expected $count" + + local dupe_stripes + dupe_stripes=$($LFS getdirstripe $stripe_dir | + awk '/0x/ {count[$1] += 1}; END { + for (idx in count) { + if (count[idx]>1) { + print "index " idx " count " count[idx] + } + } + }') + + if [[ -n "$dupe_stripes" ]] ; then + lfs getdirstripe $stripe_dir + error_noexit "Dupe MDT above: $dupe_stripes " + fi + + rm -rf $stripe_dir || + error_noexit "unlink $stripe_dir fails" +} + +test_300s() { + [ $MDS1_VERSION -lt $(version_code 2.7.55) ] && + skip "Need MDS version at least 2.7.55" && return + [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return + + mkdir $DIR/$tdir + for count in $(seq 2 $MDSCOUNT); do + test_300s_helper $count + done +} +run_test 300s "test lfs mkdir -c without -i" + + prepare_remote_file() { mkdir $DIR/$tdir/src_dir || error "create remote source failed" @@ -21268,7 +22394,7 @@ test_311() { skip "lustre < 2.8.54 does not contain LU-4825 fix" remote_mds_nodsh && skip "remote MDS with nodsh" - local old_iused=$($LFS df -i | grep OST0000 | awk '{ print $3 }') + local old_iused=$($LFS df -i | awk '/OST0000/ { print $3; exit; }') local mdts=$(comma_list $(mdts_nodes)) mkdir -p $DIR/$tdir @@ -21300,7 +22426,7 @@ test_311() { local new_iused for i in $(seq 120); do - new_iused=$($LFS df -i | grep OST0000 | awk '{ print $3 }') + new_iused=$($LFS df -i | awk '/OST0000/ { print $3; exit; }') # system may be too busy to destroy all objs in time, use # a somewhat small value to not fail autotest [ $((old_iused - new_iused)) -gt 400 ] && break @@ -21546,12 +22672,15 @@ test_317() { run_test 317 "Verify blocks get correctly update after truncate" test_318() { + local llite_name="llite.$($LFS getname $MOUNT | awk '{print $1}')" local old_max_active=$($LCTL get_param -n \ - llite.*.max_read_ahead_async_active 2>/dev/null) + ${llite_name}.max_read_ahead_async_active \ + 2>/dev/null) $LCTL set_param llite.*.max_read_ahead_async_active=256 local max_active=$($LCTL get_param -n \ - llite.*.max_read_ahead_async_active 2>/dev/null) + ${llite_name}.max_read_ahead_async_active \ + 2>/dev/null) [ $max_active -ne 256 ] && error "expected 256 but got $max_active" $LCTL set_param llite.*.max_read_ahead_async_active=0 || @@ -21559,7 +22688,7 @@ test_318() { $LCTL set_param llite.*.max_read_ahead_async_active=512 max_active=$($LCTL get_param -n \ - llite.*.max_read_ahead_async_active 2>/dev/null) + ${llite_name}.max_read_ahead_async_active 2>/dev/null) [ $max_active -eq 512 ] || error "expected 512 but got $max_active" # restore @max_active @@ -21567,9 +22696,9 @@ test_318() { llite.*.max_read_ahead_async_active=$old_max_active local old_threshold=$($LCTL get_param -n \ - llite.*.read_ahead_async_file_threshold_mb 2>/dev/null) + ${llite_name}.read_ahead_async_file_threshold_mb 2>/dev/null) local max_per_file_mb=$($LCTL get_param -n \ - llite.*.max_read_ahead_per_file_mb 2>/dev/null) + ${llite_name}.max_read_ahead_per_file_mb 2>/dev/null) local invalid=$(($max_per_file_mb + 1)) $LCTL set_param \ @@ -21581,7 +22710,7 @@ test_318() { llite.*.read_ahead_async_file_threshold_mb=$valid || error "set $valid should succeed" local threshold=$($LCTL get_param -n \ - llite.*.read_ahead_async_file_threshold_mb 2>/dev/null) + ${llite_name}.read_ahead_async_file_threshold_mb 2>/dev/null) [ $threshold -eq $valid ] || error \ "expect threshold $valid got $threshold" $LCTL set_param \ @@ -21615,6 +22744,10 @@ test_319() { run_test 319 "lost lease lock on migrate error" test_398a() { # LU-4198 + local ost1_imp=$(get_osc_import_name client ost1) + local imp_name=$($LCTL list_param osc.