X-Git-Url: https://git.whamcloud.com/?p=fs%2Flustre-release.git;a=blobdiff_plain;f=lustre%2Ftests%2Fsanity.sh;h=a92bf03ee6d77df0c6403b88cd1fb9e7f8a141f3;hp=61e6b0ddf11df9cc8bf3a5c24ecc75569f932f40;hb=f602b5ec7f45713122abd615a97a13d7c97d460e;hpb=305548c41a00e189262abdd27fa00f5f1dd5a16a diff --git a/lustre/tests/sanity.sh b/lustre/tests/sanity.sh index 61e6b0d..a92bf03 100755 --- a/lustre/tests/sanity.sh +++ b/lustre/tests/sanity.sh @@ -8,16 +8,20 @@ set -e ONLY=${ONLY:-"$*"} -# bug number for skipped test: LU-9693 LU-6493 LU-9693 -ALWAYS_EXCEPT="$SANITY_EXCEPT 42a 42b 42c" +# bug number for skipped test: LU-9693 LU-6493 LU-9693 LU-11058 +ALWAYS_EXCEPT="$SANITY_EXCEPT 42a 42b 42c 77k" # UPDATE THE COMMENT ABOVE WITH BUG NUMBERS WHEN CHANGING ALWAYS_EXCEPT! -# skipped tests: LU-8411 LU-9096 LU-9054 LU-10680 .. -ALWAYS_EXCEPT=" 407 253 312 160f 160g $ALWAYS_EXCEPT" +# skipped tests: LU-8411 LU-9096 LU-9054 .. +ALWAYS_EXCEPT=" 407 253 312 $ALWAYS_EXCEPT" + +if $SHARED_KEY; then +# bug number for skipped tests: LU-9795 (all below) + ALWAYS_EXCEPT="$ALWAYS_EXCEPT 17n 60a 133g 300f" +fi # Check Grants after these tests GRANT_CHECK_LIST="$GRANT_CHECK_LIST 42a 42b 42c 42d 42e 63a 63b 64a 64b 64c" - SRCDIR=$(cd $(dirname $0); echo $PWD) export PATH=$PATH:/sbin @@ -59,6 +63,11 @@ init_test_env $@ . ${CONFIG:=$LUSTRE/tests/cfg/${NAME}.sh} init_logging +if [[ $MDSCOUNT -gt 1 ]]; then + # bug number: LU-11161 + ALWAYS_EXCEPT+=" 160g" +fi + # 5 12 (min)" [ "$SLOW" = "no" ] && EXCEPT_SLOW="27m 64b 68 71 115 300o" @@ -69,11 +78,10 @@ if [ $(facet_fstype $SINGLEMDS) = "zfs" ]; then [ "$SLOW" = "no" ] && EXCEPT_SLOW="$EXCEPT_SLOW 51b" fi -# Get the SLES version so we can make decisions on if a test should be run +# Get the SLES distro version # # Returns a version string that should only be used in comparing # strings returned by version_code() - sles_version_code() { local version=$(grep VERSION_ID /etc/os-release | cut -d'"' -f2) @@ -83,6 +91,8 @@ sles_version_code() version_code $sles_version } +# Check if we are running on Ubuntu or SLES so we can make decisions on +# what tests to run if [ -r /etc/SuSE-release ]; then sles_version=$(sles_version_code) [ $sles_version -lt $(version_code 11.4.0) ] && @@ -91,6 +101,21 @@ if [ -r /etc/SuSE-release ]; then [ $sles_version -lt $(version_code 12.0.0) ] && # bug number for skipped test: LU-3703 ALWAYS_EXCEPT="$ALWAYS_EXCEPT 234" +elif [ -r /etc/os-release ]; then + if grep -qi ubuntu /etc/os-release; then + ubuntu_version=$(version_code $(sed -n -e 's/"//g' \ + -e 's/^VERSION=//p' \ + /etc/os-release | + awk '{ print $1 }')) + + if [[ $ubuntu_version -gt $(version_code 16.0.0) ]]; then + # bug number for skipped test: + # LU-10334 LU-10335 LU-10335 LU-10335 + ALWAYS_EXCEPT+=" 103a 130a 130b 130c" + # LU-10335 LU-10335 LU-10366 + ALWAYS_EXCEPT+=" 130d 130e 410" + fi + fi fi FAIL_ON_ERROR=false @@ -110,8 +135,7 @@ setup() { check_swap_layouts_support() { $LCTL get_param -n llite.*.sbi_flags | grep -q layout || - { skip "Does not support layout lock."; return 0; } - return 1 + skip "Does not support layout lock." } check_and_setup_lustre @@ -170,7 +194,7 @@ run_test 0c "check import proc" test_0d() { # LU-3397 [ $(lustre_version_code mgs) -lt $(version_code 2.10.57) ] && - skip "proc exports not supported before 2.10.57" && return + skip "proc exports not supported before 2.10.57" local mgs_exp="mgs.MGS.exports" local client_uuid=$($LCTL get_param -n mgc.*.uuid) @@ -185,7 +209,7 @@ test_0d() { # LU-3397 $LCTL get_param mgc.*.import | tee $temp_imp # Check if client uuid is found in MGS export for exp_client_nid in $(do_facet mgs $LCTL get_param -N $mgs_exp.*); do - [ $(do_facet mgs $LCTL get_param $exp_client_nid.uuid) == \ + [ $(do_facet mgs $LCTL get_param -n $exp_client_nid.uuid) == \ $client_uuid ] && break; done @@ -274,7 +298,8 @@ test_6a() { run_test 6a "touch f6a; chmod f6a; $RUNAS chmod f6a (should return error) ==" test_6c() { - [ $RUNAS_ID -eq $UID ] && skip_env "RUNAS_ID = UID = $UID" && return + [ $RUNAS_ID -eq $UID ] && skip_env "RUNAS_ID = UID = $UID" + touch $DIR/$tfile chown $RUNAS_ID $DIR/$tfile || error "chown $RUNAS_ID $file failed" $CHECKSTAT -t file -u \#$RUNAS_ID $DIR/$tfile || @@ -286,7 +311,8 @@ test_6c() { run_test 6c "touch f6c; chown f6c; $RUNAS chown f6c (should return error) ==" test_6e() { - [ $RUNAS_ID -eq $UID ] && skip_env "RUNAS_ID = UID = $UID" && return + [ $RUNAS_ID -eq $UID ] && skip_env "RUNAS_ID = UID = $UID" + touch $DIR/$tfile chgrp $RUNAS_ID $DIR/$tfile || error "chgrp $RUNAS_ID $file failed" $CHECKSTAT -t file -u \#$UID -g \#$RUNAS_ID $DIR/$tfile || @@ -298,7 +324,8 @@ test_6e() { run_test 6e "touch+chgrp $tfile; $RUNAS chgrp $tfile (should return error)" test_6g() { - [ $RUNAS_ID -eq $UID ] && skip_env "RUNAS_ID = UID = $UID" && return + [ $RUNAS_ID -eq $UID ] && skip_env "RUNAS_ID = UID = $UID" + test_mkdir $DIR/$tdir chmod 777 $DIR/$tdir || error "chmod 0777 $tdir failed" $RUNAS mkdir $DIR/$tdir/d || error "mkdir $tdir/d failed" @@ -325,7 +352,8 @@ test_6g() { run_test 6g "verify new dir in sgid dir inherits group" test_6h() { # bug 7331 - [ $RUNAS_ID -eq $UID ] && skip_env "RUNAS_ID = UID = $UID" && return + [ $RUNAS_ID -eq $UID ] && skip_env "RUNAS_ID = UID = $UID" + touch $DIR/$tfile || error "touch failed" chown $RUNAS_ID:$RUNAS_GID $DIR/$tfile || error "initial chown failed" $RUNAS -G$RUNAS_GID chown $RUNAS_ID:0 $DIR/$tfile && @@ -537,9 +565,11 @@ test_17g() { run_test 17g "symlinks: really long symlink name and inode boundaries" test_17h() { #bug 17378 - remote_mds_nodsh && skip "remote MDS with nodsh" && return - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return + [ $PARALLEL == "yes" ] && skip "skip parallel run" + remote_mds_nodsh && skip "remote MDS with nodsh" + local mdt_idx + test_mkdir $DIR/$tdir mdt_idx=$($LFS getdirstripe -i $DIR/$tdir) $LFS setstripe -c -1 $DIR/$tdir @@ -550,11 +580,13 @@ test_17h() { #bug 17378 run_test 17h "create objects: lov_free_memmd() doesn't lbug" test_17i() { #bug 20018 - remote_mds_nodsh && skip "remote MDS with nodsh" && return - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return - test_mkdir -c1 $DIR/$tdir + [ $PARALLEL == "yes" ] && skip "skip parallel run" + remote_mds_nodsh && skip "remote MDS with nodsh" + local foo=$DIR/$tdir/$tfile local mdt_idx + + test_mkdir -c1 $DIR/$tdir mdt_idx=$($LFS getdirstripe -i $DIR/$tdir) ln -s $foo $foo || error "create symlink failed" #define OBD_FAIL_MDS_READLINK_EPROTO 0x143 @@ -565,11 +597,11 @@ test_17i() { #bug 20018 run_test 17i "don't panic on short symlink (should return error)" test_17k() { #bug 22301 + [ $PARALLEL == "yes" ] && skip "skip parallel run" [[ -z "$(which rsync 2>/dev/null)" ]] && - skip "no rsync command" && return 0 + skip "no rsync command" rsync --help | grep -q xattr || skip_env "$(rsync --version | head -n1) does not support xattrs" - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return 0 test_mkdir $DIR/$tdir test_mkdir $DIR/$tdir.new touch $DIR/$tdir/$tfile @@ -581,7 +613,8 @@ run_test 17k "symlinks: rsync with xattrs enabled" test_17l() { # LU-279 [[ -z "$(which getfattr 2>/dev/null)" ]] && - skip "no getfattr command" && return 0 + skip "no getfattr command" + test_mkdir $DIR/$tdir touch $DIR/$tdir/$tfile ln -s $DIR/$tdir/$tfile $DIR/$tdir/$tfile.lnk @@ -599,19 +632,17 @@ run_test 17l "Ensure lgetxattr's returned xattr size is consistent" # LU-1540 test_17m() { - local short_sym="0123456789" - local wdir=$DIR/$tdir - local i - - remote_mds_nodsh && skip "remote MDS with nodsh" && return + [ $PARALLEL == "yes" ] && skip "skip parallel run" + [ "$(facet_fstype $SINGLEMDS)" != "ldiskfs" ] && + skip_env "ldiskfs only test" + remote_mds_nodsh && skip "remote MDS with nodsh" [ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.2.0) ] && [ $(lustre_version_code $SINGLEMDS) -le $(version_code 2.2.93) ] && - skip "MDS 2.2.0-2.2.93 do not NUL-terminate symlinks" && return + skip "MDS 2.2.0-2.2.93 do not NUL-terminate symlinks" - [ "$(facet_fstype $SINGLEMDS)" != "ldiskfs" ] && - skip "ldiskfs only test" && return 0 - - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return + local short_sym="0123456789" + local wdir=$DIR/$tdir + local i test_mkdir $wdir long_sym=$short_sym @@ -638,7 +669,7 @@ test_17m() { ln -sf ${short_sym} $wdir/short-$i || error "short_sym failed" done - local mds_index=$(($($LFS getstripe -M $wdir) + 1)) + local mds_index=$(($($LFS getstripe -m $wdir) + 1)) local devname=$(mdsdevname $mds_index) echo "stop and checking mds${mds_index}:" @@ -652,6 +683,7 @@ test_17m() { df $MOUNT > /dev/null 2>&1 [ $rc -eq 0 ] || error "e2fsck detected error for short/long symlink: rc=$rc" + rm -f $wdir/* } run_test 17m "run e2fsck against MDT which contains short/long symlink" @@ -676,19 +708,16 @@ check_fs_consistency_17n() { } test_17n() { - local i - - remote_mds_nodsh && skip "remote MDS with nodsh" && return + [[ $MDSCOUNT -lt 2 ]] && skip_env "needs >= 2 MDTs" + [ $PARALLEL == "yes" ] && skip "skip parallel run" + [ "$(facet_fstype $SINGLEMDS)" != "ldiskfs" ] && + skip_env "ldiskfs only test" + remote_mds_nodsh && skip "remote MDS with nodsh" [ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.2.0) ] && [ $(lustre_version_code $SINGLEMDS) -le $(version_code 2.2.93) ] && - skip "MDS 2.2.0-2.2.93 do not NUL-terminate symlinks" && return - - [ "$(facet_fstype $SINGLEMDS)" != "ldiskfs" ] && - skip "ldiskfs only test" && return 0 + skip "MDS 2.2.0-2.2.93 do not NUL-terminate symlinks" - [[ $MDSCOUNT -lt 2 ]] && skip "needs >= 2 MDTs" && return - - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return + local i test_mkdir $DIR/$tdir for ((i=0; i<10; i++)); do @@ -710,7 +739,7 @@ test_17n() { error "e2fsck report error after unlink files under remote dir" [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.4.50) ] && - skip "lustre < 2.4.50 does not support migrate mv " && return + skip "lustre < 2.4.50 does not support migrate mv" for ((i = 0; i < 10; i++)); do mkdir -p $DIR/$tdir/remote_dir_${i} @@ -731,9 +760,9 @@ test_17n() { run_test 17n "run e2fsck against master/slave MDT which contains remote dir" test_17o() { - remote_mds_nodsh && skip "remote MDS with nodsh" && return + remote_mds_nodsh && skip "remote MDS with nodsh" [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.3.64) ] && - skip "Need MDS version at least 2.3.64" && return + skip "Need MDS version at least 2.3.64" local wdir=$DIR/${tdir}o local mdt_index @@ -741,7 +770,7 @@ test_17o() { test_mkdir $wdir touch $wdir/$tfile - mdt_index=$($LFS getstripe -M $wdir/$tfile) + mdt_index=$($LFS getstripe -m $wdir/$tfile) mdt_index=$((mdt_index + 1)) cancel_lru_locks mdc @@ -778,7 +807,8 @@ run_test 19b "ls -l .../f19 (should return error) ==============" test_19c() { [ $RUNAS_ID -eq $UID ] && - skip_env "RUNAS_ID = UID = $UID -- skipping" && return + skip_env "RUNAS_ID = UID = $UID -- skipping" + $RUNAS touch $DIR/$tfile && error "create non-root file failed" || true } run_test 19c "$RUNAS touch .../f19 (should return error) ==" @@ -880,6 +910,19 @@ test_23c() { } run_test 23c "O_APPEND size checks for tiny writes" +# LU-11069 file offset is correct after appending writes +test_23d() { + local file=$DIR/$tfile + local offset + + echo CentaurHauls > $file + offset=$($MULTIOP $file oO_WRONLY:O_APPEND:w13Zp) + if ((offset != 26)); then + error "wrong offset, expected 26, got '$offset'" + fi +} +run_test 23d "file offset is correct after appending writes" + # rename sanity test_24a() { echo '-- same directory rename' @@ -1043,7 +1086,8 @@ cleanup_multiop_pause() { } test_24q() { - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return + [ $PARALLEL == "yes" ] && skip "skip parallel run" + test_mkdir $DIR/R13a test_mkdir $DIR/R13b local DIRINO=$(ls -lid $DIR/R13a | awk '{ print $1 }') @@ -1095,12 +1139,6 @@ test_24u() { # bug12192 } run_test 24u "create stripe file" -page_size() { - local size - size=$(getconf PAGE_SIZE 2>/dev/null) - echo -n ${size:-4096} -} - simple_cleanup_common() { local rc=0 trap 0 @@ -1120,12 +1158,14 @@ max_pages_per_rpc() { } test_24v() { - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return + [ $PARALLEL == "yes" ] && skip "skip parallel run" + local nrfiles=${COUNT:-100000} + local fname="$DIR/$tdir/$tfile" + # Performance issue on ZFS see LU-4072 (c.f. LU-2887) [ $(facet_fstype $SINGLEMDS) = "zfs" ] && nrfiles=${COUNT:-10000} - local fname="$DIR/$tdir/$tfile" test_mkdir "$(dirname $fname)" # assume MDT0000 has the fewest inodes local stripes=$($LFS getdirstripe -c $(dirname $fname)) @@ -1156,7 +1196,7 @@ test_24v() { # take into account of overhead in lu_dirpage header and end mark in # each page, plus one in rpc_num calculation. local dirent_size=$((32 + (${#tfile} | 7) + 1 + 8)) - local page_entries=$((($(page_size) - 24) / dirent_size)) + local page_entries=$(((PAGE_SIZE - 24) / dirent_size)) local mdt_idx=$($LFS getdirstripe -i $(dirname $fname)) local rpc_pages=$(max_pages_per_rpc $mdt_idx) local rpc_max=$((nrfiles / (page_entries * rpc_pages) + stripes)) @@ -1182,12 +1222,11 @@ test_24w() { # bug21506 run_test 24w "Reading a file larger than 4Gb" test_24x() { - [[ $MDSCOUNT -lt 2 ]] && skip "needs >= 2 MDTs" && return - + [[ $MDSCOUNT -lt 2 ]] && skip_env "needs >= 2 MDTs" + [ $PARALLEL == "yes" ] && skip "skip parallel run" [[ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.7.56) ]] && - skip "Need MDS version at least 2.7.56" && return + skip "Need MDS version at least 2.7.56" - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return local MDTIDX=1 local remote_dir=$DIR/$tdir/remote_dir @@ -1215,14 +1254,15 @@ test_24x() { run_test 24x "cross MDT rename/link" test_24y() { - [[ $MDSCOUNT -lt 2 ]] && skip "needs >= 2 MDTs" && return - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return + [[ $MDSCOUNT -lt 2 ]] && skip_env "needs >= 2 MDTs" + [ $PARALLEL == "yes" ] && skip "skip parallel run" + local remote_dir=$DIR/$tdir/remote_dir local mdtidx=1 test_mkdir $DIR/$tdir $LFS mkdir -i $mdtidx $remote_dir || - error "create remote directory failed" + error "create remote directory failed" test_mkdir $remote_dir/src_dir touch $remote_dir/src_file @@ -1261,7 +1301,8 @@ test_24A() { # LU-3182 run_test 24A "readdir() returns correct number of entries." test_24B() { # LU-4805 - [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return + [ $MDSCOUNT -lt 2 ] && skip_env "needs >= 2 MDTs" + local count test_mkdir $DIR/$tdir @@ -1286,7 +1327,7 @@ test_24B() { # LU-4805 run_test 24B "readdir for striped dir return correct number of entries" test_24C() { - [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return + [ $MDSCOUNT -lt 2 ] && skip_env "needs >= 2 MDTs" mkdir $DIR/$tdir mkdir $DIR/$tdir/d0 @@ -1315,8 +1356,8 @@ test_24C() { run_test 24C "check .. in striped dir" test_24E() { - [[ $MDSCOUNT -lt 4 ]] && skip "needs >= 4 MDTs" && return - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return + [[ $MDSCOUNT -lt 4 ]] && skip_env "needs >= 4 MDTs" + [ $PARALLEL == "yes" ] && skip "skip parallel run" mkdir -p $DIR/$tdir mkdir $DIR/$tdir/src_dir @@ -1427,7 +1468,8 @@ test_27a() { run_test 27a "one stripe file" test_27b() { - [[ $OSTCOUNT -lt 2 ]] && skip_env "needs >= 2 OSTs" && return + [[ $OSTCOUNT -lt 2 ]] && skip_env "needs >= 2 OSTs" + test_mkdir $DIR/$tdir $LFS setstripe -c 2 $DIR/$tdir/$tfile || error "setstripe failed" $LFS getstripe -c $DIR/$tdir/$tfile @@ -1450,7 +1492,8 @@ run_test 27d "create file with default settings" test_27e() { # LU-5839 adds check for existed layout before setting it [[ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.7.56) ]] && - skip "Need MDS version at least 2.7.56" && return + skip "Need MDS version at least 2.7.56" + test_mkdir $DIR/$tdir $LFS setstripe -c 2 $DIR/$tdir/$tfile || error "setstripe failed" $LFS setstripe -c 2 $DIR/$tdir/$tfile && error "setstripe worked twice" @@ -1513,13 +1556,12 @@ test_27l() { run_test 27l "check setstripe permissions (should return error)" test_27m() { - [[ $OSTCOUNT -lt 2 ]] && skip_env "needs >= 2 OSTs" && return + [[ $OSTCOUNT -lt 2 ]] && skip_env "needs >= 2 OSTs" ORIGFREE=$($LCTL get_param -n lov.$FSNAME-clilov-*.kbytesavail | head -n1) if [[ $ORIGFREE -gt $MAXFREE ]]; then skip "$ORIGFREE > $MAXFREE skipping out-of-space test on OST0" - return fi trap simple_cleanup_common EXIT test_mkdir $DIR/$tdir @@ -1571,7 +1613,7 @@ exhaust_precreations() { local ofacet=ost$((OSTIDX + 1)) test_mkdir -p -c1 $DIR/$tdir - local mdtidx=$($LFS getstripe -M $DIR/$tdir) + local mdtidx=$($LFS getstripe -m $DIR/$tdir) local mfacet=mds$((mdtidx + 1)) echo OSTIDX=$OSTIDX MDTIDX=$mdtidx @@ -1606,10 +1648,10 @@ exhaust_all_precreations() { } test_27n() { - [[ $OSTCOUNT -lt 2 ]] && skip_env "needs >= 2 OSTs" && return - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return - remote_mds_nodsh && skip "remote MDS with nodsh" && return - remote_ost_nodsh && skip "remote OST with nodsh" && return + [[ $OSTCOUNT -lt 2 ]] && skip_env "needs >= 2 OSTs" + [ $PARALLEL == "yes" ] && skip "skip parallel run" + remote_mds_nodsh && skip "remote MDS with nodsh" + remote_ost_nodsh && skip "remote OST with nodsh" reset_enospc rm -f $DIR/$tdir/$tfile @@ -1622,10 +1664,10 @@ test_27n() { run_test 27n "create file with some full OSTs" test_27o() { - [[ $OSTCOUNT -lt 2 ]] && skip_env "needs >= 2 OSTs" && return - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return - remote_mds_nodsh && skip "remote MDS with nodsh" && return - remote_ost_nodsh && skip "remote OST with nodsh" && return + [[ $OSTCOUNT -lt 2 ]] && skip_env "needs >= 2 OSTs" + [ $PARALLEL == "yes" ] && skip "skip parallel run" + remote_mds_nodsh && skip "remote MDS with nodsh" + remote_ost_nodsh && skip "remote OST with nodsh" reset_enospc rm -f $DIR/$tdir/$tfile @@ -1639,10 +1681,10 @@ test_27o() { run_test 27o "create file with all full OSTs (should error)" test_27p() { - [[ $OSTCOUNT -lt 2 ]] && skip_env "needs >= 2 OSTs" && return - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return - remote_mds_nodsh && skip "remote MDS with nodsh" && return - remote_ost_nodsh && skip "remote OST with nodsh" && return + [[ $OSTCOUNT -lt 2 ]] && skip_env "needs >= 2 OSTs" + [ $PARALLEL == "yes" ] && skip "skip parallel run" + remote_mds_nodsh && skip "remote MDS with nodsh" + remote_ost_nodsh && skip "remote OST with nodsh" reset_enospc rm -f $DIR/$tdir/$tfile @@ -1662,10 +1704,10 @@ test_27p() { run_test 27p "append to a truncated file with some full OSTs" test_27q() { - [[ $OSTCOUNT -lt 2 ]] && skip_env "needs >= 2 OSTs" && return - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return - remote_mds_nodsh && skip "remote MDS with nodsh" && return - remote_ost_nodsh && skip "remote OST with nodsh" && return + [[ $OSTCOUNT -lt 2 ]] && skip_env "needs >= 2 OSTs" + [ $PARALLEL == "yes" ] && skip "skip parallel run" + remote_mds_nodsh && skip "remote MDS with nodsh" + remote_ost_nodsh && skip "remote OST with nodsh" reset_enospc rm -f $DIR/$tdir/$tfile @@ -1686,10 +1728,10 @@ test_27q() { run_test 27q "append to truncated file with all OSTs full (should error)" test_27r() { - [[ $OSTCOUNT -lt 2 ]] && skip_env "needs >= 2 OSTs" && return - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return - remote_mds_nodsh && skip "remote MDS with nodsh" && return - remote_ost_nodsh && skip "remote OST with nodsh" && return + [[ $OSTCOUNT -lt 2 ]] && skip_env "needs >= 2 OSTs" + [ $PARALLEL == "yes" ] && skip "skip parallel run" + remote_mds_nodsh && skip "remote MDS with nodsh" + remote_ost_nodsh && skip "remote OST with nodsh" reset_enospc rm -f $DIR/$tdir/$tfile @@ -1723,8 +1765,9 @@ test_27t() { # bug 10864 run_test 27t "check that utils parse path correctly" test_27u() { # bug 4900 - [[ $OSTCOUNT -lt 2 ]] && skip_env "needs >= 2 OSTs" && return - remote_mds_nodsh && skip "remote MDS with nodsh" && return + [[ $OSTCOUNT -lt 2 ]] && skip_env "needs >= 2 OSTs" + remote_mds_nodsh && skip "remote MDS with nodsh" + local index local list=$(comma_list $(mdts_nodes)) @@ -1747,10 +1790,10 @@ test_27u() { # bug 4900 run_test 27u "skip object creation on OSC w/o objects" test_27v() { # bug 4900 - [[ $OSTCOUNT -lt 2 ]] && skip_env "needs >= 2 OSTs" && return - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return - remote_mds_nodsh && skip "remote MDS with nodsh" && return - remote_ost_nodsh && skip "remote OST with nodsh" && return + [[ $OSTCOUNT -lt 2 ]] && skip_env "needs >= 2 OSTs" + [ $PARALLEL == "yes" ] && skip "skip parallel run" + remote_mds_nodsh && skip "remote MDS with nodsh" + remote_ost_nodsh && skip "remote OST with nodsh" exhaust_all_precreations 0x215 reset_enospc @@ -1788,7 +1831,7 @@ run_test 27w "check $LFS setstripe -S and getstrip -d options" test_27wa() { [[ $OSTCOUNT -lt 2 ]] && - skip_env "skipping multiple stripe count/offset test" && return + skip_env "skipping multiple stripe count/offset test" test_mkdir $DIR/$tdir for i in $(seq 1 $OSTCOUNT); do @@ -1805,9 +1848,10 @@ test_27wa() { run_test 27wa "check $LFS setstripe -c -i options" test_27x() { - remote_ost_nodsh && skip "remote OST with nodsh" && return - [[ $OSTCOUNT -lt 2 ]] && skip_env "needs >= 2 OSTs" && return - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return + remote_ost_nodsh && skip "remote OST with nodsh" + [[ $OSTCOUNT -lt 2 ]] && skip_env "needs >= 2 OSTs" + [ $PARALLEL == "yes" ] && skip "skip parallel run" + OFFSET=$(($OSTCOUNT - 1)) OSTIDX=0 local OST=$(ostname_from_index $OSTIDX) @@ -1827,10 +1871,10 @@ test_27x() { run_test 27x "create files while OST0 is degraded" test_27y() { - [[ $OSTCOUNT -lt 2 ]] && skip_env "needs >= 2 OSTs" && return - remote_mds_nodsh && skip "remote MDS with nodsh" && return - remote_ost_nodsh && skip "remote OST with nodsh" && return - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return + [[ $OSTCOUNT -lt 2 ]] && skip_env "needs >= 2 OSTs" + remote_mds_nodsh && skip "remote MDS with nodsh" + remote_ost_nodsh && skip "remote OST with nodsh" + [ $PARALLEL == "yes" ] && skip "skip parallel run" local mdtosc=$(get_mdtosc_proc_path $SINGLEMDS $FSNAME-OST0000) local last_id=$(do_facet $SINGLEMDS lctl get_param -n \ @@ -1838,7 +1882,7 @@ test_27y() { local next_id=$(do_facet $SINGLEMDS lctl get_param -n \ osc.$mdtosc.prealloc_next_id) local fcount=$((last_id - next_id)) - [[ $fcount -eq 0 ]] && skip "not enough space on OST0" && return + [[ $fcount -eq 0 ]] && skip "not enough space on OST0" [[ $fcount -gt $OSTCOUNT ]] && fcount=$OSTCOUNT local MDS_OSCS=$(do_facet $SINGLEMDS lctl dl | @@ -2013,10 +2057,10 @@ check_seq_oid() } test_27z() { - remote_ost_nodsh && skip "remote OST with nodsh" && return - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return - test_mkdir $DIR/$tdir + [ $PARALLEL == "yes" ] && skip "skip parallel run" + remote_ost_nodsh && skip "remote OST with nodsh" + test_mkdir $DIR/$tdir $LFS setstripe -c 1 -i 0 -S 64k $DIR/$tdir/$tfile-1 || { error "setstripe -c -1 failed"; return 1; } # We need to send a write to every object to get parent FID info set. @@ -2038,20 +2082,20 @@ test_27z() { run_test 27z "check SEQ/OID on the MDT and OST filesystems" test_27A() { # b=19102 - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return + [ $PARALLEL == "yes" ] && skip "skip parallel run" save_layout_restore_at_exit $MOUNT - - $SETSTRIPE -c 0 -i -1 -S 0 $MOUNT - wait_update $HOSTNAME "$GETSTRIPE -c $MOUNT | sed 's/ *//g'" "1" 20 || - error "stripe count $($GETSTRIPE -c $MOUNT) != 1" - local default_size=$($GETSTRIPE -S $MOUNT) - local default_offset=$($GETSTRIPE -i $MOUNT) + $LFS setstripe -c 0 -i -1 -S 0 $MOUNT + wait_update $HOSTNAME "$GETSTRIPE -c $MOUNT | sed 's/ *//g'" "1" 20 || + error "stripe count $($GETSTRIPE -c $MOUNT) != 1" + local default_size=$($GETSTRIPE -S $MOUNT) + local default_offset=$($GETSTRIPE -i $MOUNT) local dsize=$(do_facet $SINGLEMDS \ "$LCTL get_param -n lod.$(facet_svc $SINGLEMDS)*.stripesize") - [ $default_size -eq $dsize ] || - error "stripe size $default_size != $dsize" - [ $default_offset -eq -1 ] ||error "stripe offset $default_offset != -1" + [ $default_size -eq $dsize ] || + error "stripe size $default_size != $dsize" + [ $default_offset -eq -1 ] || + error "stripe offset $default_offset != -1" } run_test 27A "check filesystem-wide default LOV EA values" @@ -2079,7 +2123,7 @@ test_27B() { # LU-2523 run_test 27B "call setstripe on open unlinked file/rename victim" test_27C() { #LU-2871 - [[ $OSTCOUNT -lt 2 ]] && skip "needs >= 2 OSTs" && return + [[ $OSTCOUNT -lt 2 ]] && skip_env "needs >= 2 OSTs" declare -a ost_idx local index @@ -2117,9 +2161,10 @@ test_27C() { #LU-2871 run_test 27C "check full striping across all OSTs" test_27D() { - [ $OSTCOUNT -lt 2 ] && skip "needs >= 2 OSTs" && return - [ -n "$FILESET" ] && skip "SKIP due to FILESET set" && return - remote_mds_nodsh && skip "remote MDS with nodsh" && return + [ $OSTCOUNT -lt 2 ] && skip_env "needs >= 2 OSTs" + [ -n "$FILESET" ] && skip "SKIP due to FILESET set" + remote_mds_nodsh && skip "remote MDS with nodsh" + local POOL=${POOL:-testpool} local first_ost=0 local last_ost=$(($OSTCOUNT - 1)) @@ -2155,9 +2200,9 @@ run_test 27D "validate llapi_layout API" # Verify that default_easize is increased from its initial value after # accessing a widely striped file. test_27E() { - [ $OSTCOUNT -lt 2 ] && skip "needs >= 2 OSTs" && return + [ $OSTCOUNT -lt 2 ] && skip_env "needs >= 2 OSTs" [ $(lustre_version_code client) -lt $(version_code 2.5.57) ] && - skip "client does not have LU-3338 fix" && return + skip "client does not have LU-3338 fix" # 72 bytes is the minimum space required to store striping # information for a file striped across one OST: @@ -2184,10 +2229,11 @@ test_27E() { run_test 27E "check that default extended attribute size properly increases" test_27F() { # LU-5346/LU-7975 - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return + [ $PARALLEL == "yes" ] && skip "skip parallel run" + [[ $OSTCOUNT -lt 2 ]] && skip "needs >= 2 OSTs" [[ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.8.51) ]] && - skip "Need MDS version at least 2.8.51" && return - remote_ost_nodsh && skip "remote OST with nodsh" && return + skip "Need MDS version at least 2.8.51" + remote_ost_nodsh && skip "remote OST with nodsh" test_mkdir $DIR/$tdir rm -f $DIR/$tdir/f0 @@ -2222,6 +2268,57 @@ test_27F() { # LU-5346/LU-7975 } run_test 27F "Client resend delayed layout creation with non-zero size" +test_27G() { #LU-10629 + [ -n "$FILESET" ] && skip "SKIP due to FILESET set" && return + remote_mds_nodsh && skip "remote MDS with nodsh" && return + local POOL=${POOL:-testpool} + local ostrange="0 0 1" + + test_mkdir $DIR/$tdir + pool_add $POOL || error "pool_add failed" + pool_add_targets $POOL $ostrange || error "pool_add_targets failed" + $LFS setstripe -p $POOL $DIR/$tdir + + local pool=$($LFS getstripe -p $DIR/$tdir) + + [ "$pool" = "$POOL" ] || error "Striping failed got '$pool' not '$POOL'" + + $LFS setstripe -d $DIR/$tdir + + pool=$($LFS getstripe -p $DIR/$tdir) + + rmdir $DIR/$tdir + + [ -z "$pool" ] || error "'$pool' is not empty" +} +run_test 27G "Clear OST pool from stripe" + +test_27H() { + [[ $(lustre_version_code $SINGLEMDS) -le $(version_code 2.11.54) ]] && + skip "Need MDS version newer than 2.11.54" + [[ $OSTCOUNT -lt 3 ]] && skip_env "needs >= 3 OSTs" + test_mkdir $DIR/$tdir + $LFS setstripe -o 0 -o 2 $DIR/$tdir || error "setstripe failed" + touch $DIR/$tdir/$tfile + $LFS getstripe -c $DIR/$tdir/$tfile + [ $($LFS getstripe -c $DIR/$tdir/$tfile) -eq 2 ] || + error "two-stripe file doesn't have two stripes" + + dd if=/dev/zero of=$DIR/$tdir/$tfile bs=4k count=4 || error "dd failed" + $LFS getstripe -y $DIR/$tdir/$tfile + (( $($LFS getstripe -y $DIR/$tdir/$tfile | + egrep -c "l_ost_idx: [02]$") == "2" )) || + error "expected l_ost_idx: [02]$ not matched" + + # make sure ost list have been cleared + local stripesize=$($GETSTRIPE -S $DIR/$tdir) + $LFS setstripe -S $((stripesize * 4)) -i 1 \ + -c $((OSTCOUNT - 1)) $DIR/$tdir || error "setstripe" + touch $DIR/$tdir/f3 + $LVERIFY $DIR/$tdir $DIR/$tdir/f3 || error "lverify failed" +} +run_test 27H "Set specific OSTs stripe" + # createtest also checks that device nodes are created and # then visible correctly (#2091) test_28() { # bug 2091 @@ -2231,7 +2328,8 @@ test_28() { # bug 2091 run_test 28 "create/mknod/mkdir with bad file types ============" test_29() { - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return 0 + [ $PARALLEL == "yes" ] && skip "skip parallel run" + sync; sleep 1; sync # flush out any dirty pages from previous tests cancel_lru_locks test_mkdir $DIR/d29 @@ -2296,7 +2394,8 @@ test_30b() { run_test 30b "execute binary from Lustre as non-root ===========" test_30c() { # b=22376 - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return + [ $PARALLEL == "yes" ] && skip "skip parallel run" + cp `which ls` $DIR || cp /bin/ls $DIR chmod a-rw $DIR/ls cancel_lru_locks mdc @@ -2344,7 +2443,8 @@ test_31e() { # bug 2904 run_test 31e "remove of open non-empty directory ===============" test_31f() { # bug 4554 - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return + [ $PARALLEL == "yes" ] && skip "skip parallel run" + set -vx test_mkdir $DIR/d31f $SETSTRIPE -S 1048576 -c 1 $DIR/d31f @@ -2496,7 +2596,7 @@ test_31o() { # LU-2901 run_test 31o "duplicate hard links with same filename" test_31p() { - [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return + [ $MDSCOUNT -lt 2 ] && skip_env "needs >= 2 MDTs" test_mkdir $DIR/$tdir $LFS setdirstripe -i0 -c2 $DIR/$tdir/striped_dir @@ -2525,7 +2625,8 @@ cleanup_test32_mount() { } test_32a() { - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return + [ $PARALLEL == "yes" ] && skip "skip parallel run" + echo "== more mountpoints and symlinks =================" [ -e $DIR/$tdir ] && rm -fr $DIR/$tdir trap cleanup_test32_mount EXIT @@ -2539,7 +2640,8 @@ test_32a() { run_test 32a "stat d32a/ext2-mountpoint/.. =====================" test_32b() { - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return + [ $PARALLEL == "yes" ] && skip "skip parallel run" + [ -e $DIR/$tdir ] && rm -fr $DIR/$tdir trap cleanup_test32_mount EXIT test_mkdir -p $DIR/$tdir/ext2-mountpoint @@ -2552,7 +2654,8 @@ test_32b() { run_test 32b "open d32b/ext2-mountpoint/.. =====================" test_32c() { - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return + [ $PARALLEL == "yes" ] && skip "skip parallel run" + [ -e $DIR/$tdir ] && rm -fr $DIR/$tdir trap cleanup_test32_mount EXIT test_mkdir -p $DIR/$tdir/ext2-mountpoint @@ -2566,7 +2669,8 @@ test_32c() { run_test 32c "stat d32c/ext2-mountpoint/../d2/test_dir =========" test_32d() { - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return + [ $PARALLEL == "yes" ] && skip "skip parallel run" + [ -e $DIR/$tdir ] && rm -fr $DIR/$tdir trap cleanup_test32_mount EXIT test_mkdir -p $DIR/$tdir/ext2-mountpoint @@ -2627,7 +2731,8 @@ test_32h() { run_test 32h "open d32h/symlink->tmp/symlink->lustre-subdir/${tdir}2" test_32i() { - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return + [ $PARALLEL == "yes" ] && skip "skip parallel run" + [ -e $DIR/$tdir ] && rm -fr $DIR/$tdir trap cleanup_test32_mount EXIT test_mkdir -p $DIR/$tdir/ext2-mountpoint @@ -2641,7 +2746,8 @@ test_32i() { run_test 32i "stat d32i/ext2-mountpoint/../test_file ===========" test_32j() { - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return + [ $PARALLEL == "yes" ] && skip "skip parallel run" + [ -e $DIR/$tdir ] && rm -fr $DIR/$tdir trap cleanup_test32_mount EXIT test_mkdir -p $DIR/$tdir/ext2-mountpoint @@ -2655,7 +2761,8 @@ test_32j() { run_test 32j "open d32j/ext2-mountpoint/../test_file ===========" test_32k() { - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return + [ $PARALLEL == "yes" ] && skip "skip parallel run" + rm -fr $DIR/$tdir trap cleanup_test32_mount EXIT test_mkdir -p $DIR/$tdir/ext2-mountpoint @@ -2670,7 +2777,8 @@ test_32k() { run_test 32k "stat d32k/ext2-mountpoint/../d2/test_file ========" test_32l() { - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return + [ $PARALLEL == "yes" ] && skip "skip parallel run" + rm -fr $DIR/$tdir trap cleanup_test32_mount EXIT test_mkdir -p $DIR/$tdir/ext2-mountpoint @@ -2749,7 +2857,8 @@ test_32p() { run_test 32p "open d32p/symlink->tmp/symlink->lustre-root/$tfile" test_32q() { - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return + [ $PARALLEL == "yes" ] && skip "skip parallel run" + [ -e $DIR/$tdir ] && rm -fr $DIR/$tdir trap cleanup_test32_mount EXIT test_mkdir -p $DIR/$tdir/ext2-mountpoint @@ -2762,7 +2871,8 @@ test_32q() { run_test 32q "stat follows mountpoints in Lustre (should return error)" test_32r() { - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return + [ $PARALLEL == "yes" ] && skip "skip parallel run" + [ -e $DIR/$tdir ] && rm -fr $DIR/$tdir trap cleanup_test32_mount EXIT test_mkdir -p $DIR/$tdir/ext2-mountpoint @@ -2805,13 +2915,14 @@ test_33b() { run_test 33b "test open file with malformed flags (No panic)" test_33c() { - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return + [ $PARALLEL == "yes" ] && skip "skip parallel run" + remote_ost_nodsh && skip "remote OST with nodsh" + local ostnum local ostname local write_bytes local all_zeros - remote_ost_nodsh && skip "remote OST with nodsh" && return all_zeros=: rm -fr $DIR/$tdir test_mkdir $DIR/$tdir @@ -2871,8 +2982,9 @@ test_33c() { run_test 33c "test llobdstat and write_bytes" test_33d() { - [[ $MDSCOUNT -lt 2 ]] && skip "needs >= 2 MDTs" && return - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return + [[ $MDSCOUNT -lt 2 ]] && skip_env "needs >= 2 MDTs" + [ $PARALLEL == "yes" ] && skip "skip parallel run" + local MDTIDX=1 local remote_dir=$DIR/$tdir/remote_dir @@ -2896,7 +3008,7 @@ test_33d() { run_test 33d "openfile with 444 modes and malformed flags under remote dir" test_33e() { - [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return + [ $MDSCOUNT -lt 2 ] && skip_env "needs >= 2 MDTs" mkdir $DIR/$tdir @@ -2947,8 +3059,8 @@ cleanup_33f() { } test_33f() { - [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return - remote_mds_nodsh && skip "remote MDS with nodsh" && return + [ $MDSCOUNT -lt 2 ] && skip_env "needs >= 2 MDTs" + remote_mds_nodsh && skip "remote MDS with nodsh" mkdir $DIR/$tdir chmod go+rwx $DIR/$tdir @@ -3040,7 +3152,8 @@ test_34e() { run_test 34e "create objects, some with size and some without ==" test_34f() { # bug 6242, 6243 - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return + [ $PARALLEL == "yes" ] && skip "skip parallel run" + SIZE34F=48000 rm -f $DIR/f34f $MCREATE $DIR/f34f || error "mcreate failed" @@ -3055,7 +3168,8 @@ test_34f() { # bug 6242, 6243 run_test 34f "read from a file with no objects until EOF =======" test_34g() { - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return + [ $PARALLEL == "yes" ] && skip "skip parallel run" + dd if=/dev/zero of=$DIR/$tfile bs=1 count=100 seek=$TEST_34_SIZE || error "dd failed" $TRUNCATE $DIR/$tfile $((TEST_34_SIZE / 2))|| error "truncate failed" @@ -3075,7 +3189,8 @@ test_34g() { run_test 34g "truncate long file ===============================" test_34h() { - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return + [ $PARALLEL == "yes" ] && skip "skip parallel run" + local gid=10 local sz=1000 @@ -3088,7 +3203,7 @@ test_34h() { # Since just timed wait is not good enough, let's do a sync write # that way we are sure enough time for a roundtrip + processing # passed + 2 seconds of extra margin. - dd if=/dev/zero of=$DIR/${tfile}-1 bs=4096 oflag=direct count=1 + dd if=/dev/zero of=$DIR/${tfile}-1 bs=$PAGE_SIZE oflag=direct count=1 rm $DIR/${tfile}-1 sleep 2 @@ -3139,7 +3254,8 @@ test_36d() { run_test 36d "non-root OST utime check (open, utime)" test_36e() { - [ $RUNAS_ID -eq $UID ] && skip_env "RUNAS_ID = UID = $UID -- skipping" && return + [ $RUNAS_ID -eq $UID ] && skip_env "RUNAS_ID = UID = $UID -- skipping" + test_mkdir $DIR/$tdir touch $DIR/$tdir/$tfile $RUNAS utime $DIR/$tdir/$tfile && @@ -3175,15 +3291,17 @@ subr_36fh() { } test_36f() { - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return + [ $PARALLEL == "yes" ] && skip "skip parallel run" + #define OBD_FAIL_OST_BRW_PAUSE_BULK 0x214 subr_36fh "0x80000214" } run_test 36f "utime on file racing with OST BRW write ==========" test_36g() { - remote_ost_nodsh && skip "remote OST with nodsh" && return - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return + remote_ost_nodsh && skip "remote OST with nodsh" + [ $PARALLEL == "yes" ] && skip "skip parallel run" + local fmd_max_age local fmd_before local fmd_after @@ -3209,14 +3327,15 @@ test_36g() { run_test 36g "filter mod data cache expiry =====================" test_36h() { - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return + [ $PARALLEL == "yes" ] && skip "skip parallel run" + #define OBD_FAIL_OST_BRW_PAUSE_BULK2 0x227 subr_36fh "0x80000227" } run_test 36h "utime on file racing with OST BRW write ==========" test_36i() { - [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return + [ $MDSCOUNT -lt 2 ] && skip_env "needs >= 2 MDTs" test_mkdir $DIR/$tdir $LFS setdirstripe -i0 -c$MDSCOUNT $DIR/$tdir/striped_dir @@ -3344,9 +3463,9 @@ run_test 39c "mtime change on rename ===========================" # bug 21114 test_39d() { - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return - touch $DIR1/$tfile + [ $PARALLEL == "yes" ] && skip "skip parallel run" + touch $DIR1/$tfile touch -m -d @$TEST_39_MTIME $DIR1/$tfile for (( i=0; i < 2; i++ )) ; do @@ -3362,7 +3481,8 @@ run_test 39d "create, utime, stat ==============================" # bug 21114 test_39e() { - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return + [ $PARALLEL == "yes" ] && skip "skip parallel run" + touch $DIR1/$tfile local mtime1=`stat -c %Y $DIR1/$tfile` @@ -3381,7 +3501,8 @@ run_test 39e "create, stat, utime, stat ========================" # bug 21114 test_39f() { - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return + [ $PARALLEL == "yes" ] && skip "skip parallel run" + touch $DIR1/$tfile mtime1=`stat -c %Y $DIR1/$tfile` @@ -3401,7 +3522,8 @@ run_test 39f "create, stat, sleep, utime, stat =================" # bug 11063 test_39g() { - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return + [ $PARALLEL == "yes" ] && skip "skip parallel run" + echo hello >> $DIR1/$tfile local mtime1=`stat -c %Y $DIR1/$tfile` @@ -3421,7 +3543,8 @@ run_test 39g "write, chmod, stat ===============================" # bug 11063 test_39h() { - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return + [ $PARALLEL == "yes" ] && skip "skip parallel run" + touch $DIR1/$tfile sleep 1 @@ -3447,7 +3570,8 @@ test_39h() { run_test 39h "write, utime within one second, stat =============" test_39i() { - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return + [ $PARALLEL == "yes" ] && skip "skip parallel run" + touch $DIR1/$tfile sleep 1 @@ -3469,7 +3593,8 @@ test_39i() { run_test 39i "write, rename, stat ==============================" test_39j() { - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return + [ $PARALLEL == "yes" ] && skip "skip parallel run" + start_full_debug_logging touch $DIR1/$tfile sleep 1 @@ -3501,7 +3626,8 @@ test_39j() { run_test 39j "write, rename, close, stat =======================" test_39k() { - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return + [ $PARALLEL == "yes" ] && skip "skip parallel run" + touch $DIR1/$tfile sleep 1 @@ -3530,8 +3656,9 @@ run_test 39k "write, utime, close, stat ========================" TEST_39_ATIME=`date -d "1 year" +%s` test_39l() { - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return - remote_mds_nodsh && skip "remote MDS with nodsh" && return + [ $PARALLEL == "yes" ] && skip "skip parallel run" + remote_mds_nodsh && skip "remote MDS with nodsh" + local atime_diff=$(do_facet $SINGLEMDS \ lctl get_param -n mdd.*MDT0000*.atime_diff) rm -rf $DIR/$tdir @@ -3579,7 +3706,8 @@ test_39l() { run_test 39l "directory atime update ===========================" test_39m() { - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return + [ $PARALLEL == "yes" ] && skip "skip parallel run" + touch $DIR1/$tfile sleep 2 local far_past_mtime=$(date -d "May 29 1953" +%s) @@ -3600,7 +3728,8 @@ test_39m() { run_test 39m "test atime and mtime before 1970" test_39n() { # LU-3832 - remote_mds_nodsh && skip "remote MDS with nodsh" && return + remote_mds_nodsh && skip "remote MDS with nodsh" + local atime_diff=$(do_facet $SINGLEMDS \ lctl get_param -n mdd.*MDT0000*.atime_diff) local atime0 @@ -3652,7 +3781,8 @@ test_39o() { run_test 39o "directory cached attributes updated after create" test_39p() { - [[ $MDSCOUNT -lt 2 ]] && skip "needs >= 2 MDTs" && return + [[ $MDSCOUNT -lt 2 ]] && skip_env "needs >= 2 MDTs" + local MDTIDX=1 TESTDIR=$DIR/$tdir/$tdir [ -e $TESTDIR ] && rm -rf $TESTDIR @@ -3761,7 +3891,8 @@ setup_test42() { # Tests 42* verify that our behaviour is correct WRT caching, file closure, # file truncation, and file removal. test_42a() { - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return + [ $PARALLEL == "yes" ] && skip "skip parallel run" + setup_test42 cancel_lru_locks $OSC stop_writeback @@ -3777,7 +3908,8 @@ test_42a() { run_test 42a "ensure that we don't flush on close" test_42b() { - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return + [ $PARALLEL == "yes" ] && skip "skip parallel run" + setup_test42 cancel_lru_locks $OSC stop_writeback @@ -3834,27 +3966,29 @@ trunc_test() { } test_42c() { - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return - trunc_test 42c 1024 - [ $BEFOREWRITES -eq $AFTERWRITES ] && \ - error "beforewrites $BEFOREWRITES == afterwrites $AFTERWRITES on truncate" - rm $file + [ $PARALLEL == "yes" ] && skip "skip parallel run" + + trunc_test 42c 1024 + [ $BEFOREWRITES -eq $AFTERWRITES ] && + error "beforewrites $BEFOREWRITES == afterwrites $AFTERWRITES on truncate" + rm $file } run_test 42c "test partial truncate of file with cached dirty data" test_42d() { - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return - trunc_test 42d 0 - [ $BEFOREWRITES -eq $AFTERWRITES ] || \ - error "beforewrites $BEFOREWRITES != afterwrites $AFTERWRITES on truncate" - rm $file + [ $PARALLEL == "yes" ] && skip "skip parallel run" + + trunc_test 42d 0 + [ $BEFOREWRITES -eq $AFTERWRITES ] || + error "beforewrites $BEFOREWRITES != afterwrites $AFTERWRITES on truncate" + rm $file } run_test 42d "test complete truncate of file with cached dirty data" test_42e() { # bug22074 - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return + [ $PARALLEL == "yes" ] && skip "skip parallel run" + local TDIR=$DIR/${tdir}e - local pagesz=$(page_size) local pages=16 # hardcoded 16 pages, don't change it. local files=$((OSTCOUNT * 500)) # hopefully 500 files on each OST local proc_osc0="osc.${FSNAME}-OST0000-osc-[^MDT]*" @@ -3900,7 +4034,7 @@ test_42e() { # bug22074 $LCTL set_param $proc_osc0/rpc_stats 0 for ((;i<$files; i++)); do [ $($GETSTRIPE -i $TDIR/f$i) -eq 0 ] || continue - dd if=/dev/zero of=$TDIR/f$i bs=$pagesz count=$pages 2>/dev/null + dd if=/dev/zero of=$TDIR/f$i bs=$PAGE_SIZE count=$pages 2>/dev/null done sync $LCTL get_param $proc_osc0/rpc_stats @@ -3955,7 +4089,8 @@ test_43a() { run_test 43a "open(RDWR) of file being executed should return -ETXTBSY" test_43b() { - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return + [ $PARALLEL == "yes" ] && skip "skip parallel run" + test_mkdir $DIR/$tdir cp -p $(which $MULTIOP) $DIR/$tdir/multiop || cp -p multiop $DIR/$tdir/multiop @@ -3979,7 +4114,8 @@ test_43c() { run_test 43c "md5sum of copy into lustre" test_44A() { # was test_44 - [[ $OSTCOUNT -lt 2 ]] && skip_env "needs >= 2 OSTs" && return + [[ $OSTCOUNT -lt 2 ]] && skip_env "needs >= 2 OSTs" + dd if=/dev/zero of=$DIR/f1 bs=4k count=1 seek=1023 dd if=$DIR/f1 bs=4k count=1 > /dev/null } @@ -3988,10 +4124,10 @@ run_test 44A "zero length read from a sparse stripe" test_44a() { local nstripe=$($LCTL lov_getconfig $DIR | grep default_stripe_count: | awk '{ print $2 }') - [ -z "$nstripe" ] && skip "can't get stripe info" && return + [ -z "$nstripe" ] && skip "can't get stripe info" [[ $nstripe -gt $OSTCOUNT ]] && - skip "Wrong default_stripe_count: $nstripe (OSTCOUNT: $OSTCOUNT)" && - return + skip "Wrong default_stripe_count: $nstripe OSTCOUNT: $OSTCOUNT" + local stride=$($LCTL lov_getconfig $DIR | grep default_stripe_size: | awk '{ print $2 }') if [[ $nstripe -eq 0 || $nstripe -eq -1 ]]; then @@ -4043,7 +4179,8 @@ do_dirty_record() { echo before $before, after $after } test_45() { - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return + [ $PARALLEL == "yes" ] && skip "skip parallel run" + f="$DIR/f45" # Obtain grants from OST if it supports it echo blah > ${f}_grant @@ -4071,14 +4208,15 @@ run_test 45 "osc io page accounting ============================" # objects offset and an assert hit when an rpc was built with 1023's mapped # offset 511 and 511's raw 511 offset. it also found general redirtying bugs. test_46() { - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return + [ $PARALLEL == "yes" ] && skip "skip parallel run" + f="$DIR/f46" stop_writeback sync - dd if=/dev/zero of=$f bs=`page_size` seek=511 count=1 + dd if=/dev/zero of=$f bs=$PAGE_SIZE seek=511 count=1 sync - dd conv=notrunc if=/dev/zero of=$f bs=`page_size` seek=1023 count=1 - dd conv=notrunc if=/dev/zero of=$f bs=`page_size` seek=511 count=1 + dd conv=notrunc if=/dev/zero of=$f bs=$PAGE_SIZE seek=1023 count=1 + dd conv=notrunc if=/dev/zero of=$f bs=$PAGE_SIZE seek=511 count=1 sync start_writeback } @@ -4089,8 +4227,8 @@ run_test 46 "dirtying a previously written page ================" test_48a() { # bug 2399 [ $(facet_fstype $SINGLEMDS) = "zfs" ] && [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.3.63) ] && - skip "MDS prior to 2.3.63 handle ZFS dir .. incorrectly" && - return + skip "MDS prior to 2.3.63 handle ZFS dir .. incorrectly" + test_mkdir $DIR/$tdir cd $DIR/$tdir mv $DIR/$tdir $DIR/$tdir.new || error "move directory failed" @@ -4187,8 +4325,9 @@ test_48e() { # bug 4134 run_test 48e "Access to recreated parent subdir (should return errors)" test_49() { # LU-1030 - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return - remote_ost_nodsh && skip "remote OST with nodsh" && return + [ $PARALLEL == "yes" ] && skip "skip parallel run" + remote_ost_nodsh && skip "remote OST with nodsh" + # get ost1 size - lustre-OST0000 ost1_size=$(do_facet ost1 $LFS df | grep ${ost1_svc} | awk '{ print $4 }') @@ -4247,9 +4386,9 @@ cleanup_print_lfs_df () { } test_51b() { - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return - local dir=$DIR/$tdir + [ $PARALLEL == "yes" ] && skip "skip parallel run" + local dir=$DIR/$tdir local nrdirs=$((65536 + 100)) # cleanup the directory @@ -4259,17 +4398,15 @@ test_51b() { $LFS df $LFS df -i - local mdtidx=$(printf "%04x" $($LFS getstripe -M $dir)) + local mdtidx=$(printf "%04x" $($LFS getstripe -m $dir)) local numfree=$(lctl get_param -n mdc.$FSNAME-MDT$mdtidx*.filesfree) [[ $numfree -lt $nrdirs ]] && - skip "not enough free inodes ($numfree) on MDT$mdtidx" && - return + skip "not enough free inodes ($numfree) on MDT$mdtidx" # need to check free space for the directories as well local blkfree=$(lctl get_param -n mdc.$FSNAME-MDT$mdtidx*.kbytesavail) numfree=$(( blkfree / $(fs_inode_ksize) )) - [[ $numfree -lt $nrdirs ]] && skip "not enough blocks ($numfree)" && - return + [[ $numfree -lt $nrdirs ]] && skip "not enough blocks ($numfree)" trap cleanup_print_lfs_df EXIT @@ -4313,8 +4450,9 @@ test_51b() { run_test 51b "exceed 64k subdirectory nlink limit on create, verify unlink" test_51d() { - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return - [[ $OSTCOUNT -lt 3 ]] && skip_env "needs >= 3 OSTs" && return + [ $PARALLEL == "yes" ] && skip "skip parallel run" + [[ $OSTCOUNT -lt 3 ]] && skip_env "needs >= 3 OSTs" + test_mkdir $DIR/$tdir createmany -o $DIR/$tdir/t- 1000 $LFS getstripe $DIR/$tdir > $TMP/$tfile @@ -4351,8 +4489,7 @@ run_test 51d "check object distribution" test_51e() { if [ "$(facet_fstype $SINGLEMDS)" != ldiskfs ]; then - skip "ldiskfs only test" - return + skip_env "ldiskfs only test" fi test_mkdir -c1 $DIR/$tdir @@ -4372,8 +4509,8 @@ test_51f() { local max=100000 local ulimit_old=$(ulimit -n) local spare=20 # number of spare fd's for scripts/libraries, etc. - local mdt=$(lfs getstripe -M $DIR/$tdir) - local numfree=$(lfs df -i $DIR/$tdir | awk '/MDT:'$mdt'/ { print $4 }') + local mdt=$($LFS getstripe -m $DIR/$tdir) + local numfree=$($LFS df -i $DIR/$tdir | awk '/MDT:'$mdt'/ { print $4 }') echo "MDT$mdt numfree=$numfree, max=$max" [[ $numfree -gt $max ]] && numfree=$max || numfree=$((numfree * 7 / 8)) @@ -4440,9 +4577,9 @@ test_52b() { run_test 52b "immutable flag test (should return errors) =======" test_53() { - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return - remote_mds_nodsh && skip "remote MDS with nodsh" && return - remote_ost_nodsh && skip "remote OST with nodsh" && return + [ $PARALLEL == "yes" ] && skip "skip parallel run" + remote_mds_nodsh && skip "remote MDS with nodsh" + remote_ost_nodsh && skip "remote OST with nodsh" local param local param_seq @@ -4506,7 +4643,7 @@ test_53() { run_test 53 "verify that MDS and OSTs agree on pre-creation ====" test_54a() { - perl -MSocket -e ';' || { skip "no Socket perl module installed" && return; } + perl -MSocket -e ';' || skip "no Socket perl module installed" $SOCKETSERVER $DIR/socket || error "$SOCKETSERVER $DIR/socket failed: $?" @@ -4520,7 +4657,7 @@ test_54b() { f="$DIR/f54b" mknod $f c 1 3 chmod 0666 $f - dd if=/dev/zero of=$f bs=$(page_size) count=1 + dd if=/dev/zero of=$f bs=$PAGE_SIZE count=1 } run_test 54b "char device works in lustre ======================" @@ -4550,25 +4687,26 @@ cleanup_54c() { } test_54c() { - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return + [ $PARALLEL == "yes" ] && skip "skip parallel run" + loopdev="$DIR/loop54c" find_loop_dev - [ -z "$LOOPNUM" ] && echo "couldn't find empty loop device" && return + [ -z "$LOOPNUM" ] && skip_env "couldn't find empty loop device" trap cleanup_54c EXIT mknod $loopdev b 7 $LOOPNUM echo "make a loop file system with $DIR/$tfile on $loopdev ($LOOPNUM)." - dd if=/dev/zero of=$DIR/$tfile bs=$(get_page_size client) seek=1024 count=1 > /dev/null + dd if=/dev/zero of=$DIR/$tfile bs=$PAGE_SIZE seek=1024 count=1 > /dev/null losetup $loopdev $DIR/$tfile || error "can't set up $loopdev for $DIR/$tfile" mkfs.ext2 $loopdev || error "mke2fs on $loopdev" test_mkdir $DIR/$tdir mount -t ext2 $loopdev $DIR/$tdir || error "error mounting $loopdev on $DIR/$tdir" - dd if=/dev/zero of=$DIR/$tdir/tmp bs=$(get_page_size client) count=30 || + dd if=/dev/zero of=$DIR/$tdir/tmp bs=$PAGE_SIZE count=30 || error "dd write" df $DIR/$tdir - dd if=$DIR/$tdir/tmp of=/dev/zero bs=$(get_page_size client) count=30 || + dd if=$DIR/$tdir/tmp of=/dev/zero bs=$PAGE_SIZE count=30 || error "dd read" cleanup_54c } @@ -4654,16 +4792,19 @@ test_56a() { $LFS getstripe --obd wrong_uuid $dir 2>&1 | grep -q "unknown obduuid" || error "$LFS getstripe --obd wrong_uuid: should return error" - [[ $OSTCOUNT -lt 2 ]] && - skip_env "skip '$LFS getstripe --obd' tests: $OSTCOUNT < 2" && - return + [[ $OSTCOUNT -lt 2 ]] && skip_env "needs >= 2 OSTs" local ostidx=1 local obduuid=$(ostuuid_from_index $ostidx) local found=$($LFS getstripe -r --obd $obduuid $dir | - egrep -c "obdidx|l_ost_idx") + grep 'lmm_stripe_offset:' | grep -c " $ostidx\$") + + filenum=$($LFS getstripe -ir $dir | grep -c "^$ostidx\$") + [[ $($LFS getstripe -id $dir) -ne $ostidx ]] || + ((filenum--)) + [[ $($LFS getstripe -id $dir/dir) -ne $ostidx ]] || + ((filenum--)) - filenum=$($LFS getstripe -ir $dir | grep "^$ostidx\$" | wc -l) [[ $found -eq $filenum ]] || error "$LFS getstripe --obd: found $found expect $filenum" [[ $($LFS getstripe -r -v --obd $obduuid $dir | @@ -4697,14 +4838,14 @@ test_56b() { run_test 56b "check $LFS getdirstripe" test_56c() { - remote_ost_nodsh && skip "remote OST with nodsh" && return + remote_ost_nodsh && skip "remote OST with nodsh" local ost_idx=0 local ost_name=$(ostname_from_index $ost_idx) local old_status=$(ost_dev_status $ost_idx) [[ -z "$old_status" ]] || - { skip_env "OST $ost_name is in $old_status status"; return 0; } + skip_env "OST $ost_name is in $old_status status" do_facet ost1 $LCTL set_param -n obdfilter.$ost_name.degraded=1 sleep_maxage @@ -4909,7 +5050,7 @@ run_test 56o "check lfs find -mtime for old files" test_56p() { [ $RUNAS_ID -eq $UID ] && - skip_env "RUNAS_ID = UID = $UID -- skipping" && return + skip_env "RUNAS_ID = UID = $UID -- skipping" local dir=$DIR/$tdir @@ -4933,7 +5074,7 @@ run_test 56p "check lfs find -uid and ! -uid" test_56q() { [ $RUNAS_ID -eq $UID ] && - skip_env "RUNAS_ID = UID = $UID -- skipping" && return + skip_env "RUNAS_ID = UID = $UID -- skipping" local dir=$DIR/$tdir @@ -5002,7 +5143,7 @@ test_56r() { run_test 56r "check lfs find -size works" test_56s() { # LU-611 #LU-9369 - [[ $OSTCOUNT -lt 2 ]] && skip "need at least 2 OSTs" && return 0 + [[ $OSTCOUNT -lt 2 ]] && skip_env "need at least 2 OSTs" local dir=$DIR/$tdir local onestripe=$(((NUMDIRS + 1) * NUMFILES)) @@ -5165,25 +5306,26 @@ test_56u() { # LU-611 run_test 56u "check lfs find -stripe-index works" test_56v() { - local MDT_IDX=0 + local mdt_idx=0 local dir=$DIR/$tdir setup_56 $dir $NUMFILES $NUMDIRS - UUID=$(mdtuuid_from_index $MDT_IDX $dir) - [ -z "$UUID" ] && error "mdtuuid_from_index cannot find MDT $MDT_IDX" + UUID=$(mdtuuid_from_index $mdt_idx $dir) + [ -z "$UUID" ] && error "mdtuuid_from_index cannot find MDT $mdt_idx" - for file in $($LFS find -mdt $UUID $dir); do - file_mdt_idx=$($LFS getstripe -M $file) - [ $file_mdt_idx -eq $MDT_IDX ] || - error "lfind -mdt $UUID != getstripe -M $file_mdt_idx" + for file in $($LFS find -m $UUID $dir); do + file_midx=$($LFS getstripe -m $file) + [ $file_midx -eq $mdt_idx ] || + error "lfs find -m $UUID != getstripe -m $file_midx" done } -run_test 56v "check 'lfs find -mdt match with lfs getstripe -M' =======" +run_test 56v "check 'lfs find -m match with lfs getstripe -m'" test_56w() { - [[ $OSTCOUNT -lt 2 ]] && skip_env "needs >= 2 OSTs" && return - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return + [[ $OSTCOUNT -lt 2 ]] && skip_env "needs >= 2 OSTs" + [ $PARALLEL == "yes" ] && skip "skip parallel run" + local dir=$DIR/$tdir setup_56 $dir $NUMFILES $NUMDIRS "-c $OSTCOUNT" "-c1" @@ -5198,8 +5340,7 @@ test_56w() { local free_space=$($LCTL get_param -n lov.$FSNAME-clilov-*.kbytesavail | head -n1) [[ $free_space -le $((required_space / 1024)) ]] && - skip_env "need $required_space, have $free_space kbytes" && - return + skip_env "need $required_space, have $free_space kbytes" local dd_bs=65536 local dd_count=$((file_size / dd_bs)) @@ -5371,7 +5512,7 @@ test_56wb() { run_test 56wb "check lfs_migrate pool support" test_56wc() { - local file1="$DIR/$tdir/file 1" + local file1="$DIR/$tdir/file1" echo -n "Creating test dir..." test_mkdir $DIR/$tdir &> /dev/null || error "cannot create dir" @@ -5472,7 +5613,8 @@ test_56wc() { run_test 56wc "check unrecognized options for lfs_migrate are passed through" test_56wd() { - [[ $OSTCOUNT -lt 2 ]] && skip_env "needs >= 2 OSTs" && return + [[ $OSTCOUNT -lt 2 ]] && skip_env "needs >= 2 OSTs" + local file1=$DIR/$tdir/file1 echo -n "Creating test dir..." @@ -5510,8 +5652,8 @@ test_56wd() { run_test 56wd "check lfs_migrate --rsync and --no-rsync work" test_56x() { - check_swap_layouts_support && return 0 - [[ $OSTCOUNT -lt 2 ]] && skip_env "needs >= 2 OSTs" && return + [[ $OSTCOUNT -lt 2 ]] && skip_env "needs >= 2 OSTs" + check_swap_layouts_support local dir=$DIR/$tdir local ref1=/etc/passwd @@ -5531,8 +5673,8 @@ test_56x() { run_test 56x "lfs migration support" test_56xa() { - check_swap_layouts_support && return 0 - [[ $OSTCOUNT -lt 2 ]] && skip_env "needs >= 2 OSTs" && return + [[ $OSTCOUNT -lt 2 ]] && skip_env "needs >= 2 OSTs" + check_swap_layouts_support local dir=$DIR/$tdir/$testnum @@ -5597,8 +5739,7 @@ check_migrate_links() { echo "${#paths[*]}" if [ ${#paths[*]} -lt $total_count -a "$begin" -eq 2 ]; then - echo "hard link list has unexpected size, skipping test" - return 0 + skip "hard link list has unexpected size, skipping test" fi if [ ${#paths[*]} -ge $total_count -a "$begin" -ne 2 ]; then error "link names should exceed xattrs size" @@ -5642,7 +5783,7 @@ check_migrate_links() { test_56xb() { [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.10.55) ] && - skip "Need MDS version at least 2.10.55" && return + skip "Need MDS version at least 2.10.55" local dir="$DIR/$tdir" @@ -5667,8 +5808,7 @@ run_test 56xb "lfs migration hard link support" test_56y() { [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.4.53) ] && - skip "No HSM $(lustre_build_version $SINGLEMDS) MDS < 2.4.53" && - return + skip "No HSM $(lustre_build_version $SINGLEMDS) MDS < 2.4.53" local res="" local dir=$DIR/$tdir @@ -5724,7 +5864,7 @@ test_56z() { # LU-4824 run_test 56z "lfs find should continue after an error" test_56aa() { # LU-5937 - [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return + [ $MDSCOUNT -lt 2 ] && skip_env "needs >= 2 MDTs" local dir=$DIR/$tdir @@ -5732,13 +5872,39 @@ test_56aa() { # LU-5937 $LFS setdirstripe -c$MDSCOUNT $dir/striped_dir createmany -o $dir/striped_dir/${tfile}- 1024 - local dirs=$(lfs find --size +8k $dir/) + local dirs=$($LFS find --size +8k $dir/) [ -n "$dirs" ] || error "lfs find --size wrong under striped dir" } run_test 56aa "lfs find --size under striped dir" +test_56ab() { # LU-10705 + test_mkdir $DIR/$tdir + dd if=/dev/zero of=$DIR/$tdir/$tfile.1 bs=8k count=1 seek=2k + dd if=/dev/zero of=$DIR/$tdir/$tfile.2 bs=4k count=1 seek=4k + dd if=/dev/zero of=$DIR/$tdir/$tfile.3 bs=1M count=2 seek=16 + # Flush writes to ensure valid blocks. Need to be more thorough for + # ZFS, since blocks are not allocated/returned to client immediately. + sync_all_data + wait_zfs_commit ost1 2 + cancel_lru_locks osc + ls -ls $DIR/$tdir + + local files=$($LFS find --size +16M $DIR/$tdir | wc -l) + + [[ $files == 3 ]] || error ">16M size files $files isn't 3 as expected" + + files=$($LFS find --blocks +1M $DIR/$tdir | wc -l) + [[ $files == 1 ]] || error ">1M blocks files $files isn't 1 as expected" + + rm -f $DIR/$tdir/$tfile.[123] +} +run_test 56ab "lfs find --blocks" + test_56ba() { + [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.10.50) ] && + skip "Need MDS version at least 2.10.50" + # Create composite files with one component local dir=$DIR/$tdir @@ -5802,7 +5968,7 @@ run_test 56ba "test lfs find --component-end, -start, -count, and -flags" test_56ca() { [[ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.10.57) ]] || - { skip "Need MDS version at least 2.10.57"; return 0; } + skip "Need MDS version at least 2.10.57" local td=$DIR/$tdir local tf=$td/$tfile @@ -5875,14 +6041,13 @@ test_56ca() { run_test 56ca "check lfs find --mirror-count|-N and --mirror-state" test_57a() { - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return + [ $PARALLEL == "yes" ] && skip "skip parallel run" # note test will not do anything if MDS is not local if [ "$(facet_fstype $SINGLEMDS)" != ldiskfs ]; then - skip "ldiskfs only test" - return + skip_env "ldiskfs only test" fi + remote_mds_nodsh && skip "remote MDS with nodsh" - remote_mds_nodsh && skip "remote MDS with nodsh" && return local MNTDEV="osd*.*MDT*.mntdev" DEV=$(do_facet $SINGLEMDS lctl get_param -n $MNTDEV) [ -z "$DEV" ] && error "can't access $MNTDEV" @@ -5897,59 +6062,61 @@ test_57a() { run_test 57a "verify MDS filesystem created with large inodes ==" test_57b() { - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return + [ $PARALLEL == "yes" ] && skip "skip parallel run" if [ "$(facet_fstype $SINGLEMDS)" != ldiskfs ]; then - skip "ldiskfs only test" - return + skip_env "ldiskfs only test" fi + remote_mds_nodsh && skip "remote MDS with nodsh" - remote_mds_nodsh && skip "remote MDS with nodsh" && return local dir=$DIR/$tdir - - local FILECOUNT=100 - local FILE1=$dir/f1 - local FILEN=$dir/f$FILECOUNT + local filecount=100 + local file1=$dir/f1 + local fileN=$dir/f$filecount rm -rf $dir || error "removing $dir" test_mkdir -c1 $dir - local mdtidx=$($LFS getstripe -M $dir) + local mdtidx=$($LFS getstripe -m $dir) local mdtname=MDT$(printf %04x $mdtidx) local facet=mds$((mdtidx + 1)) - echo "mcreating $FILECOUNT files" - createmany -m $dir/f 1 $FILECOUNT || \ - error "creating files in $dir" + echo "mcreating $filecount files" + createmany -m $dir/f 1 $filecount || error "creating files in $dir" # verify that files do not have EAs yet - $GETSTRIPE $FILE1 2>&1 | grep -q "no stripe" || error "$FILE1 has an EA" - $GETSTRIPE $FILEN 2>&1 | grep -q "no stripe" || error "$FILEN has an EA" + $LFS getstripe $file1 2>&1 | grep -q "no stripe" || + error "$file1 has an EA" + $LFS getstripe $fileN 2>&1 | grep -q "no stripe" || + error "$fileN has an EA" sync sleep 1 df $dir #make sure we get new statfs data - local MDSFREE=$(do_facet $facet \ - lctl get_param -n osd*.*$mdtname.kbytesfree) - local MDCFREE=$(lctl get_param -n mdc.*$mdtname-mdc-*.kbytesfree) + local mdsfree=$(do_facet $facet \ + lctl get_param -n osd*.*$mdtname.kbytesfree) + local mdcfree=$(lctl get_param -n mdc.*$mdtname-mdc-*.kbytesfree) + local file + echo "opening files to create objects/EAs" - local FILE - for FILE in `seq -f $dir/f%g 1 $FILECOUNT`; do - $OPENFILE -f O_RDWR $FILE > /dev/null 2>&1 || error "opening $FILE" + for file in $(seq -f $dir/f%g 1 $filecount); do + $OPENFILE -f O_RDWR $file > /dev/null 2>&1 || + error "opening $file" done # verify that files have EAs now - $GETSTRIPE $FILE1 | grep -q "obdidx" || error "$FILE1 missing EA" - $GETSTRIPE $FILEN | grep -q "obdidx" || error "$FILEN missing EA" + $LFS getstripe $file1 | grep -q "obdidx" || error "$file1 missing EA" + $LFS getstripe $fileN | grep -q "obdidx" || error "$fileN missing EA" sleep 1 #make sure we get new statfs data df $dir - local MDSFREE2=$(do_facet $facet \ - lctl get_param -n osd*.*$mdtname.kbytesfree) - local MDCFREE2=$(lctl get_param -n mdc.*$mdtname-mdc-*.kbytesfree) - if [[ $MDCFREE2 -lt $((MDCFREE - 16)) ]]; then - if [ "$MDSFREE" != "$MDSFREE2" ]; then - error "MDC before $MDCFREE != after $MDCFREE2" + local mdsfree2=$(do_facet $facet \ + lctl get_param -n osd*.*$mdtname.kbytesfree) + local mdcfree2=$(lctl get_param -n mdc.*$mdtname-mdc-*.kbytesfree) + + if [[ $mdcfree2 -lt $((mdcfree - 16)) ]]; then + if [ "$mdsfree" != "$mdsfree2" ]; then + error "MDC before $mdcfree != after $mdcfree2" else - echo "MDC before $MDCFREE != after $MDCFREE2" + echo "MDC before $mdcfree != after $mdcfree2" echo "unable to confirm if MDS has large inodes" fi fi @@ -5958,15 +6125,17 @@ test_57b() { run_test 57b "default LOV EAs are stored inside large inodes ===" test_58() { - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return + [ $PARALLEL == "yes" ] && skip "skip parallel run" [ -z "$(which wiretest 2>/dev/null)" ] && - skip_env "could not find wiretest" && return + skip_env "could not find wiretest" + wiretest } run_test 58 "verify cross-platform wire constants ==============" test_59() { - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return + [ $PARALLEL == "yes" ] && skip "skip parallel run" + echo "touch 130 files" createmany -o $DIR/f59- 130 echo "rm 130 files" @@ -5979,25 +6148,25 @@ run_test 59 "verify cancellation of llog records async =========" TEST60_HEAD="test_60 run $RANDOM" test_60a() { - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return - remote_mgs_nodsh && skip "remote MGS with nodsh" && return + [ $PARALLEL == "yes" ] && skip "skip parallel run" + remote_mgs_nodsh && skip "remote MGS with nodsh" do_facet mgs "! which run-llog.sh &> /dev/null" && do_facet mgs "! ls run-llog.sh &> /dev/null" && - skip_env "missing subtest run-llog.sh" && return + skip_env "missing subtest run-llog.sh" log "$TEST60_HEAD - from kernel mode" do_facet mgs "$LCTL set_param debug=warning; $LCTL dk > /dev/null" - do_facet mgs "sh run-llog.sh" || error "run-llog.sh failed" + do_facet mgs "bash run-llog.sh" || error "run-llog.sh failed" do_facet mgs $LCTL dk > $TMP/$tfile # LU-6388: test llog_reader local llog_reader=$(do_facet mgs "which llog_reader 2> /dev/null") llog_reader=${llog_reader:-$LUSTRE/utils/llog_reader} [ -z $(do_facet mgs ls -d $llog_reader 2> /dev/null) ] && - skip_env "missing llog_reader" && return + skip_env "missing llog_reader" local fstype=$(facet_fstype mgs) [ $fstype != ldiskfs -a $fstype != zfs ] && - skip_env "Only for ldiskfs or zfs type mgs" && return + skip_env "Only for ldiskfs or zfs type mgs" local mntpt=$(facet_mntpt mgs) local mgsdev=$(mgsdevname 1) @@ -6067,7 +6236,7 @@ test_60a() { run_test 60a "llog_test run from kernel module and test llog_reader" test_60aa() { - remote_mgs_nodsh && skip "remote MGS with nodsh" && return + remote_mgs_nodsh && skip "remote MGS with nodsh" # test old logid format if [ $(lustre_version_code mgs) -le $(version_code 3.1.53) ]; then @@ -6084,28 +6253,56 @@ test_60aa() { } run_test 60aa "llog_print works with FIDs and simple names" +test_60ab() { + # test llog_print with params + + [[ $(lustre_version_code $SINGLEMDS) -gt $(version_code 2.11.51) ]] || + skip "Need server version greater than 2.11.51" + + local yaml + local orig_val + + orig_val=$(do_facet mgs $LCTL get_param jobid_name) + do_facet mgs $LCTL set_param -P jobid_name="testname" + + yaml=$(do_facet mgs $LCTL --device MGS llog_print params | + grep jobid_name | tail -n 1) + + local param=`awk '{ print $10 }' <<< "$yaml"` + local val=`awk '{ print $12 }' <<< "$yaml"` + #return to the default + do_facet mgs $LCTL set_param -P jobid_name=$orig_val + [ $val = "testname" ] || error "bad value: $val" + [ $param = "jobid_name," ] || error "Bad param: $param" +} +run_test 60ab "llog_print params output values from set_param -P" + test_60b() { # bug 6411 - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return + [ $PARALLEL == "yes" ] && skip "skip parallel run" + dmesg > $DIR/$tfile - LLOG_COUNT=$(dmesg | awk "/$TEST60_HEAD/ { marker = 1; from_marker = 0; } - /llog.test/ { - if (marker) - from_marker++ - from_begin++ - } - END { - if (marker) - print from_marker - else - print from_begin - }") - [[ $LLOG_COUNT -gt 100 ]] && + LLOG_COUNT=$(do_facet mgs dmesg | + awk "/$TEST60_HEAD/ { marker = 1; from_marker = 0; } + /llog_[a-z]*.c:[0-9]/ { + if (marker) + from_marker++ + from_begin++ + } + END { + if (marker) + print from_marker + else + print from_begin + }") + + [[ $LLOG_COUNT -gt 120 ]] && error "CDEBUG_LIMIT not limiting messages ($LLOG_COUNT)" || true } -run_test 60b "limit repeated messages from CERROR/CWARN ========" +run_test 60b "limit repeated messages from CERROR/CWARN" test_60c() { - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return + [ $PARALLEL == "yes" ] && skip "skip parallel run" + echo "create 5000 files" createmany -o $DIR/f60c- 5000 #define OBD_FAIL_MDS_LLOG_CREATE_FAILED 0x137 @@ -6116,9 +6313,9 @@ test_60c() { run_test 60c "unlink file when mds full" test_60d() { - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return - SAVEPRINTK=$(lctl get_param -n printk) + [ $PARALLEL == "yes" ] && skip "skip parallel run" + SAVEPRINTK=$(lctl get_param -n printk) # verify "lctl mark" is even working" MESSAGE="test message ID $RANDOM $$" $LCTL mark "$MESSAGE" || error "$LCTL mark failed" @@ -6136,8 +6333,9 @@ test_60d() { run_test 60d "test printk console message masking" test_60e() { - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return - remote_mds_nodsh && skip "remote MDS with nodsh" && return + [ $PARALLEL == "yes" ] && skip "skip parallel run" + remote_mds_nodsh && skip "remote MDS with nodsh" + touch $DIR/$tfile #define OBD_FAIL_MDS_LLOG_CREATE_FAILED2 0x15b do_facet mds1 lctl set_param fail_loc=0x15b @@ -6145,10 +6343,41 @@ test_60e() { } run_test 60e "no space while new llog is being created" +test_60g() { + local pid + + test_mkdir -c $MDSCOUNT $DIR/$tdir + $LFS setdirstripe -D -i -1 -c $MDSCOUNT $DIR/$tdir + + ( + local index=0 + while true; do + mkdir $DIR/$tdir/subdir$index 2>/dev/null + rmdir $DIR/$tdir/subdir$index 2>/dev/null + index=$((index + 1)) + done + ) & + + pid=$! + + for i in $(seq 100); do + # define OBD_FAIL_OSD_TXN_START 0x19a + do_facet mds1 lctl set_param fail_loc=0x8000019a + usleep 100 + done + + kill -9 $pid + + mkdir $DIR/$tdir/new || error "mkdir failed" + rmdir $DIR/$tdir/new || error "rmdir failed" +} +run_test 60g "transaction abort won't cause MDT hung" + test_61() { - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return + [ $PARALLEL == "yes" ] && skip "skip parallel run" + f="$DIR/f61" - dd if=/dev/zero of=$f bs=$(page_size) count=1 || error "dd $f failed" + dd if=/dev/zero of=$f bs=$PAGE_SIZE count=1 || error "dd $f failed" cancel_lru_locks osc $MULTIOP $f OSMWUc || error "$MULTIOP $f failed" sync @@ -6157,13 +6386,14 @@ run_test 61 "mmap() writes don't make sync hang ================" # bug 2330 - insufficient obd_match error checking causes LBUG test_62() { - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return - f="$DIR/f62" - echo foo > $f - cancel_lru_locks osc - lctl set_param fail_loc=0x405 - cat $f && error "cat succeeded, expect -EIO" - lctl set_param fail_loc=0 + [ $PARALLEL == "yes" ] && skip "skip parallel run" + + f="$DIR/f62" + echo foo > $f + cancel_lru_locks osc + lctl set_param fail_loc=0x405 + cat $f && error "cat succeeded, expect -EIO" + lctl set_param fail_loc=0 } # This test is now irrelevant (as of bug 10718 inclusion), we no longer # match every page all of the time. @@ -6173,8 +6403,10 @@ test_62() { # Though this test is irrelevant anymore, it helped to reveal some # other grant bugs (LU-4482), let's keep it. test_63a() { # was test_63 - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return - MAX_DIRTY_MB=`lctl get_param -n osc.*.max_dirty_mb | head -n 1` + [ $PARALLEL == "yes" ] && skip "skip parallel run" + + MAX_DIRTY_MB=$(lctl get_param -n osc.*.max_dirty_mb | head -n 1) + for i in `seq 10` ; do dd if=/dev/zero of=$DIR/f63 bs=8k & sleep 5 @@ -6189,7 +6421,8 @@ run_test 63a "Verify oig_wait interruption does not crash =======" # bug 2248 - async write errors didn't return to application on sync # bug 3677 - async write errors left page locked test_63b() { - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return + [ $PARALLEL == "yes" ] && skip "skip parallel run" + debugsave lctl set_param debug=-1 @@ -6211,14 +6444,16 @@ test_63b() { run_test 63b "async write errors should be returned to fsync ===" test_64a () { - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return + [ $PARALLEL == "yes" ] && skip "skip parallel run" + df $DIR lctl get_param -n osc.*[oO][sS][cC][_-]*.cur* | grep "[0-9]" } run_test 64a "verify filter grant calculations (in kernel) =====" test_64b () { - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return + [ $PARALLEL == "yes" ] && skip "skip parallel run" + sh oos.sh $MOUNT || error "oos.sh failed: $?" } run_test 64b "check out-of-space detection on client" @@ -6233,8 +6468,6 @@ run_test 64c "verify grant shrink" want_grant() { local tgt=$1 - local page_size=$(get_page_size client) - 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) @@ -6243,15 +6476,15 @@ want_grant() { local dirty_max_pages=$($LCTL get_param -n osc.${tgt}.max_dirty_mb) - dirty_max_pages=$((dirty_max_pages * 1024 * 1024 / page_size)) + dirty_max_pages=$((dirty_max_pages * 1024 * 1024 / PAGE_SIZE)) [[ $dirty_max_pages -gt $nrpages ]] && nrpages=$dirty_max_pages - local undirty=$((nrpages * page_size)) + local undirty=$((nrpages * PAGE_SIZE)) local max_extent_pages max_extent_pages=$($LCTL get_param osc.${tgt}.import | grep grant_max_extent_size | awk '{print $2}') - max_extent_pages=$((max_extent_pages / page_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 | @@ -6280,22 +6513,24 @@ grant_chunk() { test_64d() { [ $(lustre_version_code ost1) -lt $(version_code 2.10.56) ] && - skip "OST < 2.10.55 doesn't limit grants enough" && return 0 + skip "OST < 2.10.55 doesn't limit grants enough" local tgt=$($LCTL dl | grep "0000-osc-[^mM]" | awk '{print $4}') + local file=$DIR/$tfile [[ $($LCTL get_param osc.${tgt}.import | - grep "connect_flags:.*grant_param") ]] || \ - { skip "no grant_param connect flag"; return; } + grep "connect_flags:.*grant_param") ]] || + skip "no grant_param connect flag" local olddebug=$($LCTL get_param -n debug 2> /dev/null) $LCTL set_param debug="$OLDDEBUG" 2> /dev/null || true local max_cur_granted=$(($(want_grant $tgt) + $(grant_chunk $tgt))) + stack_trap "rm -f $file" EXIT - $SETSTRIPE $DIR/$tfile -i 0 -c 1 - dd if=/dev/zero of=$DIR/$tfile bs=1M count=1000 & + $SETSTRIPE $file -i 0 -c 1 + dd if=/dev/zero of=$file bs=1M count=1000 & ddpid=$! while true @@ -6319,7 +6554,8 @@ run_test 64d "check grant limit exceed" # bug 1414 - set/get directories' stripe info test_65a() { - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return + [ $PARALLEL == "yes" ] && skip "skip parallel run" + test_mkdir $DIR/$tdir touch $DIR/$tdir/f1 $LVERIFY $DIR/$tdir $DIR/$tdir/f1 || error "lverify failed" @@ -6327,7 +6563,8 @@ test_65a() { run_test 65a "directory with no stripe info" test_65b() { - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return + [ $PARALLEL == "yes" ] && skip "skip parallel run" + test_mkdir $DIR/$tdir local STRIPESIZE=$($GETSTRIPE -S $DIR/$tdir) @@ -6339,8 +6576,9 @@ test_65b() { run_test 65b "directory setstripe -S stripe_size*2 -i 0 -c 1" test_65c() { - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return - [ $OSTCOUNT -lt 2 ] && skip "need at least 2 OSTs" && return + [ $PARALLEL == "yes" ] && skip "skip parallel run" + [ $OSTCOUNT -lt 2 ] && skip_env "need at least 2 OSTs" + test_mkdir $DIR/$tdir local stripesize=$($GETSTRIPE -S $DIR/$tdir) @@ -6352,7 +6590,8 @@ test_65c() { run_test 65c "directory setstripe -S stripe_size*4 -i 1 -c $((OSTCOUNT-1))" test_65d() { - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return + [ $PARALLEL == "yes" ] && skip "skip parallel run" + test_mkdir $DIR/$tdir local STRIPECOUNT=$($GETSTRIPE -c $DIR/$tdir) local STRIPESIZE=$($GETSTRIPE -S $DIR/$tdir) @@ -6373,11 +6612,12 @@ test_65d() { run_test 65d "directory setstripe -S stripe_size -c stripe_count" test_65e() { - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return + [ $PARALLEL == "yes" ] && skip "skip parallel run" + test_mkdir $DIR/$tdir $SETSTRIPE $DIR/$tdir || error "setstripe" - $GETSTRIPE -v $DIR/$tdir | grep "Default" || + $GETSTRIPE -v $DIR/$tdir | grep "Default" || error "no stripe info failed" touch $DIR/$tdir/f6 $LVERIFY $DIR/$tdir $DIR/$tdir/f6 || error "lverify failed" @@ -6385,14 +6625,16 @@ test_65e() { run_test 65e "directory setstripe defaults" test_65f() { - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return + [ $PARALLEL == "yes" ] && skip "skip parallel run" + test_mkdir $DIR/${tdir}f $RUNAS $SETSTRIPE $DIR/${tdir}f && error "setstripe succeeded" || true } run_test 65f "dir setstripe permission (should return error) ===" test_65g() { - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return + [ $PARALLEL == "yes" ] && skip "skip parallel run" + test_mkdir $DIR/$tdir local STRIPESIZE=$($GETSTRIPE -S $DIR/$tdir) @@ -6405,7 +6647,8 @@ test_65g() { run_test 65g "directory setstripe -d" test_65h() { - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return + [ $PARALLEL == "yes" ] && skip "skip parallel run" + test_mkdir $DIR/$tdir local STRIPESIZE=$($GETSTRIPE -S $DIR/$tdir) @@ -6418,7 +6661,7 @@ test_65h() { run_test 65h "directory stripe info inherit ====================" test_65i() { - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return + [ $PARALLEL == "yes" ] && skip "skip parallel run" save_layout_restore_at_exit $MOUNT @@ -6437,7 +6680,7 @@ test_65i() { run_test 65i "various tests to set root directory striping" test_65j() { # bug6367 - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return + [ $PARALLEL == "yes" ] && skip "skip parallel run" sync; sleep 1 @@ -6467,9 +6710,9 @@ cleaup_65k() { } test_65k() { # bug11679 - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return - [[ $OSTCOUNT -lt 2 ]] && skip_env "needs >= 2 OSTs" && return - remote_mds_nodsh && skip "remote MDS with nodsh" && return + [ $PARALLEL == "yes" ] && skip "skip parallel run" + [[ $OSTCOUNT -lt 2 ]] && skip_env "needs >= 2 OSTs" + remote_mds_nodsh && skip "remote MDS with nodsh" local disable_precreate=true [ $(lustre_version_code $SINGLEMDS) -le $(version_code 2.8.54) ] && @@ -6529,7 +6772,8 @@ test_65k() { # bug11679 run_test 65k "validate manual striping works properly with deactivated OSCs" test_65l() { # bug 12836 - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return + [ $PARALLEL == "yes" ] && skip "skip parallel run" + test_mkdir -p $DIR/$tdir/test_dir $SETSTRIPE -c -1 $DIR/$tdir/test_dir $LFS find -mtime -1 $DIR/$tdir >/dev/null @@ -6548,7 +6792,8 @@ run_test 65m "normal user can't set filesystem default stripe" # bug 2543 - update blocks count on client test_66() { - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return + [ $PARALLEL == "yes" ] && skip "skip parallel run" + COUNT=${COUNT:-8} dd if=/dev/zero of=$DIR/f66 bs=1k count=$COUNT sync; sync_all_data; sync; sync_all_data @@ -6569,8 +6814,8 @@ swap_used() { # bug5265, obdfilter oa2dentry return -ENOENT # #define OBD_FAIL_SRV_ENOENT 0x217 test_69() { - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return - remote_ost_nodsh && skip "remote OST with nodsh" && return + [ $PARALLEL == "yes" ] && skip "skip parallel run" + remote_ost_nodsh && skip "remote OST with nodsh" f="$DIR/$tfile" $SETSTRIPE -c 1 -i 0 $f @@ -6603,15 +6848,13 @@ test_71() { run_test 71 "Running dbench on lustre (don't segment fault) ====" test_72a() { # bug 5695 - Test that on 2.6 remove_suid works properly - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return + [ $PARALLEL == "yes" ] && skip "skip parallel run" [ "$RUNAS_ID" = "$UID" ] && - skip_env "RUNAS_ID = UID = $UID -- skipping" && return - + skip_env "RUNAS_ID = UID = $UID -- skipping" # Check that testing environment is properly set up. Skip if not - FAIL_ON_ERROR=false check_runas_id_ret $RUNAS_ID $RUNAS_GID $RUNAS || { + FAIL_ON_ERROR=false check_runas_id_ret $RUNAS_ID $RUNAS_GID $RUNAS || skip_env "User $RUNAS_ID does not exist - skipping" - return 0 - } + touch $DIR/$tfile chmod 777 $DIR/$tfile chmod ug+s $DIR/$tfile @@ -6631,17 +6874,15 @@ run_test 72a "Test that remove suid works properly (bug5695) ====" test_72b() { # bug 24226 -- keep mode setting when size is not changing local perm - [ "$RUNAS_ID" = "$UID" ] && \ - skip_env "RUNAS_ID = UID = $UID -- skipping" && return - [ "$RUNAS_ID" -eq 0 ] && \ - skip_env "RUNAS_ID = 0 -- skipping" && return - - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return + [ "$RUNAS_ID" = "$UID" ] && + skip_env "RUNAS_ID = UID = $UID -- skipping" + [ "$RUNAS_ID" -eq 0 ] && + skip_env "RUNAS_ID = 0 -- skipping" + [ $PARALLEL == "yes" ] && skip "skip parallel run" # Check that testing environment is properly set up. Skip if not - FAIL_ON_ERROR=false check_runas_id_ret $RUNAS_ID $RUNAS_ID $RUNAS || { + FAIL_ON_ERROR=false check_runas_id_ret $RUNAS_ID $RUNAS_ID $RUNAS || skip_env "User $RUNAS_ID does not exist - skipping" - return 0 - } + touch $DIR/${tfile}-f{g,u} test_mkdir $DIR/${tfile}-dg test_mkdir $DIR/${tfile}-du @@ -6660,7 +6901,8 @@ run_test 72b "Test that we keep mode setting if without file data changed (bug 2 # bug 3462 - multiple simultaneous MDC requests test_73() { - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return + [ $PARALLEL == "yes" ] && skip "skip parallel run" + test_mkdir $DIR/d73-1 test_mkdir $DIR/d73-2 multiop_bg_pause $DIR/d73-1/f73-1 O_c || return 1 @@ -6688,12 +6930,13 @@ test_73() { run_test 73 "multiple MDC requests (should not deadlock)" test_74a() { # bug 6149, 6184 - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return + [ $PARALLEL == "yes" ] && skip "skip parallel run" + + touch $DIR/f74a #define OBD_FAIL_LDLM_ENQUEUE_OLD_EXPORT 0x30e # # very important to OR with OBD_FAIL_ONCE (0x80000000) -- otherwise it # will spin in a tight reconnection loop - touch $DIR/f74a $LCTL set_param fail_loc=0x8000030e # get any lock that won't be difficult - lookup works. ls $DIR/f74a @@ -6704,7 +6947,8 @@ test_74a() { # bug 6149, 6184 run_test 74a "ldlm_enqueue freed-export error path, ls (shouldn't LBUG)" test_74b() { # bug 13310 - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return + [ $PARALLEL == "yes" ] && skip "skip parallel run" + #define OBD_FAIL_LDLM_ENQUEUE_OLD_EXPORT 0x30e # # very important to OR with OBD_FAIL_ONCE (0x80000000) -- otherwise it @@ -6719,7 +6963,8 @@ test_74b() { # bug 13310 run_test 74b "ldlm_enqueue freed-export error path, touch (shouldn't LBUG)" test_74c() { - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return + [ $PARALLEL == "yes" ] && skip "skip parallel run" + #define OBD_FAIL_LDLM_NEW_LOCK $LCTL set_param fail_loc=0x319 touch $DIR/$tfile && error "touch successful" @@ -6733,8 +6978,10 @@ num_inodes() { } test_76() { # Now for bug 20433, added originally in bug 1443 - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return + [ $PARALLEL == "yes" ] && skip "skip parallel run" + local CPUS=$(getconf _NPROCESSORS_ONLN 2>/dev/null) + cancel_lru_locks osc BEFORE_INODES=$(num_inodes) echo "before inodes: $BEFORE_INODES" @@ -6778,8 +7025,8 @@ set_checksums() export ORIG_CSUM_TYPE="`lctl get_param -n osc.*osc-[^mM]*.checksum_type | sed 's/.*\[\(.*\)\].*/\1/g' | head -n1`" -CKSUM_TYPES=${CKSUM_TYPES:-"crc32 adler"} -[ "$ORIG_CSUM_TYPE" = "crc32c" ] && CKSUM_TYPES="$CKSUM_TYPES crc32c" +CKSUM_TYPES=${CKSUM_TYPES:-$(lctl get_param -n osc.*osc-[^mM]*.checksum_type | + tr -d [] | head -n1)} set_checksum_type() { lctl set_param -n osc.*osc-[^mM]*.checksum_type $1 @@ -6794,8 +7041,9 @@ setup_f77() { } test_77a() { # bug 10889 - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return - $GSS && skip "could not run with gss" && return + [ $PARALLEL == "yes" ] && skip "skip parallel run" + $GSS && skip_env "could not run with gss" + [ ! -f $F77_TMP ] && setup_f77 set_checksums 1 dd if=$F77_TMP of=$DIR/$tfile bs=1M count=$F77SZ || error "dd error" @@ -6805,8 +7053,9 @@ test_77a() { # bug 10889 run_test 77a "normal checksum read/write operation" test_77b() { # bug 10889 - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return - $GSS && skip "could not run with gss" && return + [ $PARALLEL == "yes" ] && skip "skip parallel run" + $GSS && skip_env "could not run with gss" + [ ! -f $F77_TMP ] && setup_f77 #define OBD_FAIL_OSC_CHECKSUM_SEND 0x409 $LCTL set_param fail_loc=0x80000409 @@ -6836,16 +7085,15 @@ cleanup_77c() { $LCTL set_param osc.*osc-[^mM]*.checksum_dump=0 $check_ost && do_facet ost1 $LCTL set_param obdfilter.*-OST*.checksum_dump=0 - [ -n $osc_file_prefix ] && rm -f ${osc_file_prefix}* - $check_ost && [ -n $ost_file_prefix ] && + [ -n "$osc_file_prefix" ] && rm -f ${osc_file_prefix}* + $check_ost && [ -n "$ost_file_prefix" ] && do_facet ost1 rm -f ${ost_file_prefix}\* } test_77c() { - remote_ost_nodsh && skip "remote OST with nodsh" && return - - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return - $GSS && skip "could not run with gss" && return + [ $PARALLEL == "yes" ] && skip "skip parallel run" + $GSS && skip_env "could not run with gss" + remote_ost_nodsh && skip "remote OST with nodsh" local bad1 local osc_file_prefix @@ -6908,7 +7156,7 @@ test_77c() { [[ "$orig_cksum" == "$dump_cksum" ]] || error "dump content does not match on Client" - $check_ost || { skip "No need to check cksum dump on OSS"; return 0; } + $check_ost || skip "No need to check cksum dump on OSS" # check cksum dump on OSS ost_file=$(do_facet ost1 ls ${ost_file_prefix}\*) @@ -6923,8 +7171,9 @@ test_77c() { run_test 77c "checksum error on client read with debug" test_77d() { # bug 10889 - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return - $GSS && skip "could not run with gss" && return + [ $PARALLEL == "yes" ] && skip "skip parallel run" + $GSS && skip_env "could not run with gss" + #define OBD_FAIL_OSC_CHECKSUM_SEND 0x409 $LCTL set_param fail_loc=0x80000409 set_checksums 1 @@ -6945,8 +7194,9 @@ test_77d() { # bug 10889 run_test 77d "checksum error on OST direct write, read" test_77f() { # bug 10889 - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return - $GSS && skip "could not run with gss" && return + [ $PARALLEL == "yes" ] && skip "skip parallel run" + $GSS && skip_env "could not run with gss" + set_checksums 1 for algo in $CKSUM_TYPES; do cancel_lru_locks osc @@ -6963,17 +7213,20 @@ test_77f() { # bug 10889 run_test 77f "repeat checksum error on write (expect error)" test_77g() { # bug 10889 - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return - $GSS && skip "could not run with gss" && return - remote_ost_nodsh && skip "remote OST with nodsh" && return + [ $PARALLEL == "yes" ] && skip "skip parallel run" + $GSS && skip_env "could not run with gss" + remote_ost_nodsh && skip "remote OST with nodsh" [ ! -f $F77_TMP ] && setup_f77 - $SETSTRIPE -c 1 -i 0 $DIR/$tfile + local file=$DIR/$tfile + stack_trap "rm -f $file" EXIT + + $SETSTRIPE -c 1 -i 0 $file #define OBD_FAIL_OST_CHECKSUM_RECEIVE 0x21a do_facet ost1 lctl set_param fail_loc=0x8000021a set_checksums 1 - dd if=$F77_TMP of=$DIR/$tfile bs=1M count=$F77SZ || + dd if=$F77_TMP of=$file bs=1M count=$F77SZ || error "write error: rc=$?" do_facet ost1 lctl set_param fail_loc=0 set_checksums 0 @@ -6982,33 +7235,51 @@ test_77g() { # bug 10889 #define OBD_FAIL_OST_CHECKSUM_SEND 0x21b do_facet ost1 lctl set_param fail_loc=0x8000021b set_checksums 1 - cmp $F77_TMP $DIR/$tfile || error "file compare failed" + cmp $F77_TMP $file || error "file compare failed" do_facet ost1 lctl set_param fail_loc=0 set_checksums 0 } run_test 77g "checksum error on OST write, read" -test_77j() { # bug 13805 - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return - $GSS && skip "could not run with gss" && return - #define OBD_FAIL_OSC_CKSUM_ADLER_ONLY 0x40c - lctl set_param fail_loc=0x40c - remount_client $MOUNT - lctl set_param fail_loc=0 - # wait async osc connect to finish and reflect updated state value +test_77k() { # LU-10906 + [ $PARALLEL == "yes" ] && skip "skip parallel run" + $GSS && skip_env "could not run with gss" + + local cksum_param="osc.$FSNAME*.checksums" + local get_checksum="$LCTL get_param -n $cksum_param | head -n1" + local checksum local i - for (( i=0; i < OSTCOUNT; i++ )) ; do - wait_osc_import_state client ost$((i+1)) FULL + + [ "$ORIG_CSUM" ] || ORIG_CSUM=$(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 + + for i in 0 1; do + do_facet mgs $LCTL set_param -P $cksum_param=$i || + error "failed to set checksum=$i on MGS" + wait_update $HOSTNAME "$get_checksum" $i + #remount + echo "remount client, checksum should be $i" + remount_client $MOUNT || "failed to remount client" + checksum=$(eval $get_checksum) + [ $checksum -eq $i ] || error "checksum($checksum) != $i" done - for VALUE in $(lctl get_param osc.*osc-[^mM]*.checksum_type); do - PARAM=$(echo ${VALUE[0]} | cut -d "=" -f1) - algo=$(lctl get_param -n $PARAM | sed 's/.*\[\(.*\)\].*/\1/g') - [ "$algo" = "adler" ] || error "algo set to $algo instead of adler" + for opt in "checksum" "nochecksum"; do + #remount with mount option + echo "remount client with option $opt, checksum should be $i" + umount_client $MOUNT || "failed to umount client" + mount_client $MOUNT "$MOUNT_OPTS,$opt" || + "failed to mount client with option '$opt'" + checksum=$(eval $get_checksum) + [ $checksum -eq $i ] || error "checksum($checksum) != $i" + i=$((i - 1)) done - remount_client $MOUNT + + remount_client $MOUNT || "failed to remount client" } -run_test 77j "client only supporting ADLER32" +run_test 77k "enable/disable checksum correctly" [ "$ORIG_CSUM" ] && set_checksums $ORIG_CSUM || true rm -f $F77_TMP @@ -7020,8 +7291,8 @@ cleanup_test_78() { } test_78() { # bug 10901 - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return - remote_ost || { skip_env "local OST" && return; } + [ $PARALLEL == "yes" ] && skip "skip parallel run" + remote_ost || skip_env "local OST" NSEQ=5 F78SIZE=$(($(awk '/MemFree:/ { print $2 }' /proc/meminfo) / 1024)) @@ -7046,7 +7317,7 @@ test_78() { # bug 10901 head -n1) echo "Smallest OST: $SMALLESTOST" [[ $SMALLESTOST -lt 10240 ]] && - skip "too small OSTSIZE, useless to run large O_DIRECT test" && return 0 + skip "too small OSTSIZE, useless to run large O_DIRECT test" trap cleanup_test_78 EXIT @@ -7067,12 +7338,13 @@ test_78() { # bug 10901 run_test 78 "handle large O_DIRECT writes correctly ============" test_79() { # bug 12743 - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return + [ $PARALLEL == "yes" ] && skip "skip parallel run" + wait_delete_completed - BKTOTAL=$(calc_osc_kbytes kbytestotal) - BKFREE=$(calc_osc_kbytes kbytesfree) - BKAVAIL=$(calc_osc_kbytes kbytesavail) + BKTOTAL=$(calc_osc_kbytes kbytestotal) + BKFREE=$(calc_osc_kbytes kbytesfree) + BKAVAIL=$(calc_osc_kbytes kbytesavail) STRING=`df -P $MOUNT | tail -n 1 | awk '{print $2","$3","$4}'` DFTOTAL=`echo $STRING | cut -d, -f1` @@ -7098,17 +7370,19 @@ test_79() { # bug 12743 run_test 79 "df report consistency check =======================" test_80() { # bug 10718 - remote_ost_nodsh && skip "remote OST with nodsh" && return - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return - # relax strong synchronous semantics for slow backends like ZFS - local soc="obdfilter.*.sync_on_lock_cancel" - local soc_old=$(do_facet ost1 lctl get_param -n $soc | head -n1) - local hosts= - if [ "$soc_old" != "never" -a "$(facet_fstype ost1)" != "ldiskfs" ]; then - hosts=$(for host in $(seq -f "ost%g" 1 $OSTCOUNT); do - facet_active_host $host; done | sort -u) - do_nodes $hosts lctl set_param $soc=never - fi + remote_ost_nodsh && skip "remote OST with nodsh" + [ $PARALLEL == "yes" ] && skip "skip parallel run" + + # relax strong synchronous semantics for slow backends like ZFS + local soc="obdfilter.