$ost1_imp | head -n1 | + cut -d'.' -f2) + $LFS setstripe -c 1 -i 0 $DIR/$tfile $LCTL set_param ldlm.namespaces.*.lru_size=clear @@ -21623,7 +22756,7 @@ test_398a() { # LU-4198 dd if=/dev/zero of=$DIR/$tfile bs=1M count=1 oflag=direct conv=notrunc local lock_count=$($LCTL get_param -n \ - ldlm.namespaces.*-OST0000-osc-ffff*.lru_size) + ldlm.namespaces.$imp_name.lru_size) [[ $lock_count -eq 0 ]] || error "lock should be cancelled by direct IO" $LCTL set_param ldlm.namespaces.*-OST0000-osc-ffff*.lru_size=clear @@ -21631,7 +22764,7 @@ test_398a() { # LU-4198 # no lock cached, should use lockless IO and not enqueue new lock dd if=/dev/zero of=$DIR/$tfile bs=1M count=1 oflag=direct conv=notrunc lock_count=$($LCTL get_param -n \ - ldlm.namespaces.*-OST0000-osc-ffff*.lru_size) + ldlm.namespaces.$imp_name.lru_size) [[ $lock_count -eq 0 ]] || error "no lock should be held by direct IO" } run_test 398a "direct IO should cancel lock otherwise lockless" @@ -21663,6 +22796,10 @@ test_398b() { # LU-4198 run_test 398b "DIO and buffer IO race" test_398c() { # LU-4198 + local ost1_imp=$(get_osc_import_name client ost1) + local imp_name=$($LCTL list_param osc.$ost1_imp | head -n1 | + cut -d'.' -f2) + which fio || skip_env "no fio installed" saved_debug=$($LCTL get_param -n debug) @@ -21691,12 +22828,11 @@ test_398c() { # LU-4198 --filename=$DIR/$tfile [ $? -eq 0 ] || error "fio write error" - [ $($LCTL get_param -n \ - ldlm.namespaces.${FSNAME}-OST0000-osc-ffff*.lock_count) -eq 0 ] || + [ $($LCTL get_param -n ldlm.namespaces.$imp_name.lock_count) -eq 0 ] || error "Locks were requested while doing AIO" # get the percentage of 1-page I/O - pct=$($LCTL get_param osc.${FSNAME}-OST0000-osc-ffff*.rpc_stats | + pct=$($LCTL get_param osc.${imp_name}.rpc_stats | grep -A 1 'pages per rpc' | grep -v 'pages per rpc' | awk '{print $7}') [ $pct -le 50 ] || error "$pct% of I/O are 1-page" @@ -21708,11 +22844,44 @@ test_398c() { # LU-4198 --filename=$DIR/$tfile [ $? -eq 0 ] || error "fio mixed read write error" + echo "AIO with large block size ${size}M" + fio --name=rand-rw --rw=randrw --bs=${size}M --direct=1 \ + --numjobs=1 --fallocate=none --ioengine=libaio \ + --iodepth=16 --allow_file_create=0 --size=${size}M \ + --filename=$DIR/$tfile + [ $? -eq 0 ] || error "fio large block size failed" + rm -rf $DIR/$tfile $LCTL set_param debug="$saved_debug" } run_test 398c "run fio to test AIO" +test_398d() { # LU-13846 + test -f aiocp || skip_env "no aiocp installed" + local aio_file=$DIR/aio_file + + $LFS setstripe -c -1 -S 1M $DIR/$tfile $aio_file + + dd if=/dev/urandom of=$DIR/$tfile bs=1M count=64 + aiocp -a $PAGE_SIZE -b 64M -s 64M -f O_DIRECT $DIR/$tfile $aio_file + + diff $DIR/$tfile $aio_file || "file diff after aiocp" + + # make sure we don't crash and fail properly + aiocp -a 512 -b 64M -s 64M -f O_DIRECT $DIR/$tfile $aio_file && + error "aio not aligned with PAGE SIZE should fail" + + rm -rf $DIR/$tfile $aio_file +} +run_test 398d "run aiocp to verify block size > stripe size" + +test_398e() { + dd if=/dev/zero of=$DIR/$tfile bs=1234 count=1 + touch $DIR/$tfile.new + dd if=$DIR/$tfile of=$DIR/$tfile.new bs=1M count=1 oflag=direct +} +run_test 398e "O_Direct open cleared by fcntl doesn't cause hang" + test_fake_rw() { local read_write=$1 if [ "$read_write" = "write" ]; then @@ -21730,12 +22899,12 @@ test_fake_rw() { $LFS setstripe -c 1 -i 0 $DIR/$tfile # get ost1 size - $FSNAME-OST0000 - local ost1_avail_size=$($LFS df | awk /${ost1_svc}/'{ print $4 }') + local ost1_avail_size=$($LFS df $DIR | awk /${ost1_svc}/'{ print $4 }') local blocks=$((ost1_avail_size/2/1024)) # half avail space by megabytes [ $blocks -gt 1000 ] && blocks=1000 # 1G in maximum if [ "$read_write" = "read" ]; then - truncate -s $(expr 1048576 \* $blocks) $DIR/$tfile + $TRUNCATE $DIR/$tfile $(expr 1048576 \* $blocks) fi local start_time=$(date +%s.