*.sync_on_lock_cancel" + local soc_old=$(do_facet ost1 lctl get_param -n $soc | head -n1) + local hosts= + if [ "$soc_old" != "never" ] && + [ "$(facet_fstype ost1)" != "ldiskfs" ]; then + hosts=$(for host in $(seq -f "ost%g" 1 $OSTCOUNT); do + facet_active_host $host; done | sort -u) + do_nodes $hosts lctl set_param $soc=never + fi dd if=/dev/zero of=$DIR/$tfile bs=1M count=1 seek=1M sync; sleep 1; sync @@ -7127,37 +7401,39 @@ test_80() { # bug 10718 run_test 80 "Page eviction is equally fast at high offsets too ====" test_81a() { # LU-456 - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return - remote_ost_nodsh && skip "remote OST with nodsh" && return - # define OBD_FAIL_OST_MAPBLK_ENOSPC 0x228 - # MUST OR with the OBD_FAIL_ONCE (0x80000000) - do_facet ost1 lctl set_param fail_loc=0x80000228 - - # write should trigger a retry and success - $SETSTRIPE -i 0 -c 1 $DIR/$tfile - $MULTIOP $DIR/$tfile oO_CREAT:O_RDWR:O_SYNC:w4096c - RC=$? - if [ $RC -ne 0 ] ; then - error "write should success, but failed for $RC" - fi + [ $PARALLEL == "yes" ] && skip "skip parallel run" + remote_ost_nodsh && skip "remote OST with nodsh" + + # define OBD_FAIL_OST_MAPBLK_ENOSPC 0x228 + # MUST OR with the OBD_FAIL_ONCE (0x80000000) + do_facet ost1 lctl set_param fail_loc=0x80000228 + + # write should trigger a retry and success + $SETSTRIPE -i 0 -c 1 $DIR/$tfile + $MULTIOP $DIR/$tfile oO_CREAT:O_RDWR:O_SYNC:w4096c + RC=$? + if [ $RC -ne 0 ] ; then + error "write should success, but failed for $RC" + fi } run_test 81a "OST should retry write when get -ENOSPC ===============" test_81b() { # LU-456 - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return - remote_ost_nodsh && skip "remote OST with nodsh" && return - # define OBD_FAIL_OST_MAPBLK_ENOSPC 0x228 - # Don't OR with the OBD_FAIL_ONCE (0x80000000) - do_facet ost1 lctl set_param fail_loc=0x228 - - # write should retry several times and return -ENOSPC finally - $SETSTRIPE -i 0 -c 1 $DIR/$tfile - $MULTIOP $DIR/$tfile oO_CREAT:O_RDWR:O_SYNC:w4096c - RC=$? - ENOSPC=28 - if [ $RC -ne $ENOSPC ] ; then - error "dd should fail for -ENOSPC, but succeed." - fi + [ $PARALLEL == "yes" ] && skip "skip parallel run" + remote_ost_nodsh && skip "remote OST with nodsh" + + # define OBD_FAIL_OST_MAPBLK_ENOSPC 0x228 + # Don't OR with the OBD_FAIL_ONCE (0x80000000) + do_facet ost1 lctl set_param fail_loc=0x228 + + # write should retry several times and return -ENOSPC finally + $SETSTRIPE -i 0 -c 1 $DIR/$tfile + $MULTIOP $DIR/$tfile oO_CREAT:O_RDWR:O_SYNC:w4096c + RC=$? + ENOSPC=28 + if [ $RC -ne $ENOSPC ] ; then + error "dd should fail for -ENOSPC, but succeed." + fi } run_test 81b "OST should return -ENOSPC when retry still fails =======" @@ -7185,6 +7461,7 @@ run_test 82 "Basic grouplock test" test_83() { local sfile="/boot/System.map-$(uname -r)" + [ ! -f $sfile ] && skip "No $sfile found" # define OBD_FAIL_LLITE_PTASK_IO_FAIL 0x140d $LCTL set_param fail_loc=0x140d cp $sfile $DIR/$tfile || error "write failed" @@ -7195,8 +7472,8 @@ test_83() { run_test 83 "Short write in ptask ===============================" test_99() { - [ -z "$(which cvs 2>/dev/null)" ] && skip_env "could not find cvs" && - return + [ -z "$(which cvs 2>/dev/null)" ] && skip_env "could not find cvs" + test_mkdir $DIR/$tdir.cvsroot chown $RUNAS_ID $DIR/$tdir.cvsroot @@ -7226,15 +7503,13 @@ test_99() { run_test 99 "cvs strange file/directory operations" test_100() { - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return + [ $PARALLEL == "yes" ] && skip "skip parallel run" [[ "$NETTYPE" =~ tcp ]] || - { skip "TCP secure port test, not useful for NETTYPE=$NETTYPE" && - return ; } - - remote_ost_nodsh && skip "remote OST with nodsh" && return - remote_mds_nodsh && skip "remote MDS with nodsh" && return + skip_env "TCP secure port test, not useful for NETTYPE=$NETTYPE" + remote_ost_nodsh && skip "remote OST with nodsh" + remote_mds_nodsh && skip "remote MDS with nodsh" remote_servers || - { skip "useless for local single node setup" && return; } + skip "useless for local single node setup" netstat -tna | ( rc=1; while read PROT SND RCV LOCAL REMOTE STAT; do [ "$PROT" != "tcp" ] && continue @@ -7278,8 +7553,9 @@ cleanup_101a() { } test_101a() { - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return - [ $MDSCOUNT -ge 2 ] && skip "needs < 2 MDTs" && return #LU-4322 + [ $PARALLEL == "yes" ] && skip "skip parallel run" + [ $MDSCOUNT -ge 2 ] && skip_env "needs < 2 MDTs" #LU-4322 + local s local discard local nreads=10000 @@ -7367,10 +7643,12 @@ ra_check_101() { } test_101b() { - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return - [[ $OSTCOUNT -lt 2 ]] && skip_env "needs >= 2 OSTs" && return + [ $PARALLEL == "yes" ] && skip "skip parallel run" + [[ $OSTCOUNT -lt 2 ]] && skip_env "needs >= 2 OSTs" + local STRIPE_SIZE=1048576 local STRIDE_SIZE=$((STRIPE_SIZE*OSTCOUNT)) + if [ $SLOW == "yes" ]; then local FILE_LENGTH=$((STRIDE_SIZE * 64)) else @@ -7400,7 +7678,8 @@ test_101b() { run_test 101b "check stride-io mode read-ahead =================" test_101c() { - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return + [ $PARALLEL == "yes" ] && skip "skip parallel run" + local STRIPE_SIZE=1048576 local FILE_LENGTH=$((STRIPE_SIZE*100)) local nreads=10000 @@ -7438,14 +7717,15 @@ set_read_ahead() { } test_101d() { - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return + [ $PARALLEL == "yes" ] && skip "skip parallel run" + local file=$DIR/$tfile local sz_MB=${FILESIZE_101d:-500} local ra_MB=${READAHEAD_MB:-40} local free_MB=$(($(df -P $DIR | tail -n 1 | awk '{ print $4 }') / 1024)) [ $free_MB -lt $sz_MB ] && - skip "Need free space ${sz_MB}M, have ${free_MB}M" && return + skip "Need free space ${sz_MB}M, have ${free_MB}M" echo "Create test file $file size ${sz_MB}M, ${free_MB}M free" $SETSTRIPE -c -1 $file || error "setstripe failed" @@ -7481,7 +7761,8 @@ test_101d() { run_test 101d "file read with and without read-ahead enabled" test_101e() { - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return + [ $PARALLEL == "yes" ] && skip "skip parallel run" + local file=$DIR/$tfile local size_KB=500 #KB local count=100 @@ -7490,7 +7771,7 @@ test_101e() { local free_KB=$(df -P $DIR | tail -n 1 | awk '{ print $4 }') local need_KB=$((count * size_KB)) [[ $free_KB -le $need_KB ]] && - skip_env "Need free space $need_KB, have $free_KB" && return + skip_env "Need free space $need_KB, have $free_KB" echo "Creating $count ${size_KB}K test files" for ((i = 0; i < $count; i++)); do @@ -7520,7 +7801,7 @@ test_101e() { run_test 101e "check read-ahead for small read(1k) for small files(500k)" test_101f() { - which iozone || { skip "no iozone installed" && return; } + which iozone || skip_env "no iozone installed" local old_debug=$($LCTL get_param debug) old_debug=${old_debug#*=} @@ -7552,7 +7833,8 @@ run_test 101f "check mmap read performance" test_101g_brw_size_test() { local mb=$1 - local pages=$((mb * 1048576 / $(page_size))) + local pages=$((mb * 1048576 / PAGE_SIZE)) + local file=$DIR/$tfile $LCTL set_param osc.*.max_pages_per_rpc=${mb}M || { error "unable to set max_pages_per_rpc=${mb}M"; return 1; } @@ -7561,14 +7843,15 @@ test_101g_brw_size_test() { return 2 done + stack_trap "rm -f $file" EXIT $LCTL set_param -n osc.*.rpc_stats=0 # 10 RPCs should be enough for the test local count=10 - dd if=/dev/zero of=$DIR/$tfile bs=${mb}M count=$count || + dd if=/dev/zero of=$file bs=${mb}M count=$count || { error "dd write ${mb} MB blocks failed"; return 3; } cancel_lru_locks osc - dd of=/dev/null if=$DIR/$tfile bs=${mb}M count=$count || + dd of=/dev/null if=$file bs=${mb}M count=$count || { error "dd write ${mb} MB blocks failed"; return 4; } # calculate number of full-sized read and write RPCs @@ -7585,7 +7868,7 @@ test_101g_brw_size_test() { } test_101g() { - remote_ost_nodsh && skip "remote OST with nodsh" && return + remote_ost_nodsh && skip "remote OST with nodsh" local rpcs local osts=$(get_facets OST) @@ -7663,17 +7946,16 @@ cleanup_test102() { } test_102a() { - local testfile=$DIR/$tfile - - touch $testfile - - [ "$UID" != 0 ] && skip_env "must run as root" && return + [ "$UID" != 0 ] && skip "must run as root" [ -z "$(lctl get_param -n mdc.*-mdc-*.connect_flags | grep xattr)" ] && - skip_env "must have user_xattr" && return + skip_env "must have user_xattr" [ -z "$(which setfattr 2>/dev/null)" ] && - skip_env "could not find setfattr" && return + skip_env "could not find setfattr" + + local testfile=$DIR/$tfile + touch $testfile echo "set/get xattr..." setfattr -n trusted.name1 -v value1 $testfile || error "setfattr -n trusted.name1=value1 $testfile failed" @@ -7726,11 +8008,11 @@ run_test 102a "user xattr test ==================================" test_102b() { [ -z "$(which setfattr 2>/dev/null)" ] && - skip_env "could not find setfattr" && return + skip_env "could not find setfattr" + [[ $OSTCOUNT -lt 2 ]] && skip_env "needs >= 2 OSTs" # b10930: get/set/list trusted.lov xattr echo "get/set/list trusted.lov xattr ..." - [[ $OSTCOUNT -lt 2 ]] && skip_env "needs >= 2 OSTs" && return local testfile=$DIR/$tfile $SETSTRIPE -S 65536 -i 1 -c $OSTCOUNT $testfile || error "setstripe failed" @@ -7757,11 +8039,11 @@ run_test 102b "getfattr/setfattr for trusted.lov EAs ============" test_102c() { [ -z "$(which setfattr 2>/dev/null)" ] && - skip_env "could not find setfattr" && return + skip_env "could not find setfattr" + [[ $OSTCOUNT -lt 2 ]] && skip_env "needs >= 2 OSTs" # b10930: get/set/list lustre.lov xattr echo "get/set/list lustre.lov xattr ..." - [[ $OSTCOUNT -lt 2 ]] && skip_env "needs >= 2 OSTs" && return test_mkdir $DIR/$tdir chown $RUNAS_ID $DIR/$tdir local testfile=$DIR/$tdir/$tfile @@ -7818,8 +8100,9 @@ have_xattrs_include() { } test_102d() { - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return - [[ $OSTCOUNT -lt 2 ]] && skip_env "needs >= 2 OSTs" && return + [ $PARALLEL == "yes" ] && skip "skip parallel run" + [[ $OSTCOUNT -lt 2 ]] && skip_env "needs >= 2 OSTs" + XINC=$(have_xattrs_include) setup_test102 tar xf $TMP/f102.tar -C $DIR/$tdir --xattrs $XINC @@ -7829,8 +8112,9 @@ test_102d() { run_test 102d "tar restore stripe info from tarfile,not keep osts" test_102f() { - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return - [[ $OSTCOUNT -lt 2 ]] && skip_env "needs >= 2 OSTs" && return + [ $PARALLEL == "yes" ] && skip "skip parallel run" + [[ $OSTCOUNT -lt 2 ]] && skip_env "needs >= 2 OSTs" + XINC=$(have_xattrs_include) setup_test102 test_mkdir $DIR/$tdir.restore @@ -7843,21 +8127,19 @@ test_102f() { run_test 102f "tar copy files, not keep osts" grow_xattr() { - local xsize=${1:-1024} # in bytes - local file=$DIR/$tfile - [ -z "$(lctl get_param -n mdc.*.connect_flags | grep xattr)" ] && - skip "must have user_xattr" && return 0 + skip "must have user_xattr" [ -z "$(which setfattr 2>/dev/null)" ] && - skip_env "could not find setfattr" && return 0 + skip_env "could not find setfattr" [ -z "$(which getfattr 2>/dev/null)" ] && - skip_env "could not find getfattr" && return 0 - - touch $file + skip_env "could not find getfattr" + local xsize=${1:-1024} # in bytes + local file=$DIR/$tfile local value="$(generate_string $xsize)" - local xbig=trusted.big + + touch $file log "save $xbig on $file" setfattr -n $xbig -v $value $file || error "saving $xbig on $file failed" @@ -7889,14 +8171,16 @@ test_102h() { # bug 15777 run_test 102h "grow xattr from inside inode to external block" test_102ha() { - large_xattr_enabled || { skip "large_xattr disabled" && return; } + large_xattr_enabled || skip_env "ea_inode feature disabled" + grow_xattr $(max_xattr_size) } run_test 102ha "grow xattr from inside inode to external inode" test_102i() { # bug 17038 [ -z "$(which getfattr 2>/dev/null)" ] && - skip "could not find getfattr" && return + skip "could not find getfattr" + touch $DIR/$tfile ln -s $DIR/$tfile $DIR/${tfile}link getfattr -n trusted.lov $DIR/$tfile || @@ -7909,8 +8193,9 @@ test_102i() { # bug 17038 run_test 102i "lgetxattr test on symbolic link ============" test_102j() { - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return - [[ $OSTCOUNT -lt 2 ]] && skip_env "needs >= 2 OSTs" && return + [ $PARALLEL == "yes" ] && skip "skip parallel run" + [[ $OSTCOUNT -lt 2 ]] && skip_env "needs >= 2 OSTs" + XINC=$(have_xattrs_include) setup_test102 "$RUNAS" chown $RUNAS_ID $DIR/$tdir @@ -7922,7 +8207,8 @@ run_test 102j "non-root tar restore stripe info from tarfile, not keep osts ===" test_102k() { [ -z "$(which setfattr 2>/dev/null)" ] && - skip "could not find setfattr" && return + skip "could not find setfattr" + touch $DIR/$tfile # b22187 just check that does not crash for regular file. setfattr -n trusted.lov $DIR/$tfile @@ -7950,7 +8236,7 @@ run_test 102k "setfattr without parameter of value shouldn't cause a crash" test_102l() { [ -z "$(which getfattr 2>/dev/null)" ] && - skip "could not find getfattr" && return + skip "could not find getfattr" # LU-532 trusted. xattr is invisible to non-root local testfile=$DIR/$tfile @@ -7996,11 +8282,10 @@ getxattr() { # getxattr path name test_102n() { # LU-4101 mdt: protect internal xattrs [ -z "$(which setfattr 2>/dev/null)" ] && - skip "could not find setfattr" && return + skip "could not find setfattr" if [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.5.50) ] then skip "MDT < 2.5.50 allows setxattr on internal trusted xattrs" - return fi local file0=$DIR/$tfile.0 @@ -8078,10 +8363,10 @@ test_102n() { # LU-4101 mdt: protect internal xattrs run_test 102n "silently ignore setxattr on internal trusted xattrs" test_102p() { # LU-4703 setxattr did not check ownership - local testfile=$DIR/$tfile - [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.5.56) ] && - skip "MDS needs to be at least 2.5.56" && return + skip "MDS needs to be at least 2.5.56" + + local testfile=$DIR/$tfile touch $testfile @@ -8098,14 +8383,16 @@ run_test 102p "check setxattr(2) correctly fails without permission" test_102q() { [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.6.92) ] && - skip "MDS needs to be at least 2.6.92" && return + skip "MDS needs to be at least 2.6.92" + orphan_linkea_check $DIR/$tfile || error "orphan_linkea_check" } run_test 102q "flistxattr should not return trusted.link EAs for orphans" test_102r() { [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.6.93) ] && - skip "MDS needs to be at least 2.6.93" && return + skip "MDS needs to be at least 2.6.93" + touch $DIR/$tfile || error "touch" setfattr -n user.$(basename $tfile) $DIR/$tfile || error "setfattr" getfattr -n user.$(basename $tfile) $DIR/$tfile || error "getfattr" @@ -8135,6 +8422,56 @@ test_102r() { } run_test 102r "set EAs with empty values" +test_102s() { + [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.11.52) ] && + skip "MDS needs to be at least 2.11.52" + + local save="$TMP/$TESTSUITE-$TESTNAME.parameters" + + save_lustre_params client "llite.*.xattr_cache" > $save + + for cache in 0 1; do + lctl set_param llite.*.xattr_cache=$cache + + rm -f $DIR/$tfile + touch $DIR/$tfile || error "touch" + for prefix in lustre security system trusted user; do + # Note getxattr() may fail with 'Operation not + # supported' or 'No such attribute' depending + # on prefix and cache. + getfattr -n $prefix.n102s $DIR/$tfile && + error "getxattr '$prefix.n102s' should fail (cache = $cache)" + done + done + + restore_lustre_params < $save +} +run_test 102s "getting nonexistent xattrs should fail" + +test_102t() { + [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.11.52) ] && + skip "MDS needs to be at least 2.11.52" + + local save="$TMP/$TESTSUITE-$TESTNAME.parameters" + + save_lustre_params client "llite.*.xattr_cache" > $save + + for cache in 0 1; do + lctl set_param llite.*.xattr_cache=$cache + + for buf_size in 0 256; do + rm -f $DIR/$tfile + touch $DIR/$tfile || error "touch" + setfattr -n user.multiop $DIR/$tfile + $MULTIOP $DIR/$tfile oa$buf_size || + error "cannot get zero length xattr value (buf_size = $buf_size)" + done + done + + restore_lustre_params < $save +} +run_test 102t "zero length xattr values handled correctly" + run_acl_subtest() { $LUSTRE/tests/acl/run $LUSTRE/tests/acl/$1.test @@ -8142,13 +8479,13 @@ run_acl_subtest() } test_103a() { - [ "$UID" != 0 ] && skip_env "must run as root" && return + [ "$UID" != 0 ] && skip "must run as root" + $GSS && skip_env "could not run under gss" [ -z "$(lctl get_param -n mdc.*-mdc-*.connect_flags | grep acl)" ] && - skip "must have acl enabled" && return + skip_env "must have acl enabled" [ -z "$(which setfacl 2>/dev/null)" ] && - skip_env "could not find setfacl" && return - $GSS && skip "could not run under gss" && return - remote_mds_nodsh && skip "remote MDS with nodsh" && return + skip_env "could not find setfacl" + remote_mds_nodsh && skip "remote MDS with nodsh" gpasswd -a daemon bin # LU-5641 do_facet $SINGLEMDS gpasswd -a daemon bin # LU-5641 @@ -8216,7 +8553,37 @@ test_103a() { fi done } -run_test 103a "acl test =========================================" +run_test 103a "acl test" + +test_103b() { + local U + + for U in {0..511}; do + { + local O=$(printf "%04o" $U) + + umask $(printf "%04o" $((511 ^ $O))) + $LFS setstripe -c 1 $DIR/$tfile.s$O + local S=$(printf "%04o" 0$(stat -c%a $DIR/$tfile.s$O)) + + (( $S == ($O & 0666) )) || + error "lfs setstripe $DIR/$tfile.s$O '$S' != '$O'" + + $LFS setstripe -E16M -c 1 -E1G -S4M $DIR/$tfile.p$O + S=$(printf "%04o" 0$(stat -c%a $DIR/$tfile.p$O)) + (( $S == ($O & 0666) )) || + error "lfs setstripe -E $DIR/$tfile.p$O '$S' != '$O'" + + $LFS setstripe -N2 -c 1 $DIR/$tfile.m$O + S=$(printf "%04o" 0$(stat -c%a $DIR/$tfile.m$O)) + (( $S == ($O & 0666) )) || + error "lfs setstripe -N2 $DIR/$tfile.m$O '$S' != '$O'" + rm -f $DIR/$tfile.[smp]$0 + } & + done + wait +} +run_test 103b "umask lfs setstripe" test_103c() { mkdir -p $DIR/$tdir @@ -8231,7 +8598,8 @@ test_103c() { run_test 103c "'cp -rp' won't set empty acl" test_104a() { - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return + [ $PARALLEL == "yes" ] && skip "skip parallel run" + touch $DIR/$tfile lfs df || error "lfs df failed" lfs df -ih || error "lfs df -ih failed" @@ -8253,9 +8621,10 @@ test_104a() { run_test 104a "lfs df [-ih] [path] test =========================" test_104b() { - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return + [ $PARALLEL == "yes" ] && skip "skip parallel run" [ $RUNAS_ID -eq $UID ] && - skip_env "RUNAS_ID = UID = $UID -- skipping" && return + skip_env "RUNAS_ID = UID = $UID -- skipping" + chmod 666 /dev/obd denied_cnt=$(($($RUNAS $LFS check servers 2>&1 | grep "Permission denied" | wc -l))) @@ -8300,9 +8669,10 @@ test_105c() { run_test 105c "lockf when mounted without -o flock test" test_105d() { # bug 15924 - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return + [ $PARALLEL == "yes" ] && skip "skip parallel run" + test_mkdir $DIR/$tdir - flock_is_enabled || { skip "mount w/o flock enabled" && return; } + flock_is_enabled || skip_env "mount w/o flock enabled" #define OBD_FAIL_LDLM_CP_CB_WAIT 0x315 $LCTL set_param fail_loc=0x80000315 flocks_test 2 $DIR/$tdir @@ -8310,7 +8680,8 @@ test_105d() { # bug 15924 run_test 105d "flock race (should not freeze) ========" test_105e() { # bug 22660 && 22040 - flock_is_enabled || { skip "mount w/o flock enabled" && return; } + flock_is_enabled || skip_env "mount w/o flock enabled" + touch $DIR/$tfile flocks_test 3 $DIR/$tfile } @@ -8324,12 +8695,13 @@ test_106() { #bug 10921 run_test 106 "attempt exec of dir followed by chown of that dir" test_107() { - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return - CDIR=`pwd` - cd $DIR + [ $PARALLEL == "yes" ] && skip "skip parallel run" - local file=core - rm -f $file + CDIR=`pwd` + local file=core + + cd $DIR + rm -f $file local save_pattern=$(sysctl -n kernel.core_pattern) local save_uses_pid=$(sysctl -n kernel.core_uses_pid) @@ -8376,8 +8748,8 @@ run_test 110 "filename length checking" # Purpose: To verify dynamic thread (OSS) creation. # test_115() { - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return - remote_ost_nodsh && skip "remote OST with nodsh" && return + [ $PARALLEL == "yes" ] && skip "skip parallel run" + remote_ost_nodsh && skip "remote OST with nodsh" # Lustre does not stop service threads once they are started. # Reset number of running threads to default. @@ -8503,10 +8875,9 @@ free_min_max () { } test_116a() { # was previously test_116() - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return - remote_mds_nodsh && skip "remote MDS with nodsh" && return - - [[ $OSTCOUNT -lt 2 ]] && skip_env "needs >= 2 OSTs" && return + [ $PARALLEL == "yes" ] && skip "skip parallel run" + [[ $OSTCOUNT -lt 2 ]] && skip_env "needs >= 2 OSTs" + remote_mds_nodsh && skip "remote MDS with nodsh" echo -n "Free space priority " do_facet $SINGLEMDS lctl get_param -n lo*.*-mdtlov.qos_prio_free | @@ -8514,12 +8885,10 @@ test_116a() { # was previously test_116() declare -a AVAIL free_min_max - [ $MINV -eq 0 ] && skip "no free space in OST$MINI, skip" && return - [ $MINV -gt 10000000 ] && skip "too much free space in OST$MINI, skip" \ - && return + [ $MINV -eq 0 ] && skip "no free space in OST$MINI, skip" + [ $MINV -gt 10000000 ] && skip "too much free space in OST$MINI, skip" trap simple_cleanup_common EXIT - # Check if we need to generate uneven OSTs test_mkdir -p $DIR/$tdir/OST${MINI} local FILL=$((MINV / 4)) @@ -8564,9 +8933,8 @@ test_116a() { # was previously test_116() echo "ok" else echo "failed - QOS mode won't be used" - skip "QOS imbalance criteria not met" simple_cleanup_common - return + skip "QOS imbalance criteria not met" fi MINI1=$MINI @@ -8628,13 +8996,13 @@ test_116a() { # was previously test_116() run_test 116a "stripe QOS: free space balance ===================" test_116b() { # LU-2093 - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return - remote_mds_nodsh && skip "remote MDS with nodsh" && return + [ $PARALLEL == "yes" ] && skip "skip parallel run" + remote_mds_nodsh && skip "remote MDS with nodsh" #define OBD_FAIL_MDS_OSC_CREATE_FAIL 0x147 local old_rr=$(do_facet $SINGLEMDS lctl get_param -n \ lo*.$FSNAME-MDT0000-mdtlov.qos_threshold_rr | head -1) - [ -z "$old_rr" ] && skip "no QOS" && return 0 + [ -z "$old_rr" ] && skip "no QOS" do_facet $SINGLEMDS lctl set_param \ lo*.$FSNAME-MDT0000-mdtlov.qos_threshold_rr=0 mkdir -p $DIR/$tdir @@ -8649,13 +9017,14 @@ run_test 116b "QoS shouldn't LBUG if not enough OSTs found on the 2nd pass" test_117() # bug 10891 { - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return - dd if=/dev/zero of=$DIR/$tfile bs=1M count=1 - #define OBD_FAIL_OST_SETATTR_CREDITS 0x21e - lctl set_param fail_loc=0x21e - > $DIR/$tfile || error "truncate failed" - lctl set_param fail_loc=0 - echo "Truncate succeeded." + [ $PARALLEL == "yes" ] && skip "skip parallel run" + + dd if=/dev/zero of=$DIR/$tfile bs=1M count=1 + #define OBD_FAIL_OST_SETATTR_CREDITS 0x21e + lctl set_param fail_loc=0x21e + > $DIR/$tfile || error "truncate failed" + lctl set_param fail_loc=0 + echo "Truncate succeeded." rm -f $DIR/$tfile } run_test 117 "verify osd extend ==========" @@ -8685,7 +9054,8 @@ reset_async() { test_118a() #bug 11710 { - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return + [ $PARALLEL == "yes" ] && skip "skip parallel run" + reset_async $MULTIOP $DIR/$tfile oO_CREAT:O_RDWR:O_SYNC:w4096c @@ -8702,8 +9072,8 @@ run_test 118a "verify O_SYNC works ==========" test_118b() { - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return - remote_ost_nodsh && skip "remote OST with nodsh" && return + [ $PARALLEL == "yes" ] && skip "skip parallel run" + remote_ost_nodsh && skip "remote OST with nodsh" reset_async @@ -8739,12 +9109,12 @@ run_test 118b "Reclaim dirty pages on fatal error ==========" test_118c() { - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return + [ $PARALLEL == "yes" ] && skip "skip parallel run" # for 118c, restore the original resend count, LU-1940 [ "$SLOW" = "no" ] && [ -n "$OLD_RESENDCOUNT" ] && set_resend_count $OLD_RESENDCOUNT - remote_ost_nodsh && skip "remote OST with nodsh" && return + remote_ost_nodsh && skip "remote OST with nodsh" reset_async @@ -8791,8 +9161,8 @@ run_test 118c "Fsync blocks on EROFS until dirty pages are flushed ==========" test_118d() { - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return - remote_ost_nodsh && skip "remote OST with nodsh" && return + [ $PARALLEL == "yes" ] && skip "skip parallel run" + remote_ost_nodsh && skip "remote OST with nodsh" reset_async @@ -8830,11 +9200,12 @@ test_118d() run_test 118d "Fsync validation inject a delay of the bulk ==========" test_118f() { - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return - reset_async + [ $PARALLEL == "yes" ] && skip "skip parallel run" - #define OBD_FAIL_OSC_BRW_PREP_REQ2 0x40a - lctl set_param fail_loc=0x8000040a + reset_async + + #define OBD_FAIL_OSC_BRW_PREP_REQ2 0x40a + lctl set_param fail_loc=0x8000040a # Should simulate EINVAL error which is fatal $MULTIOP $DIR/$tfile oO_CREAT:O_RDWR:O_SYNC:w4096c @@ -8866,7 +9237,8 @@ test_118f() { run_test 118f "Simulate unrecoverable OSC side error ==========" test_118g() { - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return + [ $PARALLEL == "yes" ] && skip "skip parallel run" + reset_async #define OBD_FAIL_OSC_BRW_PREP_REQ 0x406 @@ -8902,10 +9274,10 @@ test_118g() { run_test 118g "Don't stay in wait if we got local -ENOMEM ==========" test_118h() { - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return - remote_ost_nodsh && skip "remote OST with nodsh" && return + [ $PARALLEL == "yes" ] && skip "skip parallel run" + remote_ost_nodsh && skip "remote OST with nodsh" - reset_async + reset_async #define OBD_FAIL_OST_BRW_WRITE_BULK 0x20e set_nodes_failloc "$(osts_nodes)" 0x20e @@ -8940,10 +9312,10 @@ run_test 118h "Verify timeout in handling recoverables errors ==========" [ "$SLOW" = "no" ] && [ -n "$OLD_RESENDCOUNT" ] && set_resend_count $OLD_RESENDCOUNT test_118i() { - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return - remote_ost_nodsh && skip "remote OST with nodsh" && return + [ $PARALLEL == "yes" ] && skip "skip parallel run" + remote_ost_nodsh && skip "remote OST with nodsh" - reset_async + reset_async #define OBD_FAIL_OST_BRW_WRITE_BULK 0x20e set_nodes_failloc "$(osts_nodes)" 0x20e @@ -8981,10 +9353,10 @@ run_test 118i "Fix error before timeout in recoverable error ==========" [ "$SLOW" = "no" ] && set_resend_count 4 test_118j() { - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return - remote_ost_nodsh && skip "remote OST with nodsh" && return + [ $PARALLEL == "yes" ] && skip "skip parallel run" + remote_ost_nodsh && skip "remote OST with nodsh" - reset_async + reset_async #define OBD_FAIL_OST_BRW_WRITE_BULK2 0x220 set_nodes_failloc "$(osts_nodes)" 0x220 @@ -9018,8 +9390,8 @@ run_test 118j "Simulate unrecoverable OST side error ==========" test_118k() { - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return - remote_ost_nodsh && skip "remote OSTs with nodsh" && return + [ $PARALLEL == "yes" ] && skip "skip parallel run" + remote_ost_nodsh && skip "remote OSTs with nodsh" #define OBD_FAIL_OST_BRW_WRITE_BULK 0x20e set_nodes_failloc "$(osts_nodes)" 0x20e @@ -9039,10 +9411,10 @@ test_118k() } run_test 118k "bio alloc -ENOMEM and IO TERM handling =========" -test_118l() +test_118l() # LU-646 { - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return - # LU-646 + [ $PARALLEL == "yes" ] && skip "skip parallel run" + test_mkdir $DIR/$tdir $MULTIOP $DIR/$tdir Dy || error "fsync dir failed" rm -rf $DIR/$tdir @@ -9051,7 +9423,8 @@ run_test 118l "fsync dir" test_118m() # LU-3066 { - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return + [ $PARALLEL == "yes" ] && skip "skip parallel run" + test_mkdir $DIR/$tdir $MULTIOP $DIR/$tdir DY || error "fdatasync dir failed" rm -rf $DIR/$tdir @@ -9065,8 +9438,8 @@ test_118n() local begin local end - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return - remote_ost_nodsh && skip "remote OSTs with nodsh" && return + [ $PARALLEL == "yes" ] && skip "skip parallel run" + remote_ost_nodsh && skip "remote OSTs with nodsh" # Sleep to avoid a cached response. #define OBD_STATFS_CACHE_SECONDS 1 @@ -9106,14 +9479,14 @@ run_test 119a "Short directIO read must return actual read amount" test_119b() # bug 11737 { - [[ $OSTCOUNT -lt 2 ]] && skip_env "needs >= 2 OSTs" && return + [[ $OSTCOUNT -lt 2 ]] && skip_env "needs >= 2 OSTs" - $SETSTRIPE -c 2 $DIR/$tfile || error "setstripe failed" - dd if=/dev/zero of=$DIR/$tfile bs=1M count=1 seek=1 || error "dd failed" - sync - $MULTIOP $DIR/$tfile oO_RDONLY:O_DIRECT:r$((2048 * 1024)) || \ - error "direct read failed" - rm -f $DIR/$tfile + $SETSTRIPE -c 2 $DIR/$tfile || error "setstripe failed" + dd if=/dev/zero of=$DIR/$tfile bs=1M count=1 seek=1 || error "dd failed" + sync + $MULTIOP $DIR/$tfile oO_RDONLY:O_DIRECT:r$((2048 * 1024)) || + error "direct read failed" + rm -f $DIR/$tfile } run_test 119b "Sparse directIO read must return actual read amount" @@ -9128,7 +9501,8 @@ run_test 119c "Testing for direct read hitting hole" test_119d() # bug 15950 { - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return + [ $PARALLEL == "yes" ] && skip "skip parallel run" + MAX_RPCS_IN_FLIGHT=`$LCTL get_param -n osc.*OST0000-osc-[^mM]*.max_rpcs_in_flight` $LCTL set_param -n osc.*OST0000-osc-[^mM]*.max_rpcs_in_flight 1 BSIZE=1048576 @@ -9153,11 +9527,11 @@ test_119d() # bug 15950 run_test 119d "The DIO path should try to send a new rpc once one is completed" test_120a() { - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return - remote_mds_nodsh && skip "remote MDS with nodsh" && return + [ $PARALLEL == "yes" ] && skip "skip parallel run" + remote_mds_nodsh && skip "remote MDS with nodsh" test_mkdir $DIR/$tdir $LCTL get_param -n mdc.*.connect_flags | grep -q early_lock_cancel || - { skip "no early lock cancel on server"; return 0; } + skip_env "no early lock cancel on server" lru_resize_disable mdc lru_resize_disable osc @@ -9185,11 +9559,12 @@ test_120a() { run_test 120a "Early Lock Cancel: mkdir test" test_120b() { - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return - remote_mds_nodsh && skip "remote MDS with nodsh" && return + [ $PARALLEL == "yes" ] && skip "skip parallel run" + remote_mds_nodsh && skip "remote MDS with nodsh" test_mkdir $DIR/$tdir $LCTL get_param -n mdc.*.connect_flags | grep -q early_lock_cancel || - { skip "no early lock cancel on server"; return 0; } + skip_env "no early lock cancel on server" + lru_resize_disable mdc lru_resize_disable osc cancel_lru_locks mdc @@ -9213,11 +9588,12 @@ test_120b() { run_test 120b "Early Lock Cancel: create test" test_120c() { - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return - remote_mds_nodsh && skip "remote MDS with nodsh" && return + [ $PARALLEL == "yes" ] && skip "skip parallel run" + remote_mds_nodsh && skip "remote MDS with nodsh" test_mkdir -c1 $DIR/$tdir $LCTL get_param -n mdc.*.connect_flags | grep -q early_lock_cancel || - { skip "no early lock cancel on server"; return 0; } + skip "no early lock cancel on server" + lru_resize_disable mdc lru_resize_disable osc test_mkdir -c1 $DIR/$tdir/d1 @@ -9244,11 +9620,12 @@ test_120c() { run_test 120c "Early Lock Cancel: link test" test_120d() { - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return - remote_mds_nodsh && skip "remote MDS with nodsh" && return + [ $PARALLEL == "yes" ] && skip "skip parallel run" + remote_mds_nodsh && skip "remote MDS with nodsh" test_mkdir -c1 $DIR/$tdir $LCTL get_param -n mdc.*.connect_flags | grep -q early_lock_cancel || - { skip "no early lock cancel on server"; return 0; } + skip_env "no early lock cancel on server" + lru_resize_disable mdc lru_resize_disable osc touch $DIR/$tdir @@ -9273,10 +9650,11 @@ test_120d() { run_test 120d "Early Lock Cancel: setattr test" test_120e() { - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return + [ $PARALLEL == "yes" ] && skip "skip parallel run" $LCTL get_param -n mdc.*.connect_flags | grep -q early_lock_cancel || - { skip "no early lock cancel on server"; return 0; } - remote_mds_nodsh && skip "remote MDS with nodsh" && return + skip_env "no early lock cancel on server" + remote_mds_nodsh && skip "remote MDS with nodsh" + local dlmtrace_set=false test_mkdir -c1 $DIR/$tdir @@ -9316,10 +9694,11 @@ test_120e() { run_test 120e "Early Lock Cancel: unlink test" test_120f() { - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return + [ $PARALLEL == "yes" ] && skip "skip parallel run" $LCTL get_param -n mdc.*.connect_flags | grep -q early_lock_cancel || - { skip "no early lock cancel on server"; return 0; } - remote_mds_nodsh && skip "remote MDS with nodsh" && return + skip_env "no early lock cancel on server" + remote_mds_nodsh && skip "remote MDS with nodsh" + test_mkdir -c1 $DIR/$tdir lru_resize_disable mdc lru_resize_disable osc @@ -9356,10 +9735,11 @@ test_120f() { run_test 120f "Early Lock Cancel: rename test" test_120g() { - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return + [ $PARALLEL == "yes" ] && skip "skip parallel run" $LCTL get_param -n mdc.*.connect_flags | grep -q early_lock_cancel || - { skip "no early lock cancel on server"; return 0; } - remote_mds_nodsh && skip "remote MDS with nodsh" && return + skip_env "no early lock cancel on server" + remote_mds_nodsh && skip "remote MDS with nodsh" + lru_resize_disable mdc lru_resize_disable osc count=10000 @@ -9402,7 +9782,8 @@ test_120g() { run_test 120g "Early Lock Cancel: performance test" test_121() { #bug #10589 - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return + [ $PARALLEL == "yes" ] && skip "skip parallel run" + rm -rf $DIR/$tfile writes=$(LANG=C dd if=/dev/zero of=$DIR/$tfile count=1 2>&1 | awk -F '+' '/out$/ {print $1}') #define OBD_FAIL_LDLM_CANCEL_RACE 0x310 @@ -9416,7 +9797,8 @@ test_121() { #bug #10589 run_test 121 "read cancel race =========" test_123a() { # was test 123, statahead(bug 11401) - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return + [ $PARALLEL == "yes" ] && skip "skip parallel run" + SLOWOK=0 if ! grep -q "processor.*: 1" /proc/cpuinfo; then log "testing UP system. Performance may be lower than expected." @@ -9499,7 +9881,8 @@ test_123a() { # was test 123, statahead(bug 11401) run_test 123a "verify statahead work" test_123b () { # statahead(bug 15027) - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return + [ $PARALLEL == "yes" ] && skip "skip parallel run" + test_mkdir $DIR/$tdir createmany -o $DIR/$tdir/$tfile-%d 1000 @@ -9519,10 +9902,12 @@ test_123b () { # statahead(bug 15027) run_test 123b "not panic with network error in statahead enqueue (bug 15027)" test_124a() { - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return + [ $PARALLEL == "yes" ] && skip "skip parallel run" $LCTL get_param -n mdc.*.connect_flags | grep -q lru_resize || - { skip "no lru resize on server"; return 0; } + skip_env "no lru resize on server" + local NR=2000 + test_mkdir $DIR/$tdir log "create $NR files at $DIR/$tdir" @@ -9546,21 +9931,19 @@ test_124a() { done if [[ -z "$NSDIR" || $LRU_SIZE -lt $(default_lru_size) ]]; then - skip "Not enough cached locks created!" - return 0 - fi - log "LRU=$LRU_SIZE" + skip "Not enough cached locks created!" + fi + log "LRU=$LRU_SIZE" - local SLEEP=30 + local SLEEP=30 - # We know that lru resize allows one client to hold $LIMIT locks - # for 10h. After that locks begin to be killed by client. - local MAX_HRS=10 + # We know that lru resize allows one client to hold $LIMIT locks + # for 10h. After that locks begin to be killed by client. + local MAX_HRS=10 local LIMIT=$($LCTL get_param -n $NSDIR.pool.limit) log "LIMIT=$LIMIT" if [ $LIMIT -lt $LRU_SIZE ]; then - skip "Limit is too small $LIMIT" - return 0 + skip "Limit is too small $LIMIT" fi # Make LVF so higher that sleeping for $SLEEP is enough to _start_ @@ -9621,9 +10004,9 @@ get_max_pool_limit() } test_124b() { - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return + [ $PARALLEL == "yes" ] && skip "skip parallel run" $LCTL get_param -n mdc.*.connect_flags | grep -q lru_resize || - { skip "no lru resize on server"; return 0; } + skip_env "no lru resize on server" LIMIT=$(get_max_pool_limit) @@ -9696,9 +10079,9 @@ test_124b() { run_test 124b "lru resize (performance test) =======================" test_124c() { - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return + [ $PARALLEL == "yes" ] && skip "skip parallel run" $LCTL get_param -n mdc.*.connect_flags | grep -q lru_resize || - { skip "no lru resize on server"; return 0; } + skip_env "no lru resize on server" # cache ununsed locks on client local nr=100 @@ -9729,10 +10112,11 @@ run_test 124c "LRUR cancel very aged locks" test_125() { # 13358 $LCTL get_param -n llite.*.client_type | grep -q local || - { skip "must run as local client"; return; } + skip "must run as local client" $LCTL get_param -n mdc.*-mdc-*.connect_flags | grep -q acl || - { skip "must have acl enabled"; return; } - [ -z "$(which setfacl)" ] && skip "must have setfacl tool" && return + skip_env "must have acl enabled" + [ -z "$(which setfacl)" ] && skip_env "must have setfacl tool" + test_mkdir $DIR/$tdir $LFS setstripe -S 65536 -c -1 $DIR/$tdir || error "setstripe failed" setfacl -R -m u:bin:rwx $DIR/$tdir || error "setfacl $DIR/$tdir failed" @@ -9741,10 +10125,10 @@ test_125() { # 13358 run_test 125 "don't return EPROTO when a dir has a non-default striping and ACLs" test_126() { # bug 12829/13455 + $GSS && skip_env "must run as gss disabled" $LCTL get_param -n llite.*.client_type | grep -q local || - { skip "must run as local client"; return; } - [ "$UID" != 0 ] && skip_env "$TESTNAME (must run as root)" && return - $GSS && skip "must run as gss disabled" && return + skip "must run as local client" + [ "$UID" != 0 ] && skip "must run as root, not UID $UID" $RUNAS -u 0 -g 1 touch $DIR/$tfile || error "touch failed" gid=`ls -n $DIR/$tfile | awk '{print $4}'` @@ -9754,15 +10138,16 @@ test_126() { # bug 12829/13455 run_test 126 "check that the fsgid provided by the client is taken into account" test_127a() { # bug 15521 - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return - $SETSTRIPE -i 0 -c 1 $DIR/$tfile || error "setstripe failed" - $LCTL set_param osc.*.stats=0 - FSIZE=$((2048 * 1024)) - dd if=/dev/zero of=$DIR/$tfile bs=$FSIZE count=1 - cancel_lru_locks osc - dd if=$DIR/$tfile of=/dev/null bs=$FSIZE + [ $PARALLEL == "yes" ] && skip "skip parallel run" - $LCTL get_param osc.*0000-osc-*.stats | grep samples > $DIR/${tfile}.tmp + $SETSTRIPE -i 0 -c 1 $DIR/$tfile || error "setstripe failed" + $LCTL set_param osc.*.stats=0 + FSIZE=$((2048 * 1024)) + dd if=/dev/zero of=$DIR/$tfile bs=$FSIZE count=1 + cancel_lru_locks osc + dd if=$DIR/$tfile of=/dev/null bs=$FSIZE + + $LCTL get_param osc.*0000-osc-*.stats | grep samples > $DIR/${tfile}.tmp while read NAME COUNT SAMP UNIT MIN MAX SUM SUMSQ; do echo "got $COUNT $NAME" [ ! $MIN ] && error "Missing min value for $NAME proc entry" @@ -9793,37 +10178,45 @@ test_127a() { # bug 15521 run_test 127a "verify the client stats are sane" test_127b() { # bug LU-333 - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return - $LCTL set_param llite.*.stats=0 - FSIZE=65536 # sized fixed to match PAGE_SIZE for most clients - # perform 2 reads and writes so MAX is different from SUM. - dd if=/dev/zero of=$DIR/$tfile bs=$FSIZE count=1 - dd if=/dev/zero of=$DIR/$tfile bs=$FSIZE count=1 - cancel_lru_locks osc - dd if=$DIR/$tfile of=/dev/null bs=$FSIZE count=1 - dd if=$DIR/$tfile of=/dev/null bs=$FSIZE count=1 + [ $PARALLEL == "yes" ] && skip "skip parallel run" + local name count samp unit min max sum sumsq - $LCTL get_param llite.*.stats | grep samples > $TMP/${tfile}.tmp - while read NAME COUNT SAMP UNIT MIN MAX SUM SUMSQ; do - echo "got $COUNT $NAME" - eval $NAME=$COUNT || error "Wrong proc format" + $LCTL set_param llite.*.stats=0 - case $NAME in - read_bytes) - [ $COUNT -ne 2 ] && error "count is not 2: $COUNT" - [ $MIN -ne $FSIZE ] && error "min is not $FSIZE: $MIN" - [ $MAX -ne $FSIZE ] && error "max is incorrect: $MAX" - [ $SUM -ne $((FSIZE * 2)) ] && error "sum is wrong: $SUM" - ;; - write_bytes) - [ $COUNT -ne 2 ] && error "count is not 2: $COUNT" - [ $MIN -ne $FSIZE ] && error "min is not $FSIZE: $MIN" - [ $MAX -ne $FSIZE ] && error "max is incorrect: $MAX" - [ $SUM -ne $((FSIZE * 2)) ] && error "sum is wrong: $SUM" - ;; - *) ;; - esac - done < $TMP/${tfile}.tmp + # perform 2 reads and writes so MAX is different from SUM. + dd if=/dev/zero of=$DIR/$tfile bs=$PAGE_SIZE count=1 + dd if=/dev/zero of=$DIR/$tfile bs=$PAGE_SIZE count=1 + cancel_lru_locks osc + dd if=$DIR/$tfile of=/dev/null bs=$PAGE_SIZE count=1 + dd if=$DIR/$tfile of=/dev/null bs=$PAGE_SIZE count=1 + + $LCTL get_param llite.*.stats | grep samples > $TMP/$tfile.tmp + while read name count samp unit min max sum sumsq; do + echo "got $count $name" + eval $name=$count || error "Wrong proc format" + + case $name in + read_bytes) + [ $count -ne 2 ] && error "count is not 2: $count" + [ $min -ne $PAGE_SIZE ] && + error "min is not $PAGE_SIZE: $min" + [ $max -ne $PAGE_SIZE ] && + error "max is incorrect: $max" + [ $sum -ne $((PAGE_SIZE * 2)) ] && + error "sum is wrong: $sum" + ;; + write_bytes) + [ $count -ne 2 ] && error "count is not 2: $count" + [ $min -ne $PAGE_SIZE ] && + error "min is not $PAGE_SIZE: $min" + [ $max -ne $PAGE_SIZE ] && + error "max is incorrect: $max" + [ $sum -ne $((PAGE_SIZE * 2)) ] && + error "sum is wrong: $sum" + ;; + *) ;; + esac + done < $TMP/$tfile.tmp #check that we actually got some stats [ "$read_bytes" ] || error "Missing read_bytes stats" @@ -9876,15 +10269,14 @@ check_mds_dmesg() { } test_129() { + [ $PARALLEL == "yes" ] && skip "skip parallel run" [[ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.5.56) ]] || - { skip "Need MDS version with at least 2.5.56"; return 0; } - - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return + skip "Need MDS version with at least 2.5.56" if [ "$(facet_fstype $SINGLEMDS)" != ldiskfs ]; then - skip "ldiskfs only test" - return + skip_env "ldiskfs only test" fi - remote_mds_nodsh && skip "remote MDS with nodsh" && return + remote_mds_nodsh && skip "remote MDS with nodsh" + local ENOSPC=28 local EFBIG=27 local has_warning=false @@ -9946,8 +10338,7 @@ cleanup_130() { test_130a() { local filefrag_op=$(filefrag -e 2>&1 | grep "invalid option") - [ -n "$filefrag_op" ] && skip_env "filefrag does not support FIEMAP" && - return + [ -n "$filefrag_op" ] && skip_env "filefrag does not support FIEMAP" trap cleanup_130 EXIT RETURN @@ -9960,7 +10351,7 @@ test_130a() { filefrag -ves $fm_file RC=$? [ "$(facet_fstype ost$(($($GETSTRIPE -i $fm_file) + 1)))" = "zfs" ] && - skip "ORI-366/LU-1941: FIEMAP unimplemented on ZFS" && return + skip_env "ORI-366/LU-1941: FIEMAP unimplemented on ZFS" [ $RC != 0 ] && error "filefrag $fm_file failed" filefrag_op=$(filefrag -ve -k $fm_file | @@ -9995,11 +10386,10 @@ test_130a() { run_test 130a "FIEMAP (1-stripe file)" test_130b() { - [ "$OSTCOUNT" -lt "2" ] && skip_env "needs >= 2 OSTs" && return + [ "$OSTCOUNT" -lt "2" ] && skip_env "needs >= 2 OSTs" local filefrag_op=$(filefrag -e 2>&1 | grep "invalid option") - [ -n "$filefrag_op" ] && skip_env "filefrag does not support FIEMAP" && - return + [ -n "$filefrag_op" ] && skip_env "filefrag does not support FIEMAP" trap cleanup_130 EXIT RETURN @@ -10007,7 +10397,7 @@ test_130b() { $SETSTRIPE -S 65536 -c $OSTCOUNT $fm_file || error "setstripe on $fm_file" [ "$(facet_fstype ost$(($($GETSTRIPE -i $fm_file) + 1)))" = "zfs" ] && - skip "ORI-366/LU-1941: FIEMAP unimplemented on ZFS" && return + skip_env "ORI-366/LU-1941: FIEMAP unimplemented on ZFS" dd if=/dev/zero of=$fm_file bs=1M count=$OSTCOUNT || error "dd failed on $fm_file" @@ -10055,18 +10445,17 @@ test_130b() { run_test 130b "FIEMAP ($OSTCOUNT-stripe file)" test_130c() { - [[ $OSTCOUNT -lt 2 ]] && skip_env "needs >= 2 OSTs" && return + [[ $OSTCOUNT -lt 2 ]] && skip_env "needs >= 2 OSTs" filefrag_op=$(filefrag -e 2>&1 | grep "invalid option") - [ -n "$filefrag_op" ] && skip "filefrag does not support FIEMAP" && - return + [ -n "$filefrag_op" ] && skip "filefrag does not support FIEMAP" trap cleanup_130 EXIT RETURN local fm_file=$DIR/$tfile $SETSTRIPE -S 65536 -c 2 $fm_file || error "setstripe on $fm_file" [ "$(facet_fstype ost$(($($GETSTRIPE -i $fm_file) + 1)))" = "zfs" ] && - skip "ORI-366/LU-1941: FIEMAP unimplemented on ZFS" && return + skip_env "ORI-366/LU-1941: FIEMAP unimplemented on ZFS" dd if=/dev/zero of=$fm_file seek=1 bs=1M count=1 || error "dd failed on $fm_file" @@ -10121,11 +10510,10 @@ test_130c() { run_test 130c "FIEMAP (2-stripe file with hole)" test_130d() { - [[ $OSTCOUNT -lt 3 ]] && skip_env "needs >= 3 OSTs" && return + [[ $OSTCOUNT -lt 3 ]] && skip_env "needs >= 3 OSTs" filefrag_op=$(filefrag -e 2>&1 | grep "invalid option") - [ -n "$filefrag_op" ] && skip "filefrag does not support FIEMAP" && - return + [ -n "$filefrag_op" ] && skip "filefrag does not support FIEMAP" trap cleanup_130 EXIT RETURN @@ -10133,7 +10521,7 @@ test_130d() { $SETSTRIPE -S 65536 -c $OSTCOUNT $fm_file || error "setstripe on $fm_file" [ "$(facet_fstype ost$(($($GETSTRIPE -i $fm_file) + 1)))" = "zfs" ] && - skip "ORI-366/LU-1941: FIEMAP unimplemented on ZFS" && return + skip_env "ORI-366/LU-1941: FIEMAP unimplemented on ZFS" local actual_stripe_count=$($GETSTRIPE -c $fm_file) dd if=/dev/zero of=$fm_file bs=1M count=$actual_stripe_count || @@ -10182,17 +10570,17 @@ test_130d() { run_test 130d "FIEMAP (N-stripe file)" test_130e() { - [[ $OSTCOUNT -lt 2 ]] && skip_env "needs >= 2 OSTs" && return + [[ $OSTCOUNT -lt 2 ]] && skip_env "needs >= 2 OSTs" filefrag_op=$(filefrag -e 2>&1 | grep "invalid option") - [ -n "$filefrag_op" ] && skip "filefrag does not support FIEMAP" && return + [ -n "$filefrag_op" ] && skip_env "filefrag does not support FIEMAP" trap cleanup_130 EXIT RETURN local fm_file=$DIR/$tfile $SETSTRIPE -S 131072 -c 2 $fm_file || error "setstripe on $fm_file" [ "$(facet_fstype ost$(($($GETSTRIPE -i $fm_file) + 1)))" = "zfs" ] && - skip "ORI-366/LU-1941: FIEMAP unimplemented on ZFS" && return + skip_env "ORI-366/LU-1941: FIEMAP unimplemented on ZFS" NUM_BLKS=512 EXPECTED_LEN=$(( (NUM_BLKS / 2) * 64 )) @@ -10246,8 +10634,7 @@ run_test 130e "FIEMAP (test continuation FIEMAP calls)" test_130f() { filefrag_op=$(filefrag -e 2>&1 | grep "invalid option") - [ -n "$filefrag_op" ] && skip "filefrag does not support FIEMAP" && - return + [ -n "$filefrag_op" ] && skip "filefrag does not support FIEMAP" local fm_file=$DIR/$tfile $MULTIOP $fm_file oO_RDWR:O_CREAT:O_LOV_DELAY_CREATE:T33554432c || @@ -10328,7 +10715,6 @@ check_stats() { ;; *) error "Wrong facet '$facet'" ;; esac - echo $res [ "$res" ] || error "The counter for $op on $facet was not incremented" # if the argument $3 is zero, it means any stat increment is ok. if [[ $want -gt 0 ]]; then @@ -10339,13 +10725,14 @@ check_stats() { } test_133a() { - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return - remote_ost_nodsh && skip "remote OST with nodsh" && return - remote_mds_nodsh && skip "remote MDS with nodsh" && return - + [ $PARALLEL == "yes" ] && skip "skip parallel run" + remote_ost_nodsh && skip "remote OST with nodsh" + remote_mds_nodsh && skip "remote MDS with nodsh" do_facet $SINGLEMDS $LCTL list_param mdt.*.rename_stats || - { skip "MDS doesn't support rename stats"; return; } + skip_env "MDS doesn't support rename stats" + local testdir=$DIR/${tdir}/stats_testdir + mkdir -p $DIR/${tdir} # clear stats. @@ -10386,10 +10773,14 @@ test_133a() { run_test 133a "Verifying MDT stats ========================================" test_133b() { - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return - remote_ost_nodsh && skip "remote OST with nodsh" && return - remote_mds_nodsh && skip "remote MDS with nodsh" && return + local res + + [ $PARALLEL == "yes" ] && skip "skip parallel run" + remote_ost_nodsh && skip "remote OST with nodsh" + remote_mds_nodsh && skip "remote MDS with nodsh" + local testdir=$DIR/${tdir}/stats_testdir + mkdir -p ${testdir} || error "mkdir failed" touch ${testdir}/${tfile} || error "touch failed" cancel_lru_locks mdc @@ -10407,18 +10798,53 @@ test_133b() { ls -l ${testdir}/${tfile} > /dev/null|| error "ls failed" check_stats $SINGLEMDS "getattr" 1 fi + rm -rf $DIR/${tdir} + + # when DNE is enabled, MDT uses STATFS RPC to ping other targets + # so the check below is not reliable + [ $MDSCOUNT -eq 1 ] || return 0 + + # Sleep to avoid a cached response. + #define OBD_STATFS_CACHE_SECONDS 1 + sleep 2 + do_facet $SINGLEMDS $LCTL set_param mdt.*.md_stats=clear + do_facet ost1 $LCTL set_param obdfilter.*.exports.*.stats=clear $LFS df || error "lfs failed" check_stats $SINGLEMDS "statfs" 1 - rm -rf $DIR/${tdir} + # check aggregated statfs (LU-10018) + [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.11.54) ] && + return 0 + [ $(lustre_version_code client) -lt $(version_code 2.11.54) ] && + return 0 + sleep 2 + do_facet $SINGLEMDS $LCTL set_param mdt.*.md_stats=clear + do_facet ost1 $LCTL set_param obdfilter.*.exports.*.stats=clear + df $DIR + check_stats $SINGLEMDS "statfs" 1 + + # We want to check that the client didn't send OST_STATFS to + # ost1 but the MDT also uses OST_STATFS for precreate. So some + # extra care is needed here. + if remote_mds; then + local nid=$($LCTL list_nids | head -1 | sed "s/\./\\\./g") + local param="obdfilter.$FSNAME-OST0000.exports.'$nid'.stats" + + res=$(do_facet ost1 $LCTL get_param $param | grep statfs) + [ "$res" ] && error "OST got STATFS" + fi + + return 0 } run_test 133b "Verifying extra MDT stats ==================================" test_133c() { - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return - remote_ost_nodsh && skip "remote OST with nodsh" && return - remote_mds_nodsh && skip "remote MDS with nodsh" && return + [ $PARALLEL == "yes" ] && skip "skip parallel run" + remote_ost_nodsh && skip "remote OST with nodsh" + remote_mds_nodsh && skip "remote MDS with nodsh" + local testdir=$DIR/$tdir/stats_testdir + test_mkdir -p $testdir # verify obdfilter stats. @@ -10492,28 +10918,30 @@ size_in_KMGT() { } get_rename_size() { - local size=$1 - local context=${2:-.} - local sample=$(do_facet $SINGLEMDS $LCTL get_param mdt.*.rename_stats | + local size=$1 + local context=${2:-.} + local sample=$(do_facet $SINGLEMDS $LCTL \ + get_param mdt.$FSNAME-MDT0000.rename_stats | grep -A1 $context | awk '/ '${size}'/ {print $4}' | sed -e "s/,//g") - echo $sample + echo $sample } test_133d() { - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return - remote_ost_nodsh && skip "remote OST with nodsh" && return - remote_mds_nodsh && skip "remote MDS with nodsh" && return + [ $PARALLEL == "yes" ] && skip "skip parallel run" + remote_ost_nodsh && skip "remote OST with nodsh" + remote_mds_nodsh && skip "remote MDS with nodsh" do_facet $SINGLEMDS $LCTL list_param mdt.*.rename_stats || - { skip "MDS doesn't support rename stats"; return; } + skip_env "MDS doesn't support rename stats" local testdir1=$DIR/${tdir}/stats_testdir1 local testdir2=$DIR/${tdir}/stats_testdir2 + mkdir -p $DIR/${tdir} do_facet $SINGLEMDS $LCTL set_param mdt.*.rename_stats=clear - mkdir -p ${testdir1} || error "mkdir failed" - mkdir -p ${testdir2} || error "mkdir failed" + lfs mkdir -i 0 -c 1 ${testdir1} || error "mkdir failed" + lfs mkdir -i 0 -c 1 ${testdir2} || error "mkdir failed" createmany -o $testdir1/test 512 || error "createmany failed" @@ -10534,7 +10962,9 @@ test_133d() { echo "source rename dir size: ${testdir1_size}" echo "target rename dir size: ${testdir2_size}" - local cmd="do_facet $SINGLEMDS $LCTL get_param mdt.*.rename_stats" + local cmd="do_facet $SINGLEMDS $LCTL " + cmd+="get_param mdt.$FSNAME-MDT0000.rename_stats" + eval $cmd || error "$cmd failed" local samedir=$($cmd | grep 'same_dir') local same_sample=$(get_rename_size $testdir1_size) @@ -10577,9 +11007,10 @@ test_133d() { run_test 133d "Verifying rename_stats ========================================" test_133e() { - remote_mds_nodsh && skip "remote MDS with nodsh" && return - remote_ost_nodsh && skip "remote OST with nodsh" && return - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return + remote_mds_nodsh && skip "remote MDS with nodsh" + remote_ost_nodsh && skip "remote OST with nodsh" + [ $PARALLEL == "yes" ] && skip "skip parallel run" + local testdir=$DIR/${tdir}/stats_testdir local ctr f0 f1 bs=32768 count=42 sum @@ -10624,6 +11055,18 @@ run_test 133e "Verifying OST {read,write}_bytes nid stats =================" proc_regexp="/{proc,sys}/{fs,sys,kernel/debug}/{lustre,lnet}/" +# Some versions of find (4.5.11, 4.5.14) included in CentOS 7.3-7.5 do +# not honor the -ignore_readdir_race option correctly. So we call +# error_ignore() rather than error() in these cases. See LU-11152. +error_133() { + if (find --version; do_facet mds1 find --version) | + grep -q '\b4\.5\.1[1-4]\b'; then + error_ignore LU-11152 "$@" + else + error "$@" + fi +} + test_133f() { # First without trusting modes. local proc_dirs=$(eval \ls -d $proc_regexp 2>/dev/null) @@ -10632,7 +11075,7 @@ test_133f() { find $proc_dirs -exec cat '{}' \; &> /dev/null # Second verifying readability. - $LCTL get_param -R '*' &> /dev/null || error "proc file read failed" + $LCTL get_param -R '*' &> /dev/null # Verifing writability with badarea_io. find $proc_dirs \ @@ -10641,20 +11084,20 @@ test_133f() { -not -name force_lbug \ -not -name changelog_mask \ -exec badarea_io '{}' \; || - error "find $proc_dirs failed" + error_133 "find $proc_dirs failed" } run_test 133f "Check reads/writes of client lustre proc files with bad area io" test_133g() { - remote_mds_nodsh && skip "remote MDS with nodsh" && return - remote_ost_nodsh && skip "remote OST with nodsh" && return + remote_mds_nodsh && skip "remote MDS with nodsh" + remote_ost_nodsh && skip "remote OST with nodsh" # eventually, this can also be replaced with "lctl get_param -R", # but not until that option is always available on the server local facet for facet in mds1 ost1; do [ $(lustre_version_code $facet) -le $(version_code 2.5.54) ] && - skip "Too old lustre on $facet" && continue + skip_noexit "Too old lustre on $facet" local facet_proc_dirs=$(do_facet $facet \ \\\ls -d $proc_regexp 2>/dev/null) echo "${facet}_proc_dirs='$facet_proc_dirs'" @@ -10675,7 +11118,7 @@ test_133g() { -not -name force_lbug \ -not -name changelog_mask \ -exec badarea_io '{}' \\\; || - error "$facet find $facet_proc_dirs failed" + error_133 "$facet find $facet_proc_dirs failed" done # remount the FS in case writes/reads /proc break the FS @@ -10686,12 +11129,13 @@ test_133g() { run_test 133g "Check reads/writes of server lustre proc files with bad area io" test_133h() { - remote_mds_nodsh && skip "remote MDS with nodsh" && return - remote_ost_nodsh && skip "remote OST with nodsh" && return + remote_mds_nodsh && skip "remote MDS with nodsh" + remote_ost_nodsh && skip "remote OST with nodsh" [[ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.9.54) ]] && - skip "Need MDS version at least 2.9.54" && return + skip "Need MDS version at least 2.9.54" local facet + for facet in client mds1 ost1; do local facet_proc_dirs=$(do_facet $facet \ \\\ls -d $proc_regexp 2> /dev/null) @@ -10714,9 +11158,9 @@ test_133h() { run_test 133h "Proc files should end with newlines" test_134a() { - remote_mds_nodsh && skip "remote MDS with nodsh" && return + remote_mds_nodsh && skip "remote MDS with nodsh" [[ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.7.54) ]] && - skip "Need MDS version at least 2.7.54" && return + skip "Need MDS version at least 2.7.54" mkdir -p $DIR/$tdir || error "failed to create $DIR/$tdir" cancel_lru_locks mdc @@ -10750,9 +11194,9 @@ test_134a() { run_test 134a "Server reclaims locks when reaching lock_reclaim_threshold" test_134b() { - remote_mds_nodsh && skip "remote MDS with nodsh" && return + remote_mds_nodsh && skip "remote MDS with nodsh" [[ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.7.54) ]] && - skip "Need MDS version at least 2.7.54" && return + skip "Need MDS version at least 2.7.54" mkdir -p $DIR/$tdir || error "failed to create $DIR/$tdir" cancel_lru_locks mdc @@ -10791,7 +11235,8 @@ test_134b() { run_test 134b "Server rejects lock request when reaching lock_limit_mb" test_140() { #bug-17379 - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return + [ $PARALLEL == "yes" ] && skip "skip parallel run" + test_mkdir $DIR/$tdir cd $DIR/$tdir || error "Changing to $DIR/$tdir" cp $(which stat) . || error "Copying stat to $DIR/$tdir" @@ -10830,7 +11275,8 @@ test_140() { #bug-17379 run_test 140 "Check reasonable stack depth (shouldn't LBUG) ====" test_150() { - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return + [ $PARALLEL == "yes" ] && skip "skip parallel run" + local TF="$TMP/$tfile" dd if=/dev/urandom of=$TF bs=6096 count=1 || error "dd failed" @@ -10918,30 +11364,27 @@ function set_cache() { } test_151() { - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return - remote_ost_nodsh && skip "remote OST with nodsh" && return + [ $PARALLEL == "yes" ] && skip "skip parallel run" + remote_ost_nodsh && skip "remote OST with nodsh" local CPAGES=3 local list=$(comma_list $(osts_nodes)) # check whether obdfilter is cache capable at all if ! get_osd_param $list '' read_cache_enable >/dev/null; then - echo "not cache-capable obdfilter" - return 0 + skip "not cache-capable obdfilter" fi # check cache is enabled on all obdfilters if get_osd_param $list '' read_cache_enable | grep 0; then - echo "oss cache is disabled" - return 0 + skip "oss cache is disabled" fi set_osd_param $list '' writethrough_cache_enable 1 # check write cache is enabled on all obdfilters if get_osd_param $list '' writethrough_cache_enable | grep 0; then - echo "oss write cache is NOT enabled" - return 0 + skip "oss write cache is NOT enabled" fi roc_hit_init @@ -10984,10 +11427,11 @@ test_151() { run_test 151 "test cache on oss and controls ===============================" test_152() { - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return - local TF="$TMP/$tfile" + [ $PARALLEL == "yes" ] && skip "skip parallel run" + + local TF="$TMP/$tfile" - # simulate ENOMEM during write + # simulate ENOMEM during write #define OBD_FAIL_OST_NOMEM 0x226 lctl set_param fail_loc=0x80000226 dd if=/dev/urandom of=$TF bs=6096 count=1 || error "dd failed" @@ -11138,7 +11582,7 @@ dot_lustre_fid_permission_check() { test_154A() { [[ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.4.1) ]] && - skip "Need MDS version at least 2.4.1" && return + skip "Need MDS version at least 2.4.1" local tf=$DIR/$tfile touch $tf @@ -11156,7 +11600,7 @@ run_test 154A "lfs path2fid and fid2path basic checks" test_154B() { [[ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.4.1) ]] && - skip "Need MDS version at least 2.4.1" && return + skip "Need MDS version at least 2.4.1" mkdir -p $DIR/$tdir || error "mkdir $tdir failed" touch $DIR/$tdir/$tfile || error "touch $DIR/$tdir/$tfile failed" @@ -11178,11 +11622,11 @@ test_154B() { run_test 154B "verify the ll_decode_linkea tool" test_154a() { - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return + [ $PARALLEL == "yes" ] && skip "skip parallel run" + [ -n "$FILESET" ] && skip "SKIP due to FILESET set" [[ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.2.51) ]] || - { skip "Need MDS version at least 2.2.51"; return 0; } - [ -z "$(which setfacl)" ] && skip "must have setfacl tool" && return - [ -n "$FILESET" ] && skip "SKIP due to FILESET set" && return + skip "Need MDS version at least 2.2.51" + [ -z "$(which setfacl)" ] && skip_env "must have setfacl tool" cp /etc/hosts $DIR/$tfile @@ -11208,12 +11652,11 @@ test_154a() { run_test 154a "Open-by-FID" test_154b() { - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return + [ $PARALLEL == "yes" ] && skip "skip parallel run" + [ -n "$FILESET" ] && skip "SKIP due to FILESET set" + [ $MDSCOUNT -lt 2 ] && skip_env "needs >= 2 MDTs" [[ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.2.51) ]] || - { skip "Need MDS version at least 2.2.51"; return 0; } - [ -n "$FILESET" ] && skip "SKIP due to FILESET set" && return - - [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return + skip "Need MDS version at least 2.2.51" local remote_dir=$DIR/$tdir/remote_dir local MDTIDX=1 @@ -11237,7 +11680,7 @@ run_test 154b "Open-by-FID for remote directory" test_154c() { [[ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.4.1) ]] && - skip "Need MDS version at least 2.4.1" && return + skip "Need MDS version at least 2.4.1" touch $DIR/$tfile.1 $DIR/$tfile.2 $DIR/$tfile.3 local FID1=$($LFS path2fid $DIR/$tfile.1) @@ -11264,9 +11707,9 @@ test_154c() { run_test 154c "lfs path2fid and fid2path multiple arguments" test_154d() { - remote_mds_nodsh && skip "remote MDS with nodsh" && return + remote_mds_nodsh && skip "remote MDS with nodsh" [[ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.5.53) ]] && - skip "Need MDS version at least 2.5.53" && return + skip "Need MDS version at least 2.5.53" if remote_mds; then nid=$($LCTL list_nids | sed "s/\./\\\./g") @@ -11300,7 +11743,7 @@ run_test 154d "Verify open file fid" test_154e() { [[ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.6.50) ]] && - skip "Need MDS version at least 2.6.50" && return + skip "Need MDS version at least 2.6.50" if ls -a $MOUNT | grep -q '^\.lustre$'; then error ".lustre returned by readdir" @@ -11309,7 +11752,8 @@ test_154e() run_test 154e ".lustre is not returned by readdir" test_154f() { - [ -n "$FILESET" ] && skip "SKIP due to FILESET set" && return + [ -n "$FILESET" ] && skip "SKIP due to FILESET set" + # create parent directory on a single MDT to avoid cross-MDT hardlinks test_mkdir -p -c1 $DIR/$tdir/d # test dirs inherit from its stripe @@ -11388,10 +11832,10 @@ run_test 154f "get parent fids by reading link ea" test_154g() { + [ -n "$FILESET" ] && skip "SKIP due to FILESET set" [[ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.6.92) && \ $(lustre_version_code client) -gt $(version_code 2.6.99) ]] || - { skip "Need MDS version at least 2.6.92"; return 0; } - [ -n "$FILESET" ] && skip "SKIP due to FILESET set" && return + skip "Need MDS version at least 2.6.92" mkdir -p $DIR/$tdir llapi_fid_test -d $DIR/$tdir @@ -11425,45 +11869,46 @@ test_155_small_load() { } test_155_big_load() { - remote_ost_nodsh && skip "remote OST with nodsh" && return - local temp=$TMP/$tfile - local file=$DIR/$tfile + remote_ost_nodsh && skip "remote OST with nodsh" - free_min_max - local cache_size=$(do_facet ost$((MAXI+1)) \ - "awk '/cache/ {sum+=\\\$4} END {print sum}' /proc/cpuinfo") - local large_file_size=$((cache_size * 2)) + local temp=$TMP/$tfile + local file=$DIR/$tfile - echo "OSS cache size: $cache_size KB" - echo "Large file size: $large_file_size KB" + free_min_max + local cache_size=$(do_facet ost$((MAXI+1)) \ + "awk '/cache/ {sum+=\\\$4} END {print sum}' /proc/cpuinfo") + local large_file_size=$((cache_size * 2)) - [ $MAXV -le $large_file_size ] && \ - skip_env "max available OST size needs > $large_file_size KB" && \ - return 0 + echo "OSS cache size: $cache_size KB" + echo "Large file size: $large_file_size KB" - $SETSTRIPE $file -c 1 -i $MAXI || error "$SETSTRIPE $file failed" + [ $MAXV -le $large_file_size ] && + skip_env "max available OST size needs > $large_file_size KB" - dd if=/dev/urandom of=$temp bs=$large_file_size count=1k || \ - error "dd of=$temp bs=$large_file_size count=1k failed" - cp $temp $file - ls -lh $temp $file - cancel_lru_locks osc - cmp $temp $file || error "$temp $file differ" + $SETSTRIPE $file -c 1 -i $MAXI || error "$SETSTRIPE $file failed" - rm -f $temp $file - true + dd if=/dev/urandom of=$temp bs=$large_file_size count=1k || + error "dd of=$temp bs=$large_file_size count=1k failed" + cp $temp $file + ls -lh $temp $file + cancel_lru_locks osc + cmp $temp $file || error "$temp $file differ" + + rm -f $temp $file + true } save_writethrough() { local facets=$(get_facets OST) - save_lustre_params $facets "obdfilter.