%N) @@ -21877,70 +23046,92 @@ test_401a() { #LU-7437 run_test 401a "Verify if 'lctl list_param -R' can list parameters recursively" test_401b() { - local save=$($LCTL get_param -n jobid_var) - local tmp=testing + # jobid_var may not allow arbitrary values, so use jobid_name + # if available + if $LCTL list_param jobid_name > /dev/null 2>&1; then + local testname=jobid_name tmp='testing%p' + else + local testname=jobid_var tmp=testing + fi - $LCTL set_param foo=bar jobid_var=$tmp bar=baz && + local save=$($LCTL get_param -n $testname) + + $LCTL set_param foo=bar $testname=$tmp bar=baz && error "no error returned when setting bad parameters" - local jobid_new=$($LCTL get_param -n foe jobid_var baz) + local jobid_new=$($LCTL get_param -n foe $testname baz) [[ "$jobid_new" == "$tmp" ]] || error "jobid tmp $jobid_new != $tmp" - $LCTL set_param -n fog=bam jobid_var=$save bat=fog - local jobid_old=$($LCTL get_param -n foe jobid_var bag) + $LCTL set_param -n fog=bam $testname=$save bat=fog + local jobid_old=$($LCTL get_param -n foe $testname bag) [[ "$jobid_old" == "$save" ]] || error "jobid new $jobid_old != $save" } run_test 401b "Verify 'lctl {get,set}_param' continue after error" test_401c() { - local jobid_var_old=$($LCTL get_param -n jobid_var) + # jobid_var may not allow arbitrary values, so use jobid_name + # if available + if $LCTL list_param jobid_name > /dev/null 2>&1; then + local testname=jobid_name + else + local testname=jobid_var + fi + + local jobid_var_old=$($LCTL get_param -n $testname) local jobid_var_new - $LCTL set_param jobid_var= && + $LCTL set_param $testname= && error "no error returned for 'set_param a='" - jobid_var_new=$($LCTL get_param -n jobid_var) + jobid_var_new=$($LCTL get_param -n $testname) [[ "$jobid_var_old" == "$jobid_var_new" ]] || - error "jobid_var was changed by setting without value" + error "$testname was changed by setting without value" - $LCTL set_param jobid_var && + $LCTL set_param $testname && error "no error returned for 'set_param a'" - jobid_var_new=$($LCTL get_param -n jobid_var) + jobid_var_new=$($LCTL get_param -n $testname) [[ "$jobid_var_old" == "$jobid_var_new" ]] || - error "jobid_var was changed by setting without value" + error "$testname was changed by setting without value" } run_test 401c "Verify 'lctl set_param' without value fails in either format." test_401d() { - local jobid_var_old=$($LCTL get_param -n jobid_var) + # jobid_var may not allow arbitrary values, so use jobid_name + # if available + if $LCTL list_param jobid_name > /dev/null 2>&1; then + local testname=jobid_name new_value='foo=bar%p' + else + local testname=jobid_var new_valuie=foo=bar + fi + + local jobid_var_old=$($LCTL get_param -n $testname) local jobid_var_new - local new_value="foo=bar" - $LCTL set_param jobid_var=$new_value || + $LCTL set_param $testname=$new_value || error "'set_param a=b' did not accept a value containing '='" - jobid_var_new=$($LCTL get_param -n jobid_var) + jobid_var_new=$($LCTL get_param -n $testname) [[ "$jobid_var_new" == "$new_value" ]] || error "'set_param a=b' failed on a value containing '='" - # Reset the jobid_var to test the other format - $LCTL set_param jobid_var=$jobid_var_old - jobid_var_new=$($LCTL get_param -n jobid_var) + # Reset the $testname to test the other format + $LCTL set_param $testname=$jobid_var_old + jobid_var_new=$($LCTL get_param -n $testname) [[ "$jobid_var_new" == "$jobid_var_old" ]] || - error "failed to reset jobid_var" + error "failed to reset $testname" - $LCTL set_param jobid_var $new_value || + $LCTL set_param $testname $new_value || error "'set_param a b' did not accept a value containing '='" - jobid_var_new=$($LCTL get_param -n jobid_var) + jobid_var_new=$($LCTL get_param -n $testname) [[ "$jobid_var_new" == "$new_value" ]] || error "'set_param a b' failed on a value containing '='" - $LCTL set_param jobid_var $jobid_var_old - jobid_var_new=$($LCTL get_param -n jobid_var) + $LCTL set_param $testname $jobid_var_old + jobid_var_new=$($LCTL get_param -n $testname) [[ "$jobid_var_new" == "$jobid_var_old" ]] || - error "failed to reset jobid_var" + error "failed to reset $testname" } run_test 401d "Verify 'lctl set_param' accepts values containing '='" @@ -22027,6 +23218,7 @@ test_405() { skip "Layout swap lock is not supported" check_swap_layouts_support + check_swap_layout_no_dom $DIR test_mkdir $DIR/$tdir swap_lock_test -d $DIR/$tdir || @@ -22177,6 +23369,8 @@ test_410() { [[ $CLIENT_VERSION -lt $(version_code 2.9.59) ]] && skip "Need client version at least 2.9.59" + [ -f $LUSTRE/tests/kernel/kinode.ko ] || + skip "Need MODULES build" # Create a file, and stat it from the kernel local testfile=$DIR/$tfile @@ -22896,26 +24090,24 @@ test_421f() { cnt=$(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" || + tmpdir=$(mktemp -d /tmp/lustre-XXXXXX) + stack_trap "rmdir $tmpdir" + mount_client $tmpdir "$MOUNT_OPTS,user_fid2path" || error "failed to mount client'" + stack_trap "umount_client $tmpdir" - $RUNAS $LFS rmfid $DIR $FID || error "rmfid failed" + $RUNAS $LFS rmfid $tmpdir $FID || error "rmfid failed" # rmfid should succeed - cnt=$(ls -1 $DIR/$tdir | wc -l) + cnt=$(ls -1 $tmpdir/$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'" - + chmod a+rwx $tmpdir/$tdir + touch $tmpdir/$tdir/f + ls -la $tmpdir/$tdir + FID=$(lfs path2fid $tmpdir/$tdir/f) + $RUNAS $LFS rmfid $tmpdir $FID && error "rmfid didn't fail" + return 0 } run_test 421f "rmfid checks permissions" @@ -23024,6 +24216,354 @@ test_424() { } run_test 424 "simulate ENOMEM in ptl_send_rpc bulk reply ME attach" +test_425() { + test_mkdir -c -1 $DIR/$tdir + $LFS setstripe -c -1 $DIR/$tdir + + lru_resize_disable "" 100 + stack_trap "lru_resize_enable" EXIT + + sleep 5 + + for i in $(seq $((MDSCOUNT * 125))); do + local t=$DIR/$tdir/$tfile_$i + + dd if=/dev/zero of=$t bs=4K count=1 > /dev/null 2>&1 || + error_noexit "Create file $t" + done + stack_trap "rm -rf $DIR/$tdir" EXIT + + for oscparam in $($LCTL list_param ldlm.namespaces.*osc-[-0-9a-f]*); do + local lru_size=$($LCTL get_param -n $oscparam.lru_size) + local lock_count=$($LCTL get_param -n $oscparam.lock_count) + + [ $lock_count -le $lru_size ] || + error "osc lock count $lock_count > lru size $lru_size" + done + + for mdcparam in $($LCTL list_param ldlm.namespaces.*mdc-*); do + local lru_size=$($LCTL get_param -n $mdcparam.lru_size) + local lock_count=$($LCTL get_param -n $mdcparam.lock_count) + + [ $lock_count -le $lru_size ] || + error "mdc lock count $lock_count > lru size $lru_size" + done +} +run_test 425 "lock count should not exceed lru size" + +test_426() { + splice-test -r $DIR/$tfile + splice-test -rd $DIR/$tfile + splice-test $DIR/$tfile + splice-test -d $DIR/$tfile +} +run_test 426 "splice test on Lustre" + +test_427() { + [ $MDSCOUNT -ge 2 ] || skip "needs >= 2 MDTs" + (( $MDS1_VERSION >= $(version_code 2.12.4) )) || + skip "Need MDS version at least 2.12.4" + local log + + mkdir $DIR/$tdir + mkdir $DIR/$tdir/1 + mkdir $DIR/$tdir/2 + test_mkdir -c $MDSCOUNT -i 1 $DIR/$tdir/1/dir + test_mkdir -c $MDSCOUNT -i 1 $DIR/$tdir/2/dir2 + + $LFS getdirstripe $DIR/$tdir/1/dir + + #first setfattr for creating updatelog + setfattr -n user.attr0 -v "some text" $DIR/$tdir/1/dir + +#define OBD_FAIL_OUT_OBJECT_MISS 0x1708 + do_nodes $(comma_list $(mdts_nodes)) $LCTL set_param fail_loc=0x80001708 + setfattr -n user.attr1 -v "some text" $DIR/$tdir/1/dir & + setfattr -n user.attr2 -v "another attr" $DIR/$tdir/2/dir2 & + + sleep 2 + fail mds2 + wait_recovery_complete mds2 $((2*TIMEOUT)) + + log=$(do_facet mds1 dmesg | tac | sed "/${TESTNAME//_/ }/,$ d") + echo $log | grep "get update log failed" && + error "update log corruption is detected" || true +} +run_test 427 "Failed DNE2 update request shouldn't corrupt updatelog" + +test_428() { + [ $PARALLEL == "yes" ] && skip "skip parallel run" + local cache_limit=$CACHE_MAX + + stack_trap "$LCTL set_param -n llite.*.max_cached_mb=$cache_limit" + $LCTL set_param -n llite.*.max_cached_mb=64 + + mkdir $DIR/$tdir + $LFS setstripe -c 1 $DIR/$tdir + eval touch $DIR/$tdir/$tfile.{1..$OSTCOUNT} + stack_trap "rm -f $DIR/$tdir/$tfile.*" + #test write + for f in $(seq 4); do + dd if=/dev/zero of=$DIR/$tdir/$tfile.$f bs=128M count=1 & + done + wait + + cancel_lru_locks osc + # Test read + for f in $(seq 4); do + dd if=$DIR/$tdir/$tfile.$f of=/dev/null bs=128M count=1 & + done + wait +} +run_test 428 "large block size IO should not hang" + +lseek_test_430() { + local offset + local file=$1 + + # data at [200K, 400K) + dd if=/dev/urandom of=$file bs=256K count=1 seek=1 || + error "256K->512K dd fails" + # data at [2M, 3M) + dd if=/dev/urandom of=$file bs=1M count=1 seek=2 || + error "2M->3M dd fails" + # data at [4M, 5M) + dd if=/dev/urandom of=$file bs=1M count=1 seek=4 || + error "4M->5M dd fails" + echo "Data at 256K...512K, 2M...3M and 4M...5M" + # start at first component hole #1 + printf "Seeking hole from 1000 ... " + offset=$(lseek_test -l 1000 $file) + echo $offset + [[ $offset == 1000 ]] || error "offset $offset != 1000" + printf "Seeking data from 1000 ... " + offset=$(lseek_test -d 1000 $file) + echo $offset + [[ $offset == 262144 ]] || error "offset $offset != 262144" + + # start at first component data block + printf "Seeking hole from 300000 ... " + offset=$(lseek_test -l 300000 $file) + echo $offset + [[ $offset == 524288 ]] || error "offset $offset != 524288" + printf "Seeking data from 300000 ... " + offset=$(lseek_test -d 300000 $file) + echo $offset + [[ $offset == 300000 ]] || error "offset $offset != 300000" + + # start at the first component but beyond end of object size + printf "Seeking hole from 1000000 ... " + offset=$(lseek_test -l 1000000 $file) + echo $offset + [[ $offset == 1000000 ]] || error "offset $offset != 1000000" + printf "Seeking data from 1000000 ... " + offset=$(lseek_test -d 1000000 $file) + echo $offset + [[ $offset == 2097152 ]] || error "offset $offset != 2097152" + + # start at second component stripe 2 (empty file) + printf "Seeking hole from 1500000 ... " + offset=$(lseek_test -l 1500000 $file) + echo $offset + [[ $offset == 1500000 ]] || error "offset $offset != 1500000" + printf "Seeking data from 1500000 ... " + offset=$(lseek_test -d 1500000 $file) + echo $offset + [[ $offset == 2097152 ]] || error "offset $offset != 2097152" + + # start at second component stripe 1 (all data) + printf "Seeking hole from 3000000 ... " + offset=$(lseek_test -l 3000000 $file) + echo $offset + [[ $offset == 3145728 ]] || error "offset $offset != 3145728" + printf "Seeking data from 3000000 ... " + offset=$(lseek_test -d 3000000 $file) + echo $offset + [[ $offset == 3000000 ]] || error "offset $offset != 3000000" + + dd if=/dev/urandom of=$file bs=640K count=1 seek=1 || + error "2nd dd fails" + echo "Add data block at 640K...1280K" + + # start at before new data block, in hole + printf "Seeking hole from 600000 ... " + offset=$(lseek_test -l 600000 $file) + echo $offset + [[ $offset == 600000 ]] || error "offset $offset != 600000" + printf "Seeking data from 600000 ... " + offset=$(lseek_test -d 600000 $file) + echo $offset + [[ $offset == 655360 ]] || error "offset $offset != 655360" + + # start at the first component new data block + printf "Seeking hole from 1000000 ... " + offset=$(lseek_test -l 1000000 $file) + echo $offset + [[ $offset == 1310720 ]] || error "offset $offset != 1310720" + printf "Seeking data from 1000000 ... " + offset=$(lseek_test -d 1000000 $file) + echo $offset + [[ $offset == 1000000 ]] || error "offset $offset != 1000000" + + # start at second component stripe 2, new data + printf "Seeking hole from 1200000 ... " + offset=$(lseek_test -l 1200000 $file) + echo $offset + [[ $offset == 1310720 ]] || error "offset $offset != 1310720" + printf "Seeking data from 1200000 ... " + offset=$(lseek_test -d 1200000 $file) + echo $offset + [[ $offset == 1200000 ]] || error "offset $offset != 1200000" + + # start beyond file end + printf "Using offset > filesize ... " + lseek_test -l 4000000 $file && error "lseek should fail" + printf "Using offset > filesize ... " + lseek_test -d 4000000 $file && error "lseek should fail" + + printf "Done\n\n" +} + +test_430a() { + $LCTL get_param mdc.*.import | grep -q 'connect_flags:.*seek' || + skip "MDT does not support SEEK_HOLE" + + $LCTL get_param osc.*.import | grep -q 'connect_flags:.*seek' || + skip "OST does not support SEEK_HOLE" + + local file=$DIR/$tdir/$tfile + + mkdir -p $DIR/$tdir + + $LFS setstripe -E 1M -L mdt -E eof -c2 $file + # OST stripe #1 will have continuous data at [1M, 3M) + # OST stripe #2 is empty + echo "Component #1: 1M DoM, component #2: EOF, 2 stripes 1M" + lseek_test_430 $file + rm $file + $LFS setstripe -E 1M -c2 -S 64K -E 10M -c2 -S 1M $file + echo "Component #1: 1M, 2 stripes 64K, component #2: EOF, 2 stripes 1M" + lseek_test_430 $file + rm $file + $LFS setstripe -c2 -S 512K $file + echo "Two stripes, stripe size 512K" + lseek_test_430 $file + rm $file + # FLR with stale mirror + $LFS setstripe -N -E 512K -c1 -S 64K -E eof -c2 -S 512K \ + -N -c2 -S 1M $file + echo "Mirrored file:" + echo "Component #1: 512K, stripe 64K, component #2: EOF, 2 stripes 512K" + echo "Plain 2 stripes 1M" + lseek_test_430 $file + rm $file +} +run_test 430a "lseek: SEEK_DATA/SEEK_HOLE basic functionality" + +test_430b() { + $LCTL get_param osc.*.import | grep -q 'connect_flags:.*seek' || + skip "OST does not support SEEK_HOLE" + + local offset + local file=$DIR/$tdir/$tfile + + mkdir -p $DIR/$tdir + # Empty layout lseek should fail + $MCREATE $file + # seek from 0 + printf "Seeking hole from 0 ... " + lseek_test -l 0 $file && error "lseek should fail" + printf "Seeking data from 0 ... " + lseek_test -d 0 $file && error "lseek should fail" + rm $file + + # 1M-hole file + $LFS setstripe -E 1M -c2 -E eof $file + $TRUNCATE $file 1048576 + printf "Seeking hole from 1000000 ... " + offset=$(lseek_test -l 1000000 $file) + echo $offset + [[ $offset == 1000000 ]] || error "offset $offset != 1000000" + printf "Seeking data from 1000000 ... " + lseek_test -d 1000000 $file && error "lseek should fail" + rm $file + + # full component followed by non-inited one + $LFS setstripe -E 1M -c2 -E eof $file + dd if=/dev/urandom of=$file bs=1M count=1 + printf "Seeking hole from 1000000 ... " + offset=$(lseek_test -l 1000000 $file) + echo $offset + [[ $offset == 1048576 ]] || error "offset $offset != 1048576" + printf "Seeking hole from 1048576 ... " + lseek_test -l 1048576 $file && error "lseek should fail" + # init second component and truncate back + echo "123" >> $file + $TRUNCATE $file 1048576 + printf "Seeking hole from 1000000 ... " + offset=$(lseek_test -l 1000000 $file) + echo $offset + [[ $offset == 1048576 ]] || error "offset $offset != 1048576" + printf "Seeking hole from 1048576 ... " + lseek_test -l 1048576 $file && error "lseek should fail" + # boundary checks for big values + dd if=/dev/urandom of=$file.10g bs=1 count=1 seek=10G + offset=$(lseek_test -d 0 $file.10g) + [[ $offset == 10737418240 ]] || error "offset $offset != 10737418240" + dd if=/dev/urandom of=$file.100g bs=1 count=1 seek=100G + offset=$(lseek_test -d 0 $file.100g) + [[ $offset == 107374182400 ]] || error "offset $offset != 107374182400" + return 0 +} +run_test 430b "lseek: SEEK_DATA/SEEK_HOLE special cases" + +test_430c() { + $LCTL get_param osc.*.import | grep -q 'connect_flags:.*seek' || + skip "OST does not support SEEK_HOLE" + + local file=$DIR/$tdir/$tfile + local start + + mkdir -p $DIR/$tdir + dd if=/dev/urandom of=$file bs=1k count=1 seek=5M + + # cp version 8.33+ prefers lseek over fiemap + if [[ $(cp --version | head -n1 | sed "s/[^0-9]//g") -ge 833 ]]; then + start=$SECONDS + time cp $file /dev/null + (( SECONDS - start < 5 )) || + error "cp: too long runtime $((SECONDS - start))" + + fi + # tar version 1.29+ supports SEEK_HOLE/DATA + if [[ $(tar --version | head -n1 | sed "s/[^0-9]//g") -ge 129 ]]; then + start=$SECONDS + time tar cS $file - | cat > /dev/null + (( SECONDS - start < 5 )) || + error "tar: too long runtime $((SECONDS - start))" + fi +} +run_test 430c "lseek: external tools check" + +test_431() { # LU-14187 + local file=$DIR/$tdir/$tfile + + mkdir -p $DIR/$tdir + $LFS setstripe -c 1 -i 0 $file || error "lfs setstripe failed" + dd if=/dev/urandom of=$file bs=4k count=1 + dd if=/dev/urandom of=$file bs=4k count=1 seek=10 conv=notrunc + dd if=/dev/urandom of=$file bs=4k count=1 seek=12 conv=notrunc + #define OBD_FAIL_OST_RESTART_IO 0x251 + do_facet ost1 "$LCTL set_param fail_loc=0x251" + $LFS setstripe -c 1 -i 0 $file.0 || error "lfs setstripe failed" + cp $file $file.0 + cancel_lru_locks + sync_all_data + echo 3 > /proc/sys/vm/drop_caches + diff $file $file.0 || error "data diff" +} +run_test 431 "Restart transaction for IO" + prep_801() { [[ $MDS1_VERSION -lt $(version_code 2.9.55) ]] || [[ $OST1_VERSION -lt $(version_code 2.9.55) ]] && @@ -23332,7 +24872,7 @@ test_802b() { } run_test 802b "be able to set MDTs to readonly" -test_803() { +test_803a() { [[ $MDSCOUNT -lt 2 ]] && skip_env "needs >= 2 MDTs" [ $MDS1_VERSION -lt $(version_code 2.10.54) ] && skip "MDS needs to be newer than 2.10.54" @@ -23380,7 +24920,39 @@ test_803() { [ $after_used -le $((before_used + 1)) ] || error "after ($after_used) > before ($before_used) + 1" } -run_test 803 "verify agent object for remote object" +run_test 803a "verify agent object for remote object" + +test_803b() { + [[ $MDSCOUNT -lt 2 ]] && skip_env "needs >= 2 MDTs" + [ $MDS1_VERSION -lt $(version_code 2.13.56) ] && + skip "MDS needs to be newer than 2.13.56" + [ $PARALLEL == "yes" ] && skip "skip parallel run" + + for i in $(seq 0 $((MDSCOUNT - 1))); do + $LFS mkdir -i $i $DIR/$tdir.$i || error "mkdir $tdir.$i" + done + + local before=0 + local after=0 + + local tmp + + stat $DIR/$tdir.* >/dev/null || error "stat $tdir.*" + for i in $(seq 0 $((MDSCOUNT - 1))); do + tmp=$(do_facet mds$i $LCTL get_param mdt.