*.writethrough_cache_enable" > $1 - save_lustre_params $facets "osd-*.*.writethrough_cache_enable" >> $1 + save_lustre_params $facets "osd-*.*.writethrough_cache_enable" > $1 } test_155a() { - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return + [ $PARALLEL == "yes" ] && skip "skip parallel run" + local p="$TMP/$TESTSUITE-$TESTNAME.parameters" + save_writethrough $p set_cache read on @@ -11475,8 +11920,10 @@ test_155a() { run_test 155a "Verify small file correctness: read cache:on write_cache:on" test_155b() { - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return + [ $PARALLEL == "yes" ] && skip "skip parallel run" + local p="$TMP/$TESTSUITE-$TESTNAME.parameters" + save_writethrough $p set_cache read on @@ -11488,8 +11935,10 @@ test_155b() { run_test 155b "Verify small file correctness: read cache:on write_cache:off" test_155c() { - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return + [ $PARALLEL == "yes" ] && skip "skip parallel run" + local p="$TMP/$TESTSUITE-$TESTNAME.parameters" + save_writethrough $p set_cache read off @@ -11501,8 +11950,10 @@ test_155c() { run_test 155c "Verify small file correctness: read cache:off write_cache:on" test_155d() { - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return + [ $PARALLEL == "yes" ] && skip "skip parallel run" + local p="$TMP/$TESTSUITE-$TESTNAME.parameters" + save_writethrough $p set_cache read off @@ -11514,8 +11965,10 @@ test_155d() { run_test 155d "Verify small file correctness: read cache:off write_cache:off" test_155e() { - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return + [ $PARALLEL == "yes" ] && skip "skip parallel run" + local p="$TMP/$TESTSUITE-$TESTNAME.parameters" + save_writethrough $p set_cache read on @@ -11527,8 +11980,10 @@ test_155e() { run_test 155e "Verify big file correctness: read cache:on write_cache:on" test_155f() { - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return + [ $PARALLEL == "yes" ] && skip "skip parallel run" + local p="$TMP/$TESTSUITE-$TESTNAME.parameters" + save_writethrough $p set_cache read on @@ -11540,8 +11995,10 @@ test_155f() { run_test 155f "Verify big file correctness: read cache:on write_cache:off" test_155g() { - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return + [ $PARALLEL == "yes" ] && skip "skip parallel run" + local p="$TMP/$TESTSUITE-$TESTNAME.parameters" + save_writethrough $p set_cache read off @@ -11553,8 +12010,10 @@ test_155g() { run_test 155g "Verify big file correctness: read cache:off write_cache:on" test_155h() { - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return + [ $PARALLEL == "yes" ] && skip "skip parallel run" + local p="$TMP/$TESTSUITE-$TESTNAME.parameters" + save_writethrough $p set_cache read off @@ -11566,19 +12025,18 @@ test_155h() { run_test 155h "Verify big file correctness: read cache:off write_cache:off" test_156() { - remote_ost_nodsh && skip "remote OST with nodsh" && return - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return + [ $PARALLEL == "yes" ] && skip "skip parallel run" + remote_ost_nodsh && skip "remote OST with nodsh" + [ "$(facet_fstype ost1)" = "zfs" -a \ + $(lustre_version_code ost1 -lt $(version_code 2.6.93)) ] && + skip "LU-1956/LU-2261: stats not implemented on OSD ZFS" + local CPAGES=3 local BEFORE local AFTER local file="$DIR/$tfile" local p="$TMP/$TESTSUITE-$TESTNAME.parameters" - [ "$(facet_fstype ost1)" = "zfs" -a \ - $(lustre_version_code ost1 -lt $(version_code 2.6.93)) ] && - skip "LU-1956/LU-2261: stats not implemented on OSD ZFS" && - return - save_writethrough $p roc_hit_init @@ -11713,10 +12171,10 @@ test_156() { run_test 156 "Verification of tunables" test_160a() { - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return - remote_mds_nodsh && skip "remote MDS with nodsh" && return + [ $PARALLEL == "yes" ] && skip "skip parallel run" + remote_mds_nodsh && skip "remote MDS with nodsh" [ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.2.0) ] || - { skip "Need MDS version at least 2.2.0"; return; } + skip "Need MDS version at least 2.2.0" changelog_register || error "changelog_register failed" local cl_user="${CL_USERS[$SINGLEMDS]%% *}" @@ -11833,10 +12291,10 @@ test_160a() { run_test 160a "changelog sanity" test_160b() { # LU-3587 - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return - remote_mds_nodsh && skip "remote MDS with nodsh" && return + [ $PARALLEL == "yes" ] && skip "skip parallel run" + remote_mds_nodsh && skip "remote MDS with nodsh" [ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.2.0) ] || - { skip "Need MDS version at least 2.2.0"; return; } + skip "Need MDS version at least 2.2.0" changelog_register || error "changelog_register failed" local cl_user="${CL_USERS[$SINGLEMDS]%% *}" @@ -11858,7 +12316,8 @@ test_160b() { # LU-3587 run_test 160b "Verify that very long rename doesn't crash in changelog" test_160c() { - remote_mds_nodsh && skip "remote MDS with nodsh" && return + [ $PARALLEL == "yes" ] && skip "skip parallel run" + remote_mds_nodsh && skip "remote MDS with nodsh" local rc=0 local server_version=$(lustre_version_code $SINGLEMDS) @@ -11866,8 +12325,7 @@ test_160c() { [[ $server_version -gt $(version_code 2.5.57) ]] || [[ $server_version -gt $(version_code 2.5.1) && $server_version -lt $(version_code 2.5.50) ]] || - { skip "Need MDS version at least 2.5.58 or 2.5.2+"; return; } - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return + skip "Need MDS version at least 2.5.58 or 2.5.2+" # Registration step changelog_register || error "changelog_register failed" @@ -11886,12 +12344,11 @@ test_160c() { run_test 160c "verify that changelog log catch the truncate event" test_160d() { - remote_mds_nodsh && skip "remote MDS with nodsh" && return - [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return - + remote_mds_nodsh && skip "remote MDS with nodsh" + [ $MDSCOUNT -lt 2 ] && skip_env "needs >= 2 MDTs" + [ $PARALLEL == "yes" ] && skip "skip parallel run" [[ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.7.60) ]] || - { skip "Need MDS version at least 2.7.60+"; return; } - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return + skip "Need MDS version at least 2.7.60" # Registration step changelog_register || error "changelog_register failed" @@ -11907,7 +12364,7 @@ test_160d() { run_test 160d "verify that changelog log catch the migrate event" test_160e() { - remote_mds_nodsh && skip "remote MDS with nodsh" && return + remote_mds_nodsh && skip "remote MDS with nodsh" # Create a user changelog_register || error "changelog_register failed" @@ -11945,9 +12402,12 @@ test_160f() { # Create a user changelog_register || error "first changelog_register failed" changelog_register || error "second changelog_register failed" - local cl_users=(${CL_USERS[$SINGLEMDS]}) - local cl_user1="${cl_users[0]}" - local cl_user2="${cl_users[1]}" + local cl_users + declare -A cl_user1 + declare -A cl_user2 + local user_rec1 + local user_rec2 + local i # generate some changelog records to accumulate on each MDT test_mkdir -c $MDSCOUNT $DIR/$tdir || error "test_mkdir $tdir failed" @@ -11955,11 +12415,11 @@ test_160f() { error "create $DIR/$tdir/$tfile failed" # check changelogs have been generated - nbcl=$(changelog_dump | wc -l) + local nbcl=$(changelog_dump | wc -l) [[ $nbcl -eq 0 ]] && error "no changelogs found" - # changelog_gc=1 should be set by default for param in "changelog_max_idle_time=10" \ + "changelog_gc=1" \ "changelog_min_gc_interval=2" \ "changelog_min_free_cat_entries=3"; do local MDT0=$(facet_svc $SINGLEMDS) @@ -11970,51 +12430,86 @@ test_160f() { do_nodes $mdts $LCTL set_param mdd.*.$param done + # force cl_user2 to be idle (1st part) + sleep 9 + # simulate changelog catalog almost full #define OBD_FAIL_CAT_FREE_RECORDS 0x1313 do_nodes $mdts $LCTL set_param fail_loc=0x1313 fail_val=3 - sleep 6 - local user_rec1=$(changelog_user_rec $SINGLEMDS $cl_user1) - [ -n "$user_rec1" ] || - error "User $cl_user1 not found in changelog_users" - __changelog_clear $SINGLEMDS $cl_user1 +2 - local user_rec2=$(changelog_user_rec $SINGLEMDS $cl_user1) - [ -n "$user_rec2" ] || - error "User $cl_user1 not found in changelog_users" - echo "verifying user clear: $user_rec1 + 2 == $user_rec2" - [ $((user_rec1 + 2)) == $user_rec2 ] || - error "user index expected $user_rec1 + 2, but is $user_rec2" - sleep 5 + for i in $(seq $MDSCOUNT); do + cl_users=(${CL_USERS[mds$i]}) + cl_user1[mds$i]="${cl_users[0]}" + cl_user2[mds$i]="${cl_users[1]}" + + [ -n "${cl_user1[mds$i]}" ] || + error "mds$i: no user registered" + [ -n "${cl_user2[mds$i]}" ] || + error "mds$i: only ${cl_user2[mds$i]} is registered" + + user_rec1=$(changelog_user_rec mds$i ${cl_user1[mds$i]}) + [ -n "$user_rec1" ] || + error "mds$i: User ${cl_user1[mds$i]} not registered" + __changelog_clear mds$i ${cl_user1[mds$i]} +2 + user_rec2=$(changelog_user_rec mds$i ${cl_user1[mds$i]}) + [ -n "$user_rec2" ] || + error "mds$i: User ${cl_user1[mds$i]} not registered" + echo "mds$i: verifying user ${cl_user1[mds$i]} clear: " \ + "$user_rec1 + 2 == $user_rec2" + [ $((user_rec1 + 2)) == $user_rec2 ] || + error "mds$i: user ${cl_user1[mds$i]} index expected " \ + "$user_rec1 + 2, but is $user_rec2" + user_rec2=$(changelog_user_rec mds$i ${cl_user2[mds$i]}) + [ -n "$user_rec2" ] || + error "mds$i: User ${cl_user2[mds$i]} not registered" + [ $user_rec1 == $user_rec2 ] || + error "mds$i: user ${cl_user2[mds$i]} index expected " \ + "$user_rec1, but is $user_rec2" + done + + # force cl_user2 to be idle (2nd part) and to reach + # changelog_max_idle_time + sleep 2 # generate one more changelog to trigger fail_loc - rm -rf $DIR/$tdir || error "rm -rf $tdir failed" + createmany -m $DIR/$tdir/${tfile}bis $((MDSCOUNT * 2)) || + error "create $DIR/$tdir/${tfile}bis failed" # ensure gc thread is done - wait_update_facet $SINGLEMDS \ - "ps -e -o comm= | grep chlg_gc_thread" "" 20 - - # check user still registered - changelog_users $SINGLEMDS | grep -q "$cl_user1" || - error "User $cl_user1 not found in changelog_users" - # check user2 unregistered - changelog_users $SINGLEMDS | grep -q "$cl_user2" && - error "User $cl_user2 still found in changelog_users" - - # check changelogs are present and starting at $user_rec2 + 1 - local first_rec=$($LFS changelog $(facet_svc $SINGLEMDS) | - awk '{ print $1; exit; }') + for i in $(mdts_nodes); do + wait_update $i \ + "ps -e -o comm= | grep chlg_gc_thread" "" 20 || + error "$i: GC-thread not done" + done - echo "verifying min purge: $user_rec2 + 1 == $first_rec" - [ $((user_rec2 + 1)) == $first_rec ] || - error "first index should be $user_rec2 + 1, but is $first_rec" + local first_rec + for i in $(seq $MDSCOUNT); do + # check cl_user1 still registered + changelog_users mds$i | grep -q "${cl_user1[mds$i]}" || + error "mds$i: User ${cl_user1[mds$i]} not registered" + # check cl_user2 unregistered + changelog_users mds$i | grep -q "${cl_user2[mds$i]}" && + error "mds$i: User ${cl_user2[mds$i]} still registered" + + # check changelogs are present and starting at $user_rec1 + 1 + user_rec1=$(changelog_user_rec mds$i ${cl_user1[mds$i]}) + [ -n "$user_rec1" ] || + error "mds$i: User ${cl_user1[mds$i]} not registered" + first_rec=$($LFS changelog $(facet_svc mds$i) | + awk '{ print $1; exit; }') + + echo "mds$i: verifying first index $user_rec1 + 1 == $first_rec" + [ $((user_rec1 + 1)) == $first_rec ] || + error "mds$i: first index should be $user_rec1 + 1, " \ + "but is $first_rec" + done } run_test 160f "changelog garbage collect (timestamped users)" test_160g() { - remote_mds_nodsh && skip "remote MDS with nodsh" && return + remote_mds_nodsh && skip "remote MDS with nodsh" [[ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.10.56) ]] || - { skip "Need MDS version at least 2.10.56"; return 0; } + skip "Need MDS version at least 2.10.56" local mdts=$(comma_list $(mdts_nodes)) @@ -12024,9 +12519,12 @@ test_160g() { # Create a user changelog_register || error "first changelog_register failed" changelog_register || error "second changelog_register failed" - local cl_users=(${CL_USERS[$SINGLEMDS]}) - local cl_user1="${cl_users[0]}" - local cl_user2="${cl_users[1]}" + local cl_users + declare -A cl_user1 + declare -A cl_user2 + local user_rec1 + local user_rec2 + local i # generate some changelog records to accumulate on each MDT test_mkdir -c $MDSCOUNT $DIR/$tdir || error "mkdir $tdir failed" @@ -12034,11 +12532,11 @@ test_160g() { error "create $DIR/$tdir/$tfile failed" # check changelogs have been generated - nbcl=$(changelog_dump | wc -l) + local nbcl=$(changelog_dump | wc -l) [[ $nbcl -eq 0 ]] && error "no changelogs found" - # changelog_gc=1 should be set by default for param in "changelog_max_idle_indexes=$((nbcl / 2))" \ + "changelog_gc=1" \ "changelog_min_gc_interval=2" \ "changelog_min_free_cat_entries=3"; do local MDT0=$(facet_svc $SINGLEMDS) @@ -12054,42 +12552,297 @@ test_160g() { #define OBD_FAIL_CAT_FREE_RECORDS 0x1313 do_nodes $mdts $LCTL set_param fail_loc=0x1313 fail_val=3 - local user_rec1=$(changelog_user_rec $SINGLEMDS $cl_user1) - - __changelog_clear $SINGLEMDS $cl_user1 +3 - - local user_rec2=$(changelog_user_rec $SINGLEMDS $cl_user1) + for i in $(seq $MDSCOUNT); do + cl_users=(${CL_USERS[mds$i]}) + cl_user1[mds$i]="${cl_users[0]}" + cl_user2[mds$i]="${cl_users[1]}" + + [ -n "${cl_user1[mds$i]}" ] || + error "mds$i: no user registered" + [ -n "${cl_user2[mds$i]}" ] || + error "mds$i: only ${cl_user1[mds$i]} is registered" + + user_rec1=$(changelog_user_rec mds$i ${cl_user1[mds$i]}) + [ -n "$user_rec1" ] || + error "mds$i: User ${cl_user1[mds$i]} not registered" + __changelog_clear mds$i ${cl_user1[mds$i]} +2 + user_rec2=$(changelog_user_rec mds$i ${cl_user1[mds$i]}) + [ -n "$user_rec2" ] || + error "mds$i: User ${cl_user1[mds$i]} not registered" + echo "mds$i: verifying user ${cl_user1[mds$i]} clear: " \ + "$user_rec1 + 2 == $user_rec2" + [ $((user_rec1 + 2)) == $user_rec2 ] || + error "mds$i: user ${cl_user1[mds$i]} index expected " \ + "$user_rec1 + 2, but is $user_rec2" + user_rec2=$(changelog_user_rec mds$i ${cl_user2[mds$i]}) + [ -n "$user_rec2" ] || + error "mds$i: User ${cl_user2[mds$i]} not registered" + [ $user_rec1 == $user_rec2 ] || + error "mds$i: user ${cl_user2[mds$i]} index expected " \ + "$user_rec1, but is $user_rec2" + done - echo "verifying user clear: $user_rec1 + 3 == $user_rec2" - [ $((user_rec1 + 3)) == $user_rec2 ] || - error "user index expected $user_rec1 + 3, but is $user_rec2" + # ensure we are past the previous changelog_min_gc_interval set above + sleep 2 # generate one more changelog to trigger fail_loc - rm -rf $DIR/$tdir || error "rm -rf $tdir failed" + createmany -m $DIR/$tdir/${tfile}bis $((MDSCOUNT * 2)) || + error "create $DIR/$tdir/${tfile}bis failed" # ensure gc thread is done - wait_update_facet $SINGLEMDS \ - "ps -e -o comm= | grep chlg_gc_thread" "" 20 + for i in $(mdts_nodes); do + wait_update $i \ + "ps -e -o comm= | grep chlg_gc_thread" "" 20 || + error "$i: GC-thread not done" + done - # check user still registered - [ -n "$(changelog_user_rec $SINGLEMDS $cl_user1)" ] || - error "User $cl_user1 not found in changelog_users" - # check user2 unregistered - [ -z "$(changelog_user_rec $SINGLEMDS $cl_user2)" ] || - error "User $cl_user2 still found in changelog_users" + local first_rec + for i in $(seq $MDSCOUNT); do + # check cl_user1 still registered + changelog_users mds$i | grep -q "${cl_user1[mds$i]}" || + error "mds$i: User ${cl_user1[mds$i]} not registered" + # check cl_user2 unregistered + changelog_users mds$i | grep -q "${cl_user2[mds$i]}" && + error "mds$i: User ${cl_user2[mds$i]} still registered" + + # check changelogs are present and starting at $user_rec1 + 1 + user_rec1=$(changelog_user_rec mds$i ${cl_user1[mds$i]}) + [ -n "$user_rec1" ] || + error "mds$i: User ${cl_user1[mds$i]} not registered" + first_rec=$($LFS changelog $(facet_svc mds$i) | + awk '{ print $1; exit; }') + + echo "mds$i: verifying first index $user_rec1 + 1 == $first_rec" + [ $((user_rec1 + 1)) == $first_rec ] || + error "mds$i: first index should be $user_rec1 + 1, " \ + "but is $first_rec" + done +} +run_test 160g "changelog garbage collect (old users)" - # check changelogs are present and starting at $user_rec2 + 1 - local first_rec=$($LFS changelog $(facet_svc $SINGLEMDS) | - awk '{ print $1; exit; }') +test_160h() { + remote_mds_nodsh && skip "remote MDS with nodsh" && return + [[ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.10.56) ]] || + { skip "Need MDS version at least 2.10.56"; return 0; } + + local mdts=$(comma_list $(mdts_nodes)) + + # Create a user + changelog_register || error "first changelog_register failed" + changelog_register || error "second changelog_register failed" + local cl_users + declare -A cl_user1 + declare -A cl_user2 + local user_rec1 + local user_rec2 + local i + + # generate some changelog records to accumulate on each MDT + test_mkdir -c $MDSCOUNT $DIR/$tdir || error "test_mkdir $tdir failed" + createmany -m $DIR/$tdir/$tfile $((MDSCOUNT * 2)) || + error "create $DIR/$tdir/$tfile failed" + + # check changelogs have been generated + local nbcl=$(changelog_dump | wc -l) + [[ $nbcl -eq 0 ]] && error "no changelogs found" - echo "verifying min purge: $user_rec2 + 1 == $first_rec" - [ $((user_rec2 + 1)) == $first_rec ] || - error "first index should be $user_rec2 + 1, but is $first_rec" + for param in "changelog_max_idle_time=10" \ + "changelog_gc=1" \ + "changelog_min_gc_interval=2"; do + local MDT0=$(facet_svc $SINGLEMDS) + local var="${param%=*}" + local old=$(do_facet mds1 "$LCTL get_param -n mdd.$MDT0.$var") + + stack_trap "do_nodes $mdts $LCTL set_param mdd.*.$var=$old" EXIT + do_nodes $mdts $LCTL set_param mdd.*.$param + done + + # force cl_user2 to be idle (1st part) + sleep 9 + + for i in $(seq $MDSCOUNT); do + cl_users=(${CL_USERS[mds$i]}) + cl_user1[mds$i]="${cl_users[0]}" + cl_user2[mds$i]="${cl_users[1]}" + + [ -n "${cl_user1[mds$i]}" ] || + error "mds$i: no user registered" + [ -n "${cl_user2[mds$i]}" ] || + error "mds$i: only ${cl_user2[mds$i]} is registered" + + user_rec1=$(changelog_user_rec mds$i ${cl_user1[mds$i]}) + [ -n "$user_rec1" ] || + error "mds$i: User ${cl_user1[mds$i]} not registered" + __changelog_clear mds$i ${cl_user1[mds$i]} +2 + user_rec2=$(changelog_user_rec mds$i ${cl_user1[mds$i]}) + [ -n "$user_rec2" ] || + error "mds$i: User ${cl_user1[mds$i]} not registered" + echo "mds$i: verifying user ${cl_user1[mds$i]} clear: " \ + "$user_rec1 + 2 == $user_rec2" + [ $((user_rec1 + 2)) == $user_rec2 ] || + error "mds$i: user ${cl_user1[mds$i]} index expected " \ + "$user_rec1 + 2, but is $user_rec2" + user_rec2=$(changelog_user_rec mds$i ${cl_user2[mds$i]}) + [ -n "$user_rec2" ] || + error "mds$i: User ${cl_user2[mds$i]} not registered" + [ $user_rec1 == $user_rec2 ] || + error "mds$i: user ${cl_user2[mds$i]} index expected " \ + "$user_rec1, but is $user_rec2" + done + + # force cl_user2 to be idle (2nd part) and to reach + # changelog_max_idle_time + sleep 2 + + # force each GC-thread start and block then + # one per MDT/MDD, set fail_val accordingly + #define OBD_FAIL_FORCE_GC_THREAD 0x1316 + do_nodes $mdts $LCTL set_param fail_loc=0x1316 + + # generate more changelogs to trigger fail_loc + createmany -m $DIR/$tdir/${tfile}bis $((MDSCOUNT * 2)) || + error "create $DIR/$tdir/${tfile}bis failed" + + # stop MDT to stop GC-thread, should be done in back-ground as it will + # block waiting for the thread to be released and exit + declare -A stop_pids + for i in $(seq $MDSCOUNT); do + stop mds$i & + stop_pids[mds$i]=$! + done + + for i in $(mdts_nodes); do + local facet + local nb=0 + local facets=$(facets_up_on_host $i) + + for facet in ${facets//,/ }; do + if [[ $facet == mds* ]]; then + nb=$((nb + 1)) + fi + done + # ensure each MDS's gc threads are still present and all in "R" + # state (OBD_FAIL_FORCE_GC_THREAD effect!) + [[ $(do_node $i pgrep chlg_gc_thread | wc -l) -eq $nb ]] || + error "$i: expected $nb GC-thread" + wait_update $i \ + "ps -C chlg_gc_thread -o state --no-headers | uniq" \ + "R" 20 || + error "$i: GC-thread not found in R-state" + # check umounts of each MDT on MDS have reached kthread_stop() + [[ $(do_node $i pgrep umount | wc -l) -eq $nb ]] || + error "$i: expected $nb umount" + wait_update $i \ + "ps -C umount -o state --no-headers | uniq" "D" 20 || + error "$i: umount not found in D-state" + done + + # release all GC-threads + do_nodes $mdts $LCTL set_param fail_loc=0 + + # wait for MDT stop to complete + for i in $(seq $MDSCOUNT); do + wait ${stop_pids[mds$i]} || error "mds$i: stop failed" + done + + # XXX + # may try to check if any orphan changelog records are present + # via ldiskfs/zfs and llog_reader... + + # re-start/mount MDTs + for i in $(seq $MDSCOUNT); do + start mds$i $(mdsdevname $i) $MDS_MOUNT_OPTS || + error "Fail to start mds$i" + done + + local first_rec + for i in $(seq $MDSCOUNT); do + # check cl_user1 still registered + changelog_users mds$i | grep -q "${cl_user1[mds$i]}" || + error "mds$i: User ${cl_user1[mds$i]} not registered" + # check cl_user2 unregistered + changelog_users mds$i | grep -q "${cl_user2[mds$i]}" && + error "mds$i: User ${cl_user2[mds$i]} still registered" + + # check changelogs are present and starting at $user_rec1 + 1 + user_rec1=$(changelog_user_rec mds$i ${cl_user1[mds$i]}) + [ -n "$user_rec1" ] || + error "mds$i: User ${cl_user1[mds$i]} not registered" + first_rec=$($LFS changelog $(facet_svc mds$i) | + awk '{ print $1; exit; }') + + echo "mds$i: verifying first index $user_rec1 + 1 == $first_rec" + [ $((user_rec1 + 1)) == $first_rec ] || + error "mds$i: first index should be $user_rec1 + 1, " \ + "but is $first_rec" + done } -run_test 160g "changelog garbage collect (old users)" +run_test 160h "changelog gc thread stop upon umount, orphan records delete " \ + "during mount" + +test_160i() { + + local mdts=$(comma_list $(mdts_nodes)) + + changelog_register || error "first changelog_register failed" + + # generate some changelog records to accumulate on each MDT + test_mkdir -c $MDSCOUNT $DIR/$tdir || error "mkdir $tdir failed" + createmany -m $DIR/$tdir/$tfile $((MDSCOUNT * 2)) || + error "create $DIR/$tdir/$tfile failed" + + # check changelogs have been generated + local nbcl=$(changelog_dump | wc -l) + [[ $nbcl -eq 0 ]] && error "no changelogs found" + + # simulate race between register and unregister + # XXX as fail_loc is set per-MDS, with DNE configs the race + # simulation will only occur for one MDT per MDS and for the + # others the normal race scenario will take place + #define CFS_FAIL_CHLOG_USER_REG_UNREG_RACE 0x1315 + do_nodes $mdts $LCTL set_param fail_loc=0x10001315 + do_nodes $mdts $LCTL set_param fail_val=1 + + # unregister 1st user + changelog_deregister & + local pid1=$! + # wait some time for deregister work to reach race rdv + sleep 2 + # register 2nd user + changelog_register || error "2nd user register failed" + + wait $pid1 || error "1st user deregister failed" + + local i + local last_rec + declare -A LAST_REC + for i in $(seq $MDSCOUNT); do + if changelog_users mds$i | grep "^cl"; then + # make sure new records are added with one user present + LAST_REC[mds$i]=$(changelog_users $SINGLEMDS | + awk '/^current.index:/ { print $NF }') + else + error "mds$i has no user registered" + fi + done + + # generate more changelog records to accumulate on each MDT + createmany -m $DIR/$tdir/${tfile}bis $((MDSCOUNT * 2)) || + error "create $DIR/$tdir/${tfile}bis failed" + + for i in $(seq $MDSCOUNT); do + last_rec=$(changelog_users $SINGLEMDS | + awk '/^current.index:/ { print $NF }') + echo "verify changelogs are on: $last_rec != ${LAST_REC[mds$i]}" + [ $last_rec != ${LAST_REC[mds$i]} ] || + error "changelogs are off on mds$i" + done +} +run_test 160i "changelog user register/unregister race" test_161a() { - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return + [ $PARALLEL == "yes" ] && skip "skip parallel run" + test_mkdir -c1 $DIR/$tdir cp /etc/hosts $DIR/$tdir/$tfile test_mkdir -c1 $DIR/$tdir/foo1 @@ -12128,8 +12881,9 @@ test_161a() { run_test 161a "link ea sanity" test_161b() { - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return - [ $MDSCOUNT -lt 2 ] && skip "skipping remote directory test" && return + [ $PARALLEL == "yes" ] && skip "skip parallel run" + [ $MDSCOUNT -lt 2 ] && skip_env "skipping remote directory test" + local MDTIDX=1 local remote_dir=$DIR/$tdir/remote_dir @@ -12180,10 +12934,10 @@ test_161b() { run_test 161b "link ea sanity under remote directory" test_161c() { - remote_mds_nodsh && skip "remote MDS with nodsh" && return - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return + remote_mds_nodsh && skip "remote MDS with nodsh" + [ $PARALLEL == "yes" ] && skip "skip parallel run" [[ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.1.5) ]] && - skip "Need MDS version at least 2.1.5" && return + skip "Need MDS version at least 2.1.5" # define CLF_RENAME_LAST 0x0001 # rename overwrite a target having nlink = 1 (changelog flag 0x1) @@ -12254,7 +13008,7 @@ test_161c() { run_test 161c "check CL_RENME[UNLINK] changelog record flags" test_161d() { - remote_mds_nodsh && skip "remote MDS with nodsh" && return + remote_mds_nodsh && skip "remote MDS with nodsh" local pid local fid @@ -12353,8 +13107,8 @@ test_162a() { # was test_162 run_test 162a "path lookup sanity" test_162b() { - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return - [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return + [ $PARALLEL == "yes" ] && skip "skip parallel run" + [ $MDSCOUNT -lt 2 ] && skip_env "needs >= 2 MDTs" mkdir $DIR/$tdir $LFS setdirstripe -i0 -c$MDSCOUNT -H all_char $DIR/$tdir/striped_dir || @@ -12385,12 +13139,14 @@ run_test 162b "striped directory path lookup sanity" # LU-4239: Verify fid2path works with paths 100 or more directories deep test_162c() { [[ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.7.51) ]] && - skip "Need MDS version at least 2.7.51" && return - test_mkdir $DIR/$tdir.local - test_mkdir $DIR/$tdir.remote + skip "Need MDS version at least 2.7.51" + local lpath=$tdir.local local rpath=$tdir.remote + test_mkdir $DIR/$lpath + test_mkdir $DIR/$rpath + for ((i = 0; i <= 101; i++)); do lpath="$lpath/$i" mkdir $DIR/$lpath @@ -12427,7 +13183,7 @@ test_169() { run_test 169 "parallel read and truncate should not deadlock" test_170() { - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return + [ $PARALLEL == "yes" ] && skip "skip parallel run" $LCTL clear # bug 18514 $LCTL debug_daemon start $TMP/${tfile}_log_good @@ -12482,21 +13238,22 @@ test_170() { run_test 170 "test lctl df to handle corrupted log =====================" test_171() { # bug20592 - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return -#define OBD_FAIL_PTLRPC_DUMP_LOG 0x50e - $LCTL set_param fail_loc=0x50e - $LCTL set_param fail_val=3000 - multiop_bg_pause $DIR/$tfile O_s || true - local MULTIPID=$! - kill -USR1 $MULTIPID - # cause log dump - sleep 3 - wait $MULTIPID - if dmesg | grep "recursive fault"; then - error "caught a recursive fault" - fi - $LCTL set_param fail_loc=0 - true + [ $PARALLEL == "yes" ] && skip "skip parallel run" + + #define OBD_FAIL_PTLRPC_DUMP_LOG 0x50e + $LCTL set_param fail_loc=0x50e + $LCTL set_param fail_val=3000 + multiop_bg_pause $DIR/$tfile O_s || true + local MULTIPID=$! + kill -USR1 $MULTIPID + # cause log dump + sleep 3 + wait $MULTIPID + if dmesg | grep "recursive fault"; then + error "caught a recursive fault" + fi + $LCTL set_param fail_loc=0 + true } run_test 171 "test libcfs_debug_dumplog_thread stuck in do_exit() ======" @@ -12561,69 +13318,77 @@ obdecho_test() { } test_180a() { - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return - remote_ost_nodsh && skip "remote OST with nodsh" && return - local rc=0 - local rmmod_local=0 + [ $PARALLEL == "yes" ] && skip "skip parallel run" if ! module_loaded obdecho; then - load_module obdecho/obdecho - rmmod_local=1 + load_module obdecho/obdecho && + stack_trap "rmmod obdecho" EXIT || + error "unable to load obdecho on client" fi local osc=$($LCTL dl | grep -v mdt | awk '$3 == "osc" {print $4; exit}') - local host=$(lctl get_param -n osc.$osc.import | - awk '/current_connection:/ {print $2}' ) - local target=$(lctl get_param -n osc.$osc.import | - awk '/target:/ {print $2}' ) + local host=$($LCTL get_param -n osc.$osc.import | + awk '/current_connection:/ { print $2 }' ) + local target=$($LCTL get_param -n osc.$osc.import | + awk '/target:/ { print $2 }' ) target=${target%_UUID} - [[ -n $target ]] && { setup_obdecho_osc $host $target || rc=1; } || rc=1 - [ $rc -eq 0 ] && { obdecho_test ${target}_osc client || rc=2; } - [[ -n $target ]] && cleanup_obdecho_osc $target - [ $rmmod_local -eq 1 ] && rmmod obdecho - return $rc + if [ -n "$target" ]; then + setup_obdecho_osc $host $target && + stack_trap "cleanup_obdecho_osc $target" EXIT || + { error "obdecho setup failed with $?"; return; } + + obdecho_test ${target}_osc client || + error "obdecho_test failed on ${target}_osc" + else + $LCTL get_param osc.$osc.import + error "there is no osc.$osc.import target" + fi } run_test 180a "test obdecho on osc" test_180b() { - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return - remote_ost_nodsh && skip "remote OST with nodsh" && return - local rc=0 - local rmmod_remote=0 + [ $PARALLEL == "yes" ] && skip "skip parallel run" + remote_ost_nodsh && skip "remote OST with nodsh" do_rpc_nodes $(facet_active_host ost1) load_module obdecho/obdecho && - rmmod_remote=true || error "failed to load module obdecho" - target=$(do_facet ost1 $LCTL dl | awk '/obdfilter/ {print $4;exit}') - [[ -n $target ]] && { obdecho_test $target ost1 || rc=1; } - $rmmod_remote && do_facet ost1 "rmmod obdecho" - return $rc + stack_trap "do_facet ost1 rmmod obdecho" EXIT || + error "failed to load module obdecho" + + local target=$(do_facet ost1 $LCTL dl | + awk '/obdfilter/ { print $4; exit; }') + + if [ -n "$target" ]; then + obdecho_test $target ost1 || error "obdecho_test failed with $?" + else + do_facet ost1 $LCTL dl + error "there is no obdfilter target on ost1" + fi } run_test 180b "test obdecho directly on obdfilter" test_180c() { # LU-2598 - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return - remote_ost_nodsh && skip "remote OST with nodsh" && return + [ $PARALLEL == "yes" ] && skip "skip parallel run" + remote_ost_nodsh && skip "remote OST with nodsh" [[ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.4.0) ]] && - skip "Need MDS version at least 2.4.0" && return - - local rc=0 - local rmmod_remote=false - local pages=16384 # 64MB bulk I/O RPC size - local target + skip "Need MDS version at least 2.4.0" do_rpc_nodes $(facet_active_host ost1) load_module obdecho/obdecho && - rmmod_remote=true || error "failed to load module obdecho" + stack_trap "do_facet ost1 rmmod obdecho" EXIT || + error "failed to load module obdecho" + + local target=$(do_facet ost1 $LCTL dl | + awk '/obdfilter/ { print $4; exit; }') - target=$(do_facet ost1 $LCTL dl | awk '/obdfilter/ { print $4; exit; }') if [ -n "$target" ]; then - obdecho_test "$target" ost1 "$pages" || rc=${PIPESTATUS[0]} + local pages=16384 # 64MB bulk I/O RPC size + + obdecho_test "$target" ost1 "$pages" || + error "obdecho_test with pages=$pages failed with $?" else - echo "there is no obdfilter target on ost1" - rc=2 + do_facet ost1 $LCTL dl + error "there is no obdfilter target on ost1" fi - $rmmod_remote && do_facet ost1 "rmmod obdecho" || true - return $rc } run_test 180c "test huge bulk I/O size on obdfilter, don't LASSERT" @@ -12675,11 +13440,11 @@ test_182() { run_test 182 "Test parallel modify metadata operations ================" test_183() { # LU-2275 - remote_mds_nodsh && skip "remote MDS with nodsh" && return + [ $PARALLEL == "yes" ] && skip "skip parallel run" + remote_mds_nodsh && skip "remote MDS with nodsh" [[ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.3.56) ]] && - skip "Need MDS version at least 2.3.56" && return + skip "Need MDS version at least 2.3.56" - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return mkdir -p $DIR/$tdir || error "creating dir $DIR/$tdir" echo aaa > $DIR/$tdir/$tfile @@ -12702,7 +13467,7 @@ run_test 183 "No crash or request leak in case of strange dispositions ========" # test suite 184 is for LU-2016, LU-2017 test_184a() { - check_swap_layouts_support && return 0 + check_swap_layouts_support dir0=$DIR/$tdir/$testnum test_mkdir -p -c1 $dir0 @@ -12733,7 +13498,7 @@ test_184a() { run_test 184a "Basic layout swap" test_184b() { - check_swap_layouts_support && return 0 + check_swap_layouts_support dir0=$DIR/$tdir/$testnum mkdir -p $dir0 || error "creating dir $dir0" @@ -12764,8 +13529,8 @@ run_test 184b "Forbidden layout swap (will generate errors)" test_184c() { local cmpn_arg=$(cmp -n 2>&1 | grep "invalid option") - [ -n "$cmpn_arg" ] && skip_env "cmp does not support -n" && return - check_swap_layouts_support && return 0 + [ -n "$cmpn_arg" ] && skip_env "cmp does not support -n" + check_swap_layouts_support local dir0=$DIR/$tdir/$testnum mkdir -p $dir0 || error "creating dir $dir0" @@ -12812,9 +13577,9 @@ test_184c() { run_test 184c "Concurrent write and layout swap" test_184d() { - check_swap_layouts_support && return 0 + check_swap_layouts_support [ -z "$(which getfattr 2>/dev/null)" ] && - skip "no getfattr command" && return 0 + skip_env "no getfattr command" local file1=$DIR/$tdir/$tfile-1 local file2=$DIR/$tdir/$tfile-2 @@ -12847,10 +13612,10 @@ run_test 184d "allow stripeless layouts swap" test_184e() { [[ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.6.94) ]] || - { skip "Need MDS version at least 2.6.94"; return 0; } - check_swap_layouts_support && return 0 + skip "Need MDS version at least 2.6.94" + check_swap_layouts_support [ -z "$(which getfattr 2>/dev/null)" ] && - skip "no getfattr command" && return 0 + skip_env "no getfattr command" local file1=$DIR/$tdir/$tfile-1 local file2=$DIR/$tdir/$tfile-2 @@ -12881,10 +13646,24 @@ test_184e() { } run_test 184e "Recreate layout after stripeless layout swaps" +test_184f() { + # Create a file with name longer than sizeof(struct stat) == + # 144 to see if we can get chars from the file name to appear + # in the returned striping. Note that 'f' == 0x66. + local file=$(for ((i = 0; i < 200; i++)); do echo -n f; done) + + mkdir -p $DIR/$tdir + mcreate $DIR/$tdir/$file + if lfs find --stripe-count 0x6666 $DIR/$tdir | grep $file; then + error "IOC_MDC_GETFILEINFO returned garbage striping" + fi +} +run_test 184f "IOC_MDC_GETFILEINFO for files with long names but no striping" + test_185() { # LU-2441 # LU-3553 - no volatile file support in old servers [[ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.3.60) ]] || - { skip "Need MDS version at least 2.3.60"; return 0; } + skip "Need MDS version at least 2.3.60" mkdir -p $DIR/$tdir || error "creating dir $DIR/$tdir" touch $DIR/$tdir/spoo @@ -12919,9 +13698,9 @@ test_185() { # LU-2441 run_test 185 "Volatile file support" test_187a() { - remote_mds_nodsh && skip "remote MDS with nodsh" && return + remote_mds_nodsh && skip "remote MDS with nodsh" [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.3.0) ] && - skip "Need MDS version at least 2.3.0" && return + skip "Need MDS version at least 2.3.0" local dir0=$DIR/$tdir/$testnum mkdir -p $dir0 || error "creating dir $dir0" @@ -12940,9 +13719,9 @@ test_187a() { run_test 187a "Test data version change" test_187b() { - remote_mds_nodsh && skip "remote MDS with nodsh" && return + remote_mds_nodsh && skip "remote MDS with nodsh" [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.3.0) ] && - skip "Need MDS version at least 2.3.0" && return + skip "Need MDS version at least 2.3.0" local dir0=$DIR/$tdir/$testnum mkdir -p $dir0 || error "creating dir $dir0" @@ -12958,9 +13737,9 @@ test_187b() { run_test 187b "Test data version change on volatile file" test_200() { - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return - remote_mgs_nodsh && skip "remote MGS with nodsh" && return - [ -n "$FILESET" ] && skip "SKIP due to FILESET set" && return + [ $PARALLEL == "yes" ] && skip "skip parallel run" + remote_mgs_nodsh && skip "remote MGS with nodsh" + [ -n "$FILESET" ] && skip "SKIP due to FILESET set" local POOL=${POOL:-cea1} local POOL_ROOT=${POOL_ROOT:-$DIR/d200.pools} @@ -13137,6 +13916,7 @@ else JOBENV=FAKE_JOBID fi fi +LUSTRE_JOBID_SIZE=31 # plus NUL terminator verify_jobstats() { local cmd=($1) @@ -13155,16 +13935,16 @@ verify_jobstats() { [ "$JOBENV" = "FAKE_JOBID" ] && FAKE_JOBID=id.$testnum.$(basename ${cmd[0]}).$RANDOM - JOBVAL=${!JOBENV} + JOBVAL=${!JOBENV:0:$LUSTRE_JOBID_SIZE} [ "$JOBENV" = "nodelocal" ] && { - FAKE_JOBID=id.$testnum.$(basename ${cmd[0]}).$RANDOM + FAKE_JOBID=id.$testnum.%e.$RANDOM $LCTL set_param jobid_name=$FAKE_JOBID - JOBVAL=$FAKE_JOBID + JOBVAL=${FAKE_JOBID/\%e/$(basename ${cmd[0]})} } log "Test: ${cmd[*]}" - log "Using JobID environment variable $JOBENV=$JOBVAL" + log "Using JobID environment $($LCTL get_param -n jobid_var)=$JOBVAL" if [ $JOBENV = "FAKE_JOBID" ]; then FAKE_JOBID=$JOBVAL ${cmd[*]} @@ -13175,8 +13955,10 @@ verify_jobstats() { # all files are created on OST0000 for facet in $facets; do local stats="*.$(convert_facet2label $facet).job_stats" + + # strip out libtool wrappers for in-tree executables if [ $(do_facet $facet lctl get_param $stats | - grep -c $JOBVAL) -ne 1 ]; then + sed -e 's/\.lt-/./' | grep -c $JOBVAL) -ne 1 ]; then do_facet $facet lctl get_param $stats error "No jobstats for $JOBVAL found on $facet::$stats" fi @@ -13186,30 +13968,30 @@ verify_jobstats() { jobstats_set() { local new_jobenv=$1 - do_facet mgs $LCTL conf_param $FSNAME.sys.jobid_var=$new_jobenv - wait_update $HOSTNAME "$LCTL get_param -n jobid_var" $new_jobenv + set_persistent_param_and_check client "jobid_var" \ + "$FSNAME.sys.jobid_var" $new_jobenv } test_205() { # Job stats + [ $PARALLEL == "yes" ] && skip "skip parallel run" [[ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.7.1) ]] || - { skip "Need MDS version with at least 2.7.1"; return 0; } - - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return - remote_mgs_nodsh && skip "remote MGS with nodsh" && return - remote_mds_nodsh && skip "remote MDS with nodsh" && return - remote_ost_nodsh && skip "remote OST with nodsh" && return - + skip "Need MDS version with at least 2.7.1" + remote_mgs_nodsh && skip "remote MGS with nodsh" + remote_mds_nodsh && skip "remote MDS with nodsh" + remote_ost_nodsh && skip "remote OST with nodsh" [ -z "$(lctl get_param -n mdc.*.connect_flags | grep jobstats)" ] && - skip "Server doesn't support jobstats" && return 0 - [[ $JOBID_VAR = disable ]] && skip "jobstats is disabled" && return + skip "Server doesn't support jobstats" + [[ $JOBID_VAR = disable ]] && skip_env "jobstats is disabled" local old_jobenv=$($LCTL get_param -n jobid_var) - if [ $old_jobenv != $JOBENV ]; then - jobstats_set $JOBENV - stack_trap "do_facet mgs \ - $LCTL conf_param $FSNAME.sys.jobid_var=$old_jobenv" EXIT - fi + [ $old_jobenv != $JOBENV ] && jobstats_set $JOBENV + if [[ $PERM_CMD = *"set_param -P"* ]]; then + stack_trap "do_facet mgs $PERM_CMD jobid_var=$old_jobenv" EXIT + else + stack_trap "do_facet mgs $PERM_CMD \ + $FSNAME.sys.jobid_var=$old_jobenv" EXIT + fi changelog_register local old_interval=$(do_facet $SINGLEMDS lctl get_param -n \ @@ -13285,6 +14067,12 @@ test_205() { # Job stats [ $jobids -eq 0 ] || error "Unexpected jobids when jobid_var=$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" + + verify_jobstats "touch $DIR/$tfile" $SINGLEMDS } run_test 205 "Verify job stats" @@ -13339,9 +14127,9 @@ test_208() { # for now as only exclusive open is supported. After generic lease # is done, this test suite should be revised. - Jinshan - remote_mds_nodsh && skip "remote MDS with nodsh" && return + remote_mds_nodsh && skip "remote MDS with nodsh" [[ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.4.52) ]] || - { skip "Need MDS version at least 2.4.52"; return 0; } + skip "Need MDS version at least 2.4.52" echo "==== test 1: verify get lease work" $MULTIOP $DIR/$tfile oO_CREAT:O_RDWR:eRE+eU || error "get lease error" @@ -13388,7 +14176,7 @@ run_test 208 "Exclusive open" test_209() { [ -z "$(lctl get_param -n mdc.*.connect_flags | grep disp_stripe)" ] && - skip_env "must have disp_stripe" && return + skip_env "must have disp_stripe" touch $DIR/$tfile sync; sleep 5; sync; @@ -13487,7 +14275,8 @@ check_lnet_proc_entry() { } test_215() { # for bugs 18102, 21079, 21517 - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return + [ $PARALLEL == "yes" ] && skip "skip parallel run" + local N='(0|[1-9][0-9]*)' # non-negative numeric local P='[1-9][0-9]*' # positive numeric local I='(0|-?[1-9][0-9]*|NA)' # any numeric (0 | >0 | <0) or NA if no value @@ -13566,8 +14355,8 @@ test_215() { # for bugs 18102, 21079, 21517 run_test 215 "lnet exists and has proper content - bugs 18102, 21079, 21517" test_216() { # bug 20317 - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return - remote_ost_nodsh && skip "remote OST with nodsh" && return + [ $PARALLEL == "yes" ] && skip "skip parallel run" + remote_ost_nodsh && skip "remote OST with nodsh" local node local facets=$(get_facets OST) @@ -13610,7 +14399,8 @@ test_216() { # bug 20317 run_test 216 "check lockless direct write updates file size and kms correctly" test_217() { # bug 22430 - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return + [ $PARALLEL == "yes" ] && skip "skip parallel run" + local node local nid @@ -13641,21 +14431,22 @@ test_218() { log "removing the temporary file" rm -rf $DIR/$tfile || error "tmp file removal failed" } -run_test 218 "parallel read and truncate should not deadlock =======================" +run_test 218 "parallel read and truncate should not deadlock" test_219() { - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return - # write one partial page - dd if=/dev/zero of=$DIR/$tfile bs=1024 count=1 - # set no grant so vvp_io_commit_write will do sync write - $LCTL set_param fail_loc=0x411 - # write a full page at the end of file - dd if=/dev/zero of=$DIR/$tfile bs=4096 count=1 seek=1 conv=notrunc - - $LCTL set_param fail_loc=0 - dd if=/dev/zero of=$DIR/$tfile bs=4096 count=1 seek=3 - $LCTL set_param fail_loc=0x411 - dd if=/dev/zero of=$DIR/$tfile bs=1024 count=1 seek=2 conv=notrunc + [ $PARALLEL == "yes" ] && skip "skip parallel run" + + # write one partial page + dd if=/dev/zero of=$DIR/$tfile bs=1024 count=1 + # set no grant so vvp_io_commit_write will do sync write + $LCTL set_param fail_loc=0x411 + # write a full page at the end of file + dd if=/dev/zero of=$DIR/$tfile bs=4096 count=1 seek=1 conv=notrunc + + $LCTL set_param fail_loc=0 + dd if=/dev/zero of=$DIR/$tfile bs=4096 count=1 seek=3 + $LCTL set_param fail_loc=0x411 + dd if=/dev/zero of=$DIR/$tfile bs=1024 count=1 seek=2 conv=notrunc # LU-4201 dd if=/dev/zero of=$DIR/$tfile-2 bs=1024 count=1 @@ -13664,10 +14455,11 @@ test_219() { run_test 219 "LU-394: Write partial won't cause uncontiguous pages vec at LND" test_220() { #LU-325 - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return - remote_ost_nodsh && skip "remote OST with nodsh" && return - remote_mds_nodsh && skip "remote MDS with nodsh" && return - remote_mgs_nodsh && skip "remote MGS with nodsh" && return + [ $PARALLEL == "yes" ] && skip "skip parallel run" + remote_ost_nodsh && skip "remote OST with nodsh" + remote_mds_nodsh && skip "remote MDS with nodsh" + remote_mgs_nodsh && skip "remote MGS with nodsh" + local OSTIDX=0 # create on MDT0000 so the last_id and next_id are correct @@ -13732,20 +14524,22 @@ test_220() { #LU-325 run_test 220 "preallocated MDS objects still used if ENOSPC from OST" test_221() { - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return - dd if=`which date` of=$MOUNT/date oflag=sync - chmod +x $MOUNT/date + [ $PARALLEL == "yes" ] && skip "skip parallel run" + + dd if=`which date` of=$MOUNT/date oflag=sync + chmod +x $MOUNT/date - #define OBD_FAIL_LLITE_FAULT_TRUNC_RACE 0x1401 - $LCTL set_param fail_loc=0x80001401 + #define OBD_FAIL_LLITE_FAULT_TRUNC_RACE 0x1401 + $LCTL set_param fail_loc=0x80001401 - $MOUNT/date > /dev/null - rm -f $MOUNT/date + $MOUNT/date > /dev/null + rm -f $MOUNT/date } run_test 221 "make sure fault and truncate race to not cause OOM" test_222a () { - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return + [ $PARALLEL == "yes" ] && skip "skip parallel run" + rm -rf $DIR/$tdir test_mkdir $DIR/$tdir $LFS setstripe -c 1 -i 0 $DIR/$tdir @@ -13761,7 +14555,8 @@ test_222a () { run_test 222a "AGL for ls should not trigger CLIO lock failure" test_222b () { - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return + [ $PARALLEL == "yes" ] && skip "skip parallel run" + rm -rf $DIR/$tdir test_mkdir $DIR/$tdir $LFS setstripe -c 1 -i 0 $DIR/$tdir @@ -13776,7 +14571,8 @@ test_222b () { run_test 222b "AGL for rmdir should not trigger CLIO lock failure" test_223 () { - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return + [ $PARALLEL == "yes" ] && skip "skip parallel run" + rm -rf $DIR/$tdir test_mkdir $DIR/$tdir $LFS setstripe -c 1 -i 0 $DIR/$tdir @@ -13792,30 +14588,32 @@ test_223 () { run_test 223 "osc reenqueue if without AGL lock granted =======================" test_224a() { # LU-1039, MRP-303 - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return - #define OBD_FAIL_PTLRPC_CLIENT_BULK_CB 0x508 - $LCTL set_param fail_loc=0x508 - dd if=/dev/zero of=$DIR/$tfile bs=4096 count=1 conv=fsync - $LCTL set_param fail_loc=0 - df $DIR + [ $PARALLEL == "yes" ] && skip "skip parallel run" + + #define OBD_FAIL_PTLRPC_CLIENT_BULK_CB 0x508 + $LCTL set_param fail_loc=0x508 + dd if=/dev/zero of=$DIR/$tfile bs=4096 count=1 conv=fsync + $LCTL set_param fail_loc=0 + df $DIR } run_test 224a "Don't panic on bulk IO failure" test_224b() { # LU-1039, MRP-303 - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return - dd if=/dev/zero of=$DIR/$tfile bs=4096 count=1 - cancel_lru_locks osc - #define OBD_FAIL_PTLRPC_CLIENT_BULK_CB2 0x515 - $LCTL set_param fail_loc=0x515 - dd of=/dev/null if=$DIR/$tfile bs=4096 count=1 - $LCTL set_param fail_loc=0 - df $DIR + [ $PARALLEL == "yes" ] && skip "skip parallel run" + + dd if=/dev/zero of=$DIR/$tfile bs=4096 count=1 + cancel_lru_locks osc + #define OBD_FAIL_PTLRPC_CLIENT_BULK_CB2 0x515 + $LCTL set_param fail_loc=0x515 + dd of=/dev/null if=$DIR/$tfile bs=4096 count=1 + $LCTL set_param fail_loc=0 + df $DIR } run_test 224b "Don't panic on bulk IO failure" test_224c() { # LU-6441 - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return - remote_mds_nodsh && skip "remote MDS with nodsh" && return + [ $PARALLEL == "yes" ] && skip "skip parallel run" + remote_mds_nodsh && skip "remote MDS with nodsh" local p="$TMP/$TESTSUITE-$TESTNAME.parameters" save_writethrough $p @@ -13825,17 +14623,15 @@ test_224c() { # LU-6441 osc.*.max_pages_per_rpc) local at_max=$($LCTL get_param -n at_max) local timeout=$($LCTL get_param -n timeout) - local test_at="$LCTL get_param -n at_max" + local test_at="at_max" local param_at="$FSNAME.sys.at_max" - local test_timeout="$LCTL get_param -n timeout" + local test_timeout="timeout" local param_timeout="$FSNAME.sys.timeout" $LCTL set_param -n osc.*.max_pages_per_rpc=1024 - set_conf_param_and_check client "$test_at" "$param_at" 0 || - error "conf_param at_max=0 failed" - set_conf_param_and_check client "$test_timeout" "$param_timeout" 5 || - error "conf_param timeout=5 failed" + set_persistent_param_and_check client "$test_at" "$param_at" 0 + set_persistent_param_and_check client "$test_timeout" "$param_timeout" 5 #define OBD_FAIL_PTLRPC_CLIENT_BULK_CB3 0x520 do_facet ost1 $LCTL set_param fail_loc=0x520 @@ -13844,10 +14640,9 @@ test_224c() { # LU-6441 sync do_facet ost1 $LCTL set_param fail_loc=0 - set_conf_param_and_check client "$test_at" "$param_at" $at_max || - error "conf_param at_max=$at_max failed" - set_conf_param_and_check client "$test_timeout" "$param_timeout" \ - $timeout || error "conf_param timeout=$timeout failed" + set_persistent_param_and_check client "$test_at" "$param_at" $at_max + set_persistent_param_and_check client "$test_timeout" "$param_timeout" \ + $timeout $LCTL set_param -n $pages_per_rpc restore_lustre_params < $p @@ -13857,14 +14652,12 @@ run_test 224c "Don't hang if one of md lost during large bulk RPC" MDSSURVEY=${MDSSURVEY:-$(which mds-survey 2>/dev/null || true)} test_225a () { - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return - remote_mds_nodsh && skip "remote MDS with nodsh" && return + [ $PARALLEL == "yes" ] && skip "skip parallel run" if [ -z ${MDSSURVEY} ]; then - skip_env "mds-survey not found" && return + skip_env "mds-survey not found" fi - [ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.2.51) ] || - { skip "Need MDS version at least 2.2.51"; return; } + skip "Need MDS version at least 2.2.51" local mds=$(facet_host $SINGLEMDS) local target=$(do_nodes $mds 'lctl dl' | @@ -13884,16 +14677,15 @@ test_225a () { run_test 225a "Metadata survey sanity with zero-stripe" test_225b () { - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return - remote_mds_nodsh && skip "remote MDS with nodsh" && return if [ -z ${MDSSURVEY} ]; then - skip_env "mds-survey not found" && return + skip_env "mds-survey not found" fi [ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.2.51) ] || - { skip "Need MDS version at least 2.2.51"; return; } - + skip "Need MDS version at least 2.2.51" + [ $PARALLEL == "yes" ] && skip "skip parallel run" + remote_mds_nodsh && skip "remote MDS with nodsh" if [ $($LCTL dl | grep -c osc) -eq 0 ]; then - skip_env "Need to mount OST to test" && return + skip_env "Need to mount OST to test" fi local mds=$(facet_host $SINGLEMDS) @@ -13957,10 +14749,11 @@ test_226a () { run_test 226a "call path2fid and fid2path on files of all type" test_226b () { - [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return - rm -rf $DIR/$tdir + [ $MDSCOUNT -lt 2 ] && skip_env "needs >= 2 MDTs" + local MDTIDX=1 + rm -rf $DIR/$tdir mkdir -p $DIR/$tdir $LFS setdirstripe -i $MDTIDX $DIR/$tdir/remote_dir || error "create remote directory failed" @@ -13981,8 +14774,9 @@ run_test 226b "call path2fid and fid2path on files of all type under remote dir" # LU-1299 Executing or running ldd on a truncated executable does not # cause an out-of-memory condition. test_227() { - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return - [ -z "$(which ldd)" ] && skip "should have ldd tool" && return + [ $PARALLEL == "yes" ] && skip "skip parallel run" + [ -z "$(which ldd)" ] && skip_env "should have ldd tool" + dd if=$(which date) of=$MOUNT/date bs=1k count=1 chmod +x $MOUNT/date @@ -13994,10 +14788,10 @@ run_test 227 "running truncated executable does not cause OOM" # LU-1512 try to reuse idle OI blocks test_228a() { - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return - remote_mds_nodsh && skip "remote MDS with nodsh" && return + [ $PARALLEL == "yes" ] && skip "skip parallel run" + remote_mds_nodsh && skip "remote MDS with nodsh" [ "$(facet_fstype $SINGLEMDS)" != "ldiskfs" ] && - skip "ldiskfs only test" && return + skip_env "ldiskfs only test" local MDT_DEV=$(mdsdevname ${SINGLEMDS//mds/}) local myDIR=$DIR/$tdir @@ -14036,10 +14830,10 @@ test_228a() { run_test 228a "try to reuse idle OI blocks" test_228b() { - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return - remote_mds_nodsh && skip "remote MDS with nodsh" && return + [ $PARALLEL == "yes" ] && skip "skip parallel run" + remote_mds_nodsh && skip "remote MDS with nodsh" [ "$(facet_fstype $SINGLEMDS)" != "ldiskfs" ] && - skip "ldiskfs only test" && return + skip_env "ldiskfs only test" local MDT_DEV=$(mdsdevname ${SINGLEMDS//mds/}) local myDIR=$DIR/$tdir @@ -14086,10 +14880,10 @@ run_test 228b "idle OI blocks can be reused after MDT restart" #LU-1881 test_228c() { - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return - remote_mds_nodsh && skip "remote MDS with nodsh" && return + [ $PARALLEL == "yes" ] && skip "skip parallel run" + remote_mds_nodsh && skip "remote MDS with nodsh" [ "$(facet_fstype $SINGLEMDS)" != "ldiskfs" ] && - skip "ldiskfs only test" && return + skip_env "ldiskfs only test" local MDT_DEV=$(mdsdevname ${SINGLEMDS//mds/}) local myDIR=$DIR/$tdir @@ -14132,11 +14926,10 @@ test_228c() { run_test 228c "NOT shrink the last entry in OI index node to recycle idle leaf" test_229() { # LU-2482, LU-3448 + [ $PARALLEL == "yes" ] && skip "skip parallel run" + [ $OSTCOUNT -lt 2 ] && skip_env "needs >= 2 OSTs" [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.4.53) ] && - skip "No HSM $(lustre_build_version $SINGLEMDS) MDS < 2.4.53" && - return - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return - [ $OSTCOUNT -lt 2 ] && skip "needs >= 2 OSTs" && return + skip "No HSM $(lustre_build_version $SINGLEMDS) MDS < 2.4.53" rm -f $DIR/$tfile @@ -14165,33 +14958,39 @@ test_229() { # LU-2482, LU-3448 run_test 229 "getstripe/stat/rm/attr changes work on released files" test_230a() { - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return - [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return + [ $PARALLEL == "yes" ] && skip "skip parallel run" + [ $MDSCOUNT -lt 2 ] && skip_env "needs >= 2 MDTs" + [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.11.52) ] && + skip "Need MDS version at least 2.11.52" + local MDTIDX=1 test_mkdir $DIR/$tdir test_mkdir -i0 -c1 $DIR/$tdir/test_230_local - local mdt_idx=$($GETSTRIPE -M $DIR/$tdir/test_230_local) + local mdt_idx=$($LFS getstripe -m $DIR/$tdir/test_230_local) [ $mdt_idx -ne 0 ] && error "create local directory on wrong MDT $mdt_idx" $LFS mkdir -i $MDTIDX $DIR/$tdir/test_230 || error "create remote directory failed" - local mdt_idx=$($GETSTRIPE -M $DIR/$tdir/test_230) + local mdt_idx=$($LFS getstripe -m $DIR/$tdir/test_230) [ $mdt_idx -ne $MDTIDX ] && error "create remote directory on wrong MDT $mdt_idx" createmany -o $DIR/$tdir/test_230/t- 10 || error "create files on remote directory failed" - mdt_idx=$($GETSTRIPE -M $DIR/$tdir/test_230/t-0) + mdt_idx=$($LFS getstripe -m $DIR/$tdir/test_230/t-0) [ $mdt_idx -ne $MDTIDX ] && error "create files on wrong MDT $mdt_idx" rm -r $DIR/$tdir || error "unlink remote directory failed" } run_test 230a "Create remote directory and files under the remote directory" test_230b() { - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return - [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return + [ $PARALLEL == "yes" ] && skip "skip parallel run" + [ $MDSCOUNT -lt 2 ] && skip_env "needs >= 2 MDTs" + [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.11.52) ] && + skip "Need MDS version at least 2.11.52" + local MDTIDX=1 local mdt_index local i @@ -14238,14 +15037,14 @@ test_230b() { echo "migratate to MDT1, then checking.." for ((i = 0; i < 10; i++)); do for file in $(find $migrate_dir/dir_${i}); do - mdt_index=$($LFS getstripe -M $file) + mdt_index=$($LFS getstripe -m $file) [ $mdt_index == $MDTIDX ] || error "$file is not on MDT${MDTIDX}" done done # the multiple link file should still in MDT0 - mdt_index=$($LFS getstripe -M $migrate_dir/$tfile) + mdt_index=$($LFS getstripe -m $migrate_dir/$tfile) [ $mdt_index == 0 ] || error "$file is not on MDT${MDTIDX}" @@ -14302,7 +15101,7 @@ test_230b() { echo "migrate back to MDT0, checking.." for file in $(find $migrate_dir); do - mdt_index=$($LFS getstripe -M $file) + mdt_index=$($LFS getstripe -m $file) [ $mdt_index == $MDTIDX ] || error "$file is not on MDT${MDTIDX}" done @@ -14354,10 +15153,14 @@ test_230b() { run_test 230b "migrate directory" test_230c() { - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return - remote_mds_nodsh && skip "remote MDS with nodsh" && return - [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return + [ $PARALLEL == "yes" ] && skip "skip parallel run" + [ $MDSCOUNT -lt 2 ] && skip_env "needs >= 2 MDTs" + remote_mds_nodsh && skip "remote MDS with nodsh" + [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.11.52) ] && + skip "Need MDS version at least 2.11.52" + local MDTIDX=1 + local total=3 local mdt_index local file local migrate_dir=$DIR/$tdir/migrate_dir @@ -14366,58 +15169,100 @@ test_230c() { #the directory is still accessiable. test_mkdir $DIR/$tdir test_mkdir -i0 -c1 $migrate_dir + test_mkdir -i1 -c1 $DIR/$tdir/remote_dir stat $migrate_dir - createmany -o $migrate_dir/f 10 || + createmany -o $migrate_dir/f $total || error "create files under ${migrate_dir} failed" - #failed after migrating 5 entries + # fail after migrating top dir, and this will fail only once, so the + # first sub file migration will fail (currently f3), others succeed. #OBD_FAIL_MIGRATE_ENTRIES 0x1801 - do_facet mds1 lctl set_param fail_loc=0x20001801 - do_facet mds1 lctl set_param fail_val=5 + do_facet mds1 lctl set_param fail_loc=0x1801 local t=$(ls $migrate_dir | wc -l) $LFS migrate --mdt-index $MDTIDX $migrate_dir && - error "migrate should fail after 5 