*-MDT000$i.md_stats | + awk '/getattr/ { print $2 }') + before=$((before + tmp)) + done + stat $DIR/$tdir.* >/dev/null || error "stat $tdir.*" + for i in $(seq 0 $((MDSCOUNT - 1))); do + tmp=$(do_facet mds$i $LCTL get_param mdt.*-MDT000$i.md_stats | + awk '/getattr/ { print $2 }') + after=$((after + tmp)) + done + + [ $before -eq $after ] || error "getattr count $before != $after" +} +run_test 803b "remote object can getattr from cache" test_804() { [[ $MDSCOUNT -lt 2 ]] && skip_env "needs >= 2 MDTs" @@ -23779,8 +25351,6 @@ run_test 810 "partial page writes on ZFS (LU-11663)" test_812a() { [ $OST1_VERSION -lt $(version_code 2.12.51) ] && skip "OST < 2.12.51 doesn't support this fail_loc" - [ "$SHARED_KEY" = true ] && - skip "OSC connections never go IDLE with Shared-Keys enabled" $LFS setstripe -c 1 -i 0 $DIR/$tfile # ensure ost1 is connected @@ -23802,8 +25372,6 @@ 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 @@ -24011,11 +25579,13 @@ test_815() run_test 815 "zero byte tiny write doesn't hang (LU-12382)" test_816() { - [ "$SHARED_KEY" = true ] && - skip "OSC connections never go IDLE with Shared-Keys enabled" + local ost1_imp=$(get_osc_import_name client ost1) + local imp_name=$($LCTL list_param osc.$ost1_imp | head -n1 | + cut -d'.' -f2) $LFS setstripe -c 1 -i 0 $DIR/$tfile # ensure ost1 is connected + stat $DIR/$tfile >/dev/null || error "can't stat" wait_osc_import_state client ost1 FULL # no locks, no reqs to let the connection idle @@ -24024,12 +25594,12 @@ test_816() { local before local now before=$($LCTL get_param -n \ - ldlm.namespaces.$FSNAME-OST0000-osc-[^M]*.lru_size) + ldlm.namespaces.$imp_name.lru_size) wait_osc_import_state client ost1 IDLE dd if=/dev/null of=$DIR/$tfile bs=1k count=1 conv=sync now=$($LCTL get_param -n \ - ldlm.namespaces.$FSNAME-OST0000-osc-[^M]*.lru_size) + ldlm.namespaces.$imp_name.lru_size) [ $before == $now ] || error "lru_size changed $before != $now" } run_test 816 "do not reset lru_resize on idle reconnect" @@ -24122,11 +25692,49 @@ test_820() { # open intent should update default EA size # see mdc_update_max_ea_from_body() # notice this is the very first RPC to MDS2 - cp /etc/services $DIR/$tdir/mds2 || - error "Failed to copy files to mds$n" + out=$(cp /etc/services $DIR/$tdir/mds2 2>&1) + ret=$? + echo $out + # With SSK, this situation can lead to -EPERM being returned. + # In that case, simply retry. + if [ $ret -ne 0 ] && $SHARED_KEY; then + if echo "$out" | grep -q "not permitted"; then + cp /etc/services $DIR/$tdir/mds2 + ret=$? + fi + fi + [ $ret -eq 0 ] || error "Failed to copy files to mds$n" } run_test 820 "update max EA from open intent" +test_822() { + local p="$TMP/$TESTSUITE-$TESTNAME.parameters" + + save_lustre_params mds1 \ + "osp.$FSNAME-OST*-osc-MDT0000.max_create_count" > $p + do_facet $SINGLEMDS "$LCTL set_param -n \ + osp.$FSNAME-OST*MDT0000.max_create_count=0" + do_facet $SINGLEMDS "$LCTL set_param -n \ + osp.$FSNAME-OST0000*MDT0000.max_create_count=20000" + + # wait for statfs update to clear OS_STATFS_NOPRECREATE + local maxage=$(do_facet mds1 $LCTL get_param -n \ + osp.$FSNAME-OST0000*MDT0000.maxage) + sleep $((maxage + 1)) + + #define OBD_FAIL_NET_ERROR_RPC 0x532 + do_facet mds1 "$LCTL set_param fail_loc=0x80000532 fail_val=5" + + stack_trap "restore_lustre_params < $p; rm $p" + + local count=$(do_facet $SINGLEMDS "lctl get_param -n \ + osp.$FSNAME-OST0000*MDT0000.create_count") + for i in $(seq 1 $count); do + touch $DIR/$tfile.${i} || error "touch failed" + done +} +run_test 822 "test precreate failure" + # # tests that do cleanup/setup should be run at the end #