entries" - - mkdir $migrate_dir/dir && - error "mkdir succeeds under migrating directory" - touch $migrate_dir/file && - error "touch file succeeds under migrating directory" - + error "migrate should fail" local u=$(ls $migrate_dir | wc -l) [ "$u" == "$t" ] || error "$u != $t during migration" - for file in $(find $migrate_dir); do - stat $file || error "stat $file failed" + # add new dir/file should succeed + mkdir $migrate_dir/dir || + error "mkdir failed under migrating directory" + touch $migrate_dir/file || + error "create file failed under migrating directory" + + # add file with existing name should fail + for file in $migrate_dir/f*; do + stat $file > /dev/null || error "stat $file failed" + $OPENFILE -f O_CREAT:O_EXCL $file && + error "open(O_CREAT|O_EXCL) $file should fail" + $MULTIOP $file m && error "create $file should fail" + touch $DIR/$tdir/remote_dir/$tfile || + error "touch $tfile failed" + ln $DIR/$tdir/remote_dir/$tfile $file && + error "link $file should fail" + mdt_index=$($LFS getstripe -m $file) + if [ $mdt_index == 0 ]; then + # file failed to migrate is not allowed to rename to + mv $DIR/$tdir/remote_dir/$tfile $file && + error "rename to $file should fail" + else + mv $DIR/$tdir/remote_dir/$tfile $file || + error "rename to $file failed" + fi + echo hello >> $file || error "write $file failed" done - do_facet mds1 lctl set_param fail_loc=0 - do_facet mds1 lctl set_param fail_val=0 + # resume migration with different options should fail + $LFS migrate -m 0 $migrate_dir && + error "migrate -m 0 $migrate_dir should fail" + + $LFS migrate -m $MDTIDX -c 2 $migrate_dir && + error "migrate -c 2 $migrate_dir should fail" + # resume migration should succeed $LFS migrate -m $MDTIDX $migrate_dir || - error "migrate open files should failed with open files" + error "migrate $migrate_dir failed" echo "Finish migration, then checking.." for file in $(find $migrate_dir); do - mdt_index=$($LFS getstripe -M $file) + mdt_index=$($LFS getstripe -m $file) [ $mdt_index == $MDTIDX ] || error "$file is not on MDT${MDTIDX}" done rm -rf $DIR/$tdir || error "rm dir failed after migration" } -run_test 230c "check directory accessiblity if migration is failed" +run_test 230c "check directory accessiblity if migration failed" test_230d() { - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return - [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return - local MDTIDX=1 - local mdt_index + [ $PARALLEL == "yes" ] && skip "skip parallel run" + [ $MDSCOUNT -lt 2 ] && skip_env "needs >= 2 MDTs" + [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.11.52) ] && + skip "Need MDS version at least 2.11.52" + # LU-11235 + [ "$(facet_fstype mds1)" == "zfs" ] && skip "skip ZFS backend" + local migrate_dir=$DIR/$tdir/migrate_dir + local old_index + local new_index + local old_count + local new_count + local new_hash + local mdt_index local i local j + old_index=$((RANDOM % MDSCOUNT)) + old_count=$((MDSCOUNT - old_index)) + new_index=$((RANDOM % MDSCOUNT)) + new_count=$((MDSCOUNT - new_index)) + new_hash="all_char" + + [ $old_count -gt 1 ] && old_count=$((old_count - RANDOM % old_count)) + [ $new_count -gt 1 ] && new_count=$((new_count - RANDOM % new_count)) + test_mkdir $DIR/$tdir - test_mkdir -i0 -c1 $migrate_dir + test_mkdir -i $old_index -c $old_count $migrate_dir for ((i=0; i<100; i++)); do test_mkdir -i0 -c1 $migrate_dir/dir_${i} @@ -14425,14 +15270,23 @@ test_230d() { error "create files under remote dir failed $i" done - $LFS migrate -m $MDTIDX $migrate_dir || + echo -n "Migrate from MDT$old_index " + [ $old_count -gt 1 ] && echo -n "... MDT$((old_index + old_count - 1)) " + echo -n "to MDT$new_index" + [ $new_count -gt 1 ] && echo -n " ... MDT$((new_index + new_count - 1))" + echo + + echo "$LFS migrate -m$new_index -c$new_count -H $new_hash $migrate_dir" + $LFS migrate -m $new_index -c $new_count -H $new_hash $migrate_dir || error "migrate remote dir error" echo "Finish migration, then checking.." for file in $(find $migrate_dir); do - mdt_index=$($LFS getstripe -M $file) - [ $mdt_index == $MDTIDX ] || - error "$file is not on MDT${MDTIDX}" + mdt_index=$($LFS getstripe -m $file) + if [ $mdt_index -lt $new_index ] || + [ $mdt_index -gt $((new_index + new_count - 1)) ]; then + error "$file is on MDT$mdt_index" + fi done rm -rf $DIR/$tdir || error "rm dir failed after migration" @@ -14440,8 +15294,11 @@ test_230d() { run_test 230d "check migrate big directory" test_230e() { - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return - [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return + [ $PARALLEL == "yes" ] && skip "skip parallel run" + [ $MDSCOUNT -lt 2 ] && skip_env "needs >= 2 MDTs" + [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.11.52) ] && + skip "Need MDS version at least 2.11.52" + local i local j local a_fid @@ -14457,22 +15314,22 @@ test_230e() { $LFS migrate -m 1 $DIR/$tdir/migrate_dir || error "migrate dir fails" - mdt_index=$($LFS getstripe -M $DIR/$tdir/migrate_dir) + mdt_index=$($LFS getstripe -m $DIR/$tdir/migrate_dir) [ $mdt_index == 1 ] || error "migrate_dir is not on MDT1" - mdt_index=$($LFS getstripe -M $DIR/$tdir/migrate_dir/a) + mdt_index=$($LFS getstripe -m $DIR/$tdir/migrate_dir/a) [ $mdt_index == 0 ] || error "a is not on MDT0" $LFS migrate -m 1 $DIR/$tdir/other_dir || error "migrate dir fails" - mdt_index=$($LFS getstripe -M $DIR/$tdir/other_dir) + mdt_index=$($LFS getstripe -m $DIR/$tdir/other_dir) [ $mdt_index == 1 ] || error "other_dir is not on MDT1" - mdt_index=$($LFS getstripe -M $DIR/$tdir/migrate_dir/a) + mdt_index=$($LFS getstripe -m $DIR/$tdir/migrate_dir/a) [ $mdt_index == 1 ] || error "a is not on MDT1" - mdt_index=$($LFS getstripe -M $DIR/$tdir/other_dir/b) + mdt_index=$($LFS getstripe -m $DIR/$tdir/other_dir/b) [ $mdt_index == 1 ] || error "b is not on MDT1" a_fid=$($LFS path2fid $DIR/$tdir/migrate_dir/a) @@ -14485,8 +15342,11 @@ test_230e() { run_test 230e "migrate mulitple local link files" test_230f() { - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return - [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return + [ $PARALLEL == "yes" ] && skip "skip parallel run" + [ $MDSCOUNT -lt 2 ] && skip_env "needs >= 2 MDTs" + [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.11.52) ] && + skip "Need MDS version at least 2.11.52" + local a_fid local ln_fid @@ -14501,15 +15361,15 @@ test_230f() { # a should be migrated to MDT1, since no other links on MDT0 $LFS migrate -m 1 $DIR/$tdir/migrate_dir || error "#1 migrate dir fails" - mdt_index=$($LFS getstripe -M $DIR/$tdir/migrate_dir) + mdt_index=$($LFS getstripe -m $DIR/$tdir/migrate_dir) [ $mdt_index == 1 ] || error "migrate_dir is not on MDT1" - mdt_index=$($LFS getstripe -M $DIR/$tdir/migrate_dir/a) + mdt_index=$($LFS getstripe -m $DIR/$tdir/migrate_dir/a) [ $mdt_index == 1 ] || error "a is not on MDT1" # a should stay on MDT1, because it is a mulitple link file $LFS migrate -m 0 $DIR/$tdir/migrate_dir || error "#2 migrate dir fails" - mdt_index=$($LFS getstripe -M $DIR/$tdir/migrate_dir/a) + mdt_index=$($LFS getstripe -m $DIR/$tdir/migrate_dir/a) [ $mdt_index == 1 ] || error "a is not on MDT1" $LFS migrate -m 1 $DIR/$tdir/migrate_dir || @@ -14525,7 +15385,7 @@ test_230f() { # a should be migrated to MDT0, since no other links on MDT1 $LFS migrate -m 0 $DIR/$tdir/migrate_dir || error "#4 migrate dir fails" - mdt_index=$($LFS getstripe -M $DIR/$tdir/migrate_dir/a) + mdt_index=$($LFS getstripe -m $DIR/$tdir/migrate_dir/a) [ $mdt_index == 0 ] || error "a is not on MDT0" rm -rf $DIR/$tdir || error "rm dir failed after migration" @@ -14533,8 +15393,10 @@ test_230f() { run_test 230f "migrate mulitple remote link files" test_230g() { - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return - [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return + [ $PARALLEL == "yes" ] && skip "skip parallel run" + [ $MDSCOUNT -lt 2 ] && skip_env "needs >= 2 MDTs" + [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.11.52) ] && + skip "Need MDS version at least 2.11.52" mkdir -p $DIR/$tdir/migrate_dir @@ -14545,10 +15407,11 @@ test_230g() { run_test 230g "migrate dir to non-exist MDT" test_230h() { - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return - [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return - [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.7.64) ] && - skip "Need MDS version at least 2.7.64" && return + [ $PARALLEL == "yes" ] && skip "skip parallel run" + [ $MDSCOUNT -lt 2 ] && skip_env "needs >= 2 MDTs" + [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.11.52) ] && + skip "Need MDS version at least 2.11.52" + local mdt_index mkdir -p $DIR/$tdir/migrate_dir @@ -14559,21 +15422,19 @@ test_230h() { $LFS migrate -m1 $DIR/$tdir/.. && error "migrating mountpoint2 should fail" - $LFS migrate -m1 $DIR/$tdir/migrate_dir/.. || - error "migrating $tdir fail" - - mdt_index=$($LFS getstripe -M $DIR/$tdir) - [ $mdt_index == 1 ] || error "$mdt_index != 1 after migration" - - mdt_index=$($LFS getstripe -M $DIR/$tdir/migrate_dir) - [ $mdt_index == 1 ] || error "$mdt_index != 1 after migration" + # same as mv + $LFS migrate -m1 $DIR/$tdir/migrate_dir/.. && + error "migrating $tdir/migrate_dir/.. should fail" + true } run_test 230h "migrate .. and root" test_230i() { - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return - [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return + [ $PARALLEL == "yes" ] && skip "skip parallel run" + [ $MDSCOUNT -lt 2 ] && skip_env "needs >= 2 MDTs" + [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.11.52) ] && + skip "Need MDS version at least 2.11.52" mkdir -p $DIR/$tdir/migrate_dir @@ -14585,12 +15446,29 @@ test_230i() { } run_test 230i "lfs migrate -m tolerates trailing slashes" +test_230j() { + [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" + [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.11.52) ] && + skip "Need MDS version at least 2.11.52" + + $LFS mkdir -m 0 -c 1 $DIR/$tdir || error "mkdir $tdir failed" + $LFS setstripe -E 1M -L mdt $DIR/$tdir/$tfile || + error "create $tfile failed" + cat /etc/passwd > $DIR/$tdir/$tfile + + $LFS migrate -m 1 $DIR/$tdir + + cmp /etc/passwd $DIR/$tdir/$tfile || + error "DoM file mismatch after migration" +} +run_test 230j "DoM file data not changed after dir migration" + test_231a() { # For simplicity this test assumes that max_pages_per_rpc # is the same across all OSCs local max_pages=$($LCTL get_param -n osc.*.max_pages_per_rpc | head -n1) - local bulk_size=$((max_pages * 4096)) + local bulk_size=$((max_pages * PAGE_SIZE)) local brw_size=$(do_facet ost1 $LCTL get_param -n obdfilter.*.brw_size | head -n 1) @@ -14664,7 +15542,8 @@ run_test 232a "failed lock should not block umount" test_232b() { [ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.10.58) ] || - { skip "Need MDS version at least 2.10.58"; return; } + skip "Need MDS version at least 2.10.58" + mkdir -p $DIR/$tdir $LFS setstripe -c1 -i0 $DIR/$tdir/$tfile dd if=/dev/zero of=$DIR/$tdir/$tfile bs=1M count=1 @@ -14687,10 +15566,11 @@ run_test 232b "failed data version lock should not block umount" test_233a() { [ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.3.64) ] || - { skip "Need MDS version at least 2.3.64"; return; } - [ -n "$FILESET" ] && skip "SKIP due to FILESET set" && return + skip "Need MDS version at least 2.3.64" + [ -n "$FILESET" ] && skip_env "SKIP due to FILESET set" local fid=$($LFS path2fid $MOUNT) + stat $MOUNT/.lustre/fid/$fid > /dev/null || error "cannot access $MOUNT using its FID '$fid'" } @@ -14698,10 +15578,11 @@ run_test 233a "checking that OBF of the FS root succeeds" test_233b() { [ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.5.90) ] || - { skip "Need MDS version at least 2.5.90"; return; } - [ -n "$FILESET" ] && skip "SKIP due to FILESET set" && return + skip "Need MDS version at least 2.5.90" + [ -n "$FILESET" ] && skip_env "SKIP due to FILESET set" local fid=$($LFS path2fid $MOUNT/.lustre) + stat $MOUNT/.lustre/fid/$fid > /dev/null || error "cannot access $MOUNT/.lustre using its FID '$fid'" @@ -14715,7 +15596,7 @@ test_234() { local p="$TMP/sanityN-$TESTNAME.parameters" save_lustre_params client "llite.*.xattr_cache" > $p lctl set_param llite.*.xattr_cache 1 || - { skip "xattr cache is not supported"; return 0; } + skip_env "xattr cache is not supported" mkdir -p $DIR/$tdir || error "mkdir failed" touch $DIR/$tdir/$tfile || error "touch failed" @@ -14733,7 +15614,8 @@ run_test 234 "xattr cache should not crash on ENOMEM" test_235() { [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.4.52) ] && - skip "Need MDS version at least 2.4.52" && return + skip "Need MDS version at least 2.4.52" + flock_deadlock $DIR/$tfile local RC=$? case $RC in @@ -14749,14 +15631,14 @@ run_test 235 "LU-1715: flock deadlock detection does not work properly" #LU-2935 test_236() { - check_swap_layouts_support && return 0 - test_mkdir -c1 $DIR/$tdir + check_swap_layouts_support local ref1=/etc/passwd local ref2=/etc/group local file1=$DIR/$tdir/f1 local file2=$DIR/$tdir/f2 + test_mkdir -c1 $DIR/$tdir $SETSTRIPE -c 1 $file1 || error "cannot setstripe on '$file1': rc = $?" cp $ref1 $file1 || error "cp $ref1 $file1 failed: rc = $?" $SETSTRIPE -c 2 $file2 || error "cannot setstripe on '$file2': rc = $?" @@ -14776,18 +15658,6 @@ test_236() { } run_test 236 "Layout swap on open unlinked file" -# test to verify file handle related system calls -# (name_to_handle_at/open_by_handle_at) -# The new system calls are supported in glibc >= 2.14. - -test_237() { - echo "Test file_handle syscalls" > $DIR/$tfile || - error "write failed" - check_fhandle_syscalls $DIR/$tfile || - error "check_fhandle_syscalls failed" -} -run_test 237 "Verify name_to_handle_at/open_by_handle_at syscalls" - # LU-4659 linkea consistency test_238() { local server_version=$(lustre_version_code $SINGLEMDS) @@ -14795,7 +15665,7 @@ test_238() { [[ $server_version -gt $(version_code 2.5.57) ]] || [[ $server_version -gt $(version_code 2.5.1) && $server_version -lt $(version_code 2.5.50) ]] || - { skip "Need MDS version at least 2.5.58 or 2.5.2+"; return; } + skip "Need MDS version at least 2.5.58 or 2.5.2+" touch $DIR/$tfile ln $DIR/$tfile $DIR/$tfile.lnk @@ -14814,7 +15684,8 @@ run_test 238 "Verify linkea consistency" test_239A() { # was test_239 [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.5.60) ] && - skip "Need MDS version at least 2.5.60" && return + skip "Need MDS version at least 2.5.60" + local list=$(comma_list $(mdts_nodes)) mkdir -p $DIR/$tdir @@ -14829,7 +15700,8 @@ test_239A() { # was test_239 run_test 239A "osp_sync test" test_239a() { #LU-5297 - remote_mds_nodsh && skip "remote MDS with nodsh" && return + remote_mds_nodsh && skip "remote MDS with nodsh" + touch $DIR/$tfile #define OBD_FAIL_OSP_CHECK_INVALID_REC 0x2100 do_facet $SINGLEMDS $LCTL set_param fail_loc=0x2100 @@ -14839,7 +15711,8 @@ test_239a() { #LU-5297 run_test 239a "process invalid osp sync record correctly" test_239b() { #LU-5297 - remote_mds_nodsh && skip "remote MDS with nodsh" && return + remote_mds_nodsh && skip "remote MDS with nodsh" + touch $DIR/$tfile1 #define OBD_FAIL_OSP_CHECK_ENOMEM 0x2101 do_facet $SINGLEMDS $LCTL set_param fail_loc=0x2101 @@ -14853,8 +15726,8 @@ test_239b() { #LU-5297 run_test 239b "process osp sync record with ENOMEM error correctly" test_240() { - [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return - remote_mds_nodsh && skip "remote MDS with nodsh" && return + [ $MDSCOUNT -lt 2 ] && skip_env "needs >= 2 MDTs" + remote_mds_nodsh && skip "remote MDS with nodsh" mkdir -p $DIR/$tdir @@ -14874,42 +15747,55 @@ test_240() { run_test 240 "race between ldlm enqueue and the connection RPC (no ASSERT)" test_241_bio() { - for LOOP in $(seq $1); do - dd if=$DIR/$tfile of=/dev/null bs=40960 count=1 2>/dev/null + local count=$1 + local bsize=$2 + + for LOOP in $(seq $count); do + dd if=$DIR/$tfile of=/dev/null bs=$bsize count=1 2>/dev/null cancel_lru_locks $OSC || true done } test_241_dio() { + local count=$1 + local bsize=$2 + for LOOP in $(seq $1); do - dd if=$DIR/$tfile of=/dev/null bs=40960 count=1 \ - iflag=direct 2>/dev/null + dd if=$DIR/$tfile of=/dev/null bs=$bsize count=1 iflag=direct \ + 2>/dev/null done } test_241a() { # was test_241 - dd if=/dev/zero of=$DIR/$tfile count=1 bs=40960 + local bsize=$PAGE_SIZE + + (( bsize < 40960 )) && bsize=40960 + dd if=/dev/zero of=$DIR/$tfile count=1 bs=$bsize ls -la $DIR/$tfile cancel_lru_locks $OSC - test_241_bio 1000 & + test_241_bio 1000 $bsize & PID=$! - test_241_dio 1000 + test_241_dio 1000 $bsize wait $PID } run_test 241a "bio vs dio" test_241b() { - dd if=/dev/zero of=$DIR/$tfile count=1 bs=40960 + local bsize=$PAGE_SIZE + + (( bsize < 40960 )) && bsize=40960 + dd if=/dev/zero of=$DIR/$tfile count=1 bs=$bsize ls -la $DIR/$tfile - test_241_dio 1000 & + test_241_dio 1000 $bsize & PID=$! - test_241_dio 1000 + test_241_dio 1000 $bsize wait $PID } run_test 241b "dio vs dio" test_242() { - remote_mds_nodsh && skip "remote MDS with nodsh" && return + remote_mds_nodsh && skip "remote MDS with nodsh" + mkdir -p $DIR/$tdir touch $DIR/$tdir/$tfile @@ -14965,9 +15851,10 @@ test_245() { run_test 245 "check mdc connection flag/data: multiple modify RPCs" test_246() { # LU-7371 - remote_ost_nodsh && skip "remote OST with nodsh" && return + remote_ost_nodsh && skip "remote OST with nodsh" [ $(lustre_version_code ost1) -lt $(version_code 2.7.62) ] && - skip "Need OST version >= 2.7.62" && return 0 + skip "Need OST version >= 2.7.62" + do_facet ost1 $LCTL set_param fail_val=4095 #define OBD_FAIL_OST_READ_SIZE 0x234 do_facet ost1 $LCTL set_param fail_loc=0x234 @@ -14989,7 +15876,7 @@ cleanup_247() { test_247a() { lctl get_param -n mdc.$FSNAME-MDT0000*.import | grep -q subtree || - { skip "Fileset feature is not supported"; return; } + skip_env "Fileset feature is not supported" local submount=${MOUNT}_$tdir @@ -15007,7 +15894,7 @@ run_test 247a "mount subdir as fileset" test_247b() { lctl get_param -n mdc.$FSNAME-MDT0000*.import | grep -q subtree || - { skip "Fileset feature is not supported"; return; } + skip_env "Fileset feature is not supported" local submount=${MOUNT}_$tdir @@ -15022,7 +15909,7 @@ run_test 247b "mount subdir that dose not exist" test_247c() { lctl get_param -n mdc.$FSNAME-MDT0000*.import | grep -q subtree || - { skip "Fileset feature is not supported"; return; } + skip_env "Fileset feature is not supported" local submount=${MOUNT}_$tdir @@ -15039,7 +15926,7 @@ run_test 247c "running fid2path outside root" test_247d() { lctl get_param -n mdc.$FSNAME-MDT0000*.import | grep -q subtree || - { skip "Fileset feature is not supported"; return; } + skip "Fileset feature is not supported" local submount=${MOUNT}_$tdir @@ -15058,7 +15945,7 @@ run_test 247d "running fid2path inside root" test_247e() { lctl get_param -n mdc.$FSNAME-MDT0000*.import | grep -q subtree || - { skip "Fileset feature is not supported"; return; } + skip "Fileset feature is not supported" local submount=${MOUNT}_$tdir @@ -15072,14 +15959,14 @@ run_test 247e "mount .. as fileset" test_248() { local fast_read_sav=$($LCTL get_param -n llite.*.fast_read 2>/dev/null) - [ -z "$fast_read_sav" ] && skip "no fast read support" && return + [ -z "$fast_read_sav" ] && skip "no fast read support" # create a large file for fast read verification dd if=/dev/zero of=$DIR/$tfile bs=1M count=128 > /dev/null 2>&1 # make sure the file is created correctly $CHECKSTAT -s $((128*1024*1024)) $DIR/$tfile || - { rm -f $DIR/$tfile; skip "file creation error" && return; } + { rm -f $DIR/$tfile; skip "file creation error"; } echo "Test 1: verify that fast read is 4 times faster on cache read" @@ -15124,12 +16011,11 @@ test_248() { run_test 248 "fast read verification" test_249() { # LU-7890 - rm -f $DIR/$tfile - $SETSTRIPE -c 1 $DIR/$tfile - [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.8.53) ] && - skip "Need at least version 2.8.54" + skip "Need at least version 2.8.54" + rm -f $DIR/$tfile + $SETSTRIPE -c 1 $DIR/$tfile # Offset 2T == 4k * 512M dd if=/dev/zero of=$DIR/$tfile bs=4k count=1 seek=512M || error "dd to 2T offset failed" @@ -15138,7 +16024,7 @@ run_test 249 "Write above 2T file size" test_250() { [ "$(facet_fstype ost$(($($GETSTRIPE -i $DIR/$tfile) + 1)))" = "zfs" ] \ - && skip "no 16TB file size limit on ZFS" && return + && skip "no 16TB file size limit on ZFS" $SETSTRIPE -c 1 $DIR/$tfile # ldiskfs extent file size limit is (16TB - 4KB - 1) bytes @@ -15168,6 +16054,13 @@ test_251() { run_test 251 "Handling short read and write correctly" test_252() { + remote_mds_nodsh && skip "remote MDS with nodsh" + remote_ost_nodsh && skip "remote OST with nodsh" + if [ "$(facet_fstype ost1)" != "ldiskfs" -o \ + "$(facet_fstype mds1)" != "ldiskfs" ]; then + skip_env "ldiskfs only test" + fi + local tgt local dev local out @@ -15175,14 +16068,6 @@ test_252() { local num local gen - remote_mds_nodsh && skip "remote MDS with nodsh" && return - remote_ost_nodsh && skip "remote OST with nodsh" && return - if [ "$(facet_fstype ost1)" != "ldiskfs" -o \ - "$(facet_fstype mds1)" != "ldiskfs" ]; then - skip "ldiskfs only test" - return - fi - # check lr_reader on OST0000 tgt=ost1 dev=$(facet_device $tgt) @@ -15198,8 +16083,7 @@ test_252() { tgt=mds1 dev=$(facet_device $tgt) if ! do_facet $tgt $LR_READER -h | grep -q OPTIONS; then - echo "$LR_READER does not support additional options" - return 0 + skip "$LR_READER does not support additional options" fi out=$(do_facet $tgt $LR_READER -c $dev) [ $? -eq 0 ] || error "$LR_READER failed on target $tgt device $dev" @@ -15248,21 +16132,20 @@ test_253_fill_ost() { } test_253() { + [ $PARALLEL == "yes" ] && skip "skip parallel run" + remote_mds_nodsh && skip "remote MDS with nodsh" + remote_mgs_nodsh && skip "remote MGS with nodsh" + local ostidx=0 local rc=0 - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return - remote_mds_nodsh && skip "remote MDS with nodsh" && return - remote_mgs_nodsh && skip "remote MGS with nodsh" && return - local ost_name=$($LFS osts | sed -n 's/^'$ostidx': \(.*\)_UUID .*/\1/p') # on the mdt's osc local mdtosc_proc1=$(get_mdtosc_proc_path $SINGLEMDS $ost_name) do_facet $SINGLEMDS $LCTL get_param -n \ osp.$mdtosc_proc1.reserved_mb_high || - { skip "remote MDS does not support reserved_mb_high" && - return; } + skip "remote MDS does not support reserved_mb_high" rm -rf $DIR/$tdir wait_mds_ost_sync @@ -15350,16 +16233,14 @@ test_253() { run_test 253 "Check object allocation limit" test_254() { - local cl_user - - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return - remote_mds_nodsh && skip "remote MDS with nodsh" && return + [ $PARALLEL == "yes" ] && skip "skip parallel run" + remote_mds_nodsh && skip "remote MDS with nodsh" + do_facet $SINGLEMDS $LCTL get_param -n mdd.$MDT0.changelog_size || + skip "MDS does not support changelog_size" + local cl_user local MDT0=$(facet_svc $SINGLEMDS) - do_facet $SINGLEMDS $LCTL get_param -n mdd.$MDT0.changelog_size || - { skip "MDS does not support changelog_size" && return; } - changelog_register || error "changelog_register failed" changelog_clear 0 || error "changelog_clear failed" @@ -15503,16 +16384,16 @@ ladvise_willread_performance() test_255a() { [ $(lustre_version_code ost1) -lt $(version_code 2.8.54) ] && - skip "lustre < 2.8.54 does not support ladvise " && return - remote_ost_nodsh && skip "remote OST with nodsh" && return + skip "lustre < 2.8.54 does not support ladvise " + remote_ost_nodsh && skip "remote OST with nodsh" lfs setstripe -c -1 -i 0 $DIR/$tfile || error "$tfile failed" ladvise_no_type willread $DIR/$tfile && - skip "willread ladvise is not supported" && return + skip "willread ladvise is not supported" ladvise_no_ioctl $DIR/$tfile && - skip "ladvise ioctl is not supported" && return + skip "ladvise ioctl is not supported" local size_mb=100 local size=$((size_mb * 1048576)) @@ -15590,23 +16471,20 @@ facet_meminfo() { test_255b() { [ $(lustre_version_code ost1) -lt $(version_code 2.8.54) ] && - skip "lustre < 2.8.54 does not support ladvise " && return - remote_ost_nodsh && skip "remote OST with nodsh" && return + skip "lustre < 2.8.54 does not support ladvise " + remote_ost_nodsh && skip "remote OST with nodsh" lfs setstripe -c 1 -i 0 $DIR/$tfile ladvise_no_type dontneed $DIR/$tfile && - skip "dontneed ladvise is not supported" && return + skip "dontneed ladvise is not supported" ladvise_no_ioctl $DIR/$tfile && - skip "ladvise ioctl is not supported" && return - - [ $(lustre_version_code ost1) -lt $(version_code 2.8.54) ] && - skip "lustre < 2.8.54 does not support ladvise" && return + skip "ladvise ioctl is not supported" ! $LFS ladvise -a dontneed $DIR/$tfile && [ "$(facet_fstype ost1)" = "zfs" ] && - skip "zfs-osd does not support 'ladvise dontneed'" && return + skip "zfs-osd does not support 'ladvise dontneed'" local size_mb=100 local size=$((size_mb * 1048576)) @@ -15617,6 +16495,9 @@ test_255b() { dd if=/dev/zero of=$DIR/$tfile bs=1048576 count=$size_mb || error "dd to $DIR/$tfile failed" + #force write to complete before dropping OST cache & checking memory + sync + local total=$(facet_meminfo ost1 MemTotal) echo "Total memory: $total KiB" @@ -15652,15 +16533,15 @@ test_255b() { run_test 255b "check 'lfs ladvise -a dontneed'" test_255c() { + [ $(lustre_version_code ost1) -lt $(version_code 2.10.50) ] && + skip "lustre < 2.10.53 does not support lockahead" + local count local new_count local difference local i local rc - [ $(lustre_version_code ost1) -lt $(version_code 2.10.50) ] && - skip "lustre < 2.10.53 does not support lockahead" && return - test_mkdir -p $DIR/$tdir $SETSTRIPE -i 0 $DIR/$tdir @@ -15736,19 +16617,19 @@ test_255c() { run_test 255c "suite of ladvise lockahead tests" test_256() { + [ $PARALLEL == "yes" ] && skip "skip parallel run" + remote_mds_nodsh && skip "remote MDS with nodsh" + [ "$(facet_fstype $SINGLEMDS)" != "ldiskfs" ] && + skip "ldiskfs only test" + changelog_users $SINGLEMDS | grep "^cl" && + skip "active changelog user" + local cl_user local cat_sl local mdt_dev - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return - remote_mds_nodsh && skip "remote MDS with nodsh" && return - [ "$(facet_fstype $SINGLEMDS)" != "ldiskfs" ] && - skip "ldiskfs only test" && return - mdt_dev=$(mdsdevname 1) echo $mdt_dev - changelog_users $SINGLEMDS | grep "^cl" && - skip "active changelog user" && return changelog_register || error "changelog_register failed" @@ -15792,9 +16673,9 @@ test_256() { run_test 256 "Check llog delete for empty and not full state" test_257() { - remote_mds_nodsh && skip "remote MDS with nodsh" && return + remote_mds_nodsh && skip "remote MDS with nodsh" [[ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.8.55) ]] && - skip "Need MDS version at least 2.8.55" && return + skip "Need MDS version at least 2.8.55" test_mkdir $DIR/$tdir @@ -15803,7 +16684,7 @@ test_257() { stat $DIR/$tdir #define OBD_FAIL_MDS_XATTR_REP 0x161 - local mdtidx=$($LFS getstripe -M $DIR/$tdir) + local mdtidx=$($LFS getstripe -m $DIR/$tdir) local facet=mds$((mdtidx + 1)) set_nodes_failloc $(facet_active_host $facet) 0x80000161 getfattr -n trusted.name1 $DIR/$tdir 2> /dev/null @@ -15847,6 +16728,45 @@ test_258b() { } run_test 258b "verify i_mutex security behavior" +test_259() { + local file=$DIR/$tfile + local before + local after + + [ "$(facet_fstype mds1)" != "ldiskfs" ] && + skip "ldiskfs only test" && return + + stack_trap "rm -f $file" EXIT + + wait_delete_completed + before=$(do_facet ost1 "$LCTL get_param -n osd-*.*OST0000.kbytesfree") + echo "before: $before" + + $LFS setstripe -i 0 -c 1 $file + dd if=/dev/zero of=$file bs=1M count=10 || error "couldn't write" + sync_all_data + after=$(do_facet ost1 "$LCTL get_param -n osd-*.*OST0000.kbytesfree") + echo "after write: $after" + +#define OBD_FAIL_OSD_FAIL_AT_TRUNCATE 0x2301 + do_facet ost1 $LCTL set_param fail_loc=0x2301 + $TRUNCATE $file 0 + after=$(do_facet ost1 "$LCTL get_param -n osd-*.*OST0000.kbytesfree") + echo "after truncate: $after" + + stop ost1 + do_facet ost1 $LCTL set_param fail_loc=0 + start ost1 $(ostdevname 1) $OST_MOUNT_OPTS || error "cannot start ost1" + sleep 2 + after=$(do_facet ost1 "$LCTL get_param -n osd-*.*OST0000.kbytesfree") + echo "after restart: $after" + [ $((after - before)) -ge $(fs_log_size ost1) ] && + error "missing truncate?" + + return 0 +} +run_test 259 "crash at delayed truncate" + test_260() { #define OBD_FAIL_MDC_CLOSE 0x806 $LCTL set_param fail_loc=0x80000806 @@ -15857,9 +16777,8 @@ run_test 260 "Check mdc_close fail" ### Data-on-MDT sanity tests ### test_270a() { - [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.10.55) ] && - skip "Need MDS version at least 2.10.55" && return + skip "Need MDS version at least 2.10.55" # create DoM file local dom=$DIR/$tdir/dom_file @@ -15881,7 +16800,7 @@ test_270a() { [ $($LFS getstripe -c $dom) == 0 ] || error "bad stripe count" [ $($LFS getstripe -S $dom) == 1048576 ] || error "bad stripe size" - local mdtidx=$($GETSTRIPE -M $dom) + local mdtidx=$($LFS getstripe -m $dom) local mdtname=MDT$(printf %04x $mdtidx) local facet=mds$((mdtidx + 1)) local space_check=1 @@ -15957,7 +16876,7 @@ run_test 270a "DoM: basic functionality tests" test_270b() { [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.10.55) ] && - skip "Need MDS version at least 2.10.55" && return + skip "Need MDS version at least 2.10.55" local dom=$DIR/$tdir/dom_file local max_size=1048576 @@ -15982,7 +16901,7 @@ run_test 270b "DoM: maximum size overflow checks for DoM-only file" test_270c() { [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.10.55) ] && - skip "Need MDS version at least 2.10.55" && return + skip "Need MDS version at least 2.10.55" mkdir -p $DIR/$tdir $LFS setstripe -E 1024K -L mdt $DIR/$tdir @@ -16011,7 +16930,7 @@ run_test 270c "DoM: DoM EA inheritance tests" test_270d() { [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.10.55) ] && - skip "Need MDS version at least 2.10.55" && return + skip "Need MDS version at least 2.10.55" mkdir -p $DIR/$tdir $LFS setstripe -E 1024K -L mdt $DIR/$tdir @@ -16033,7 +16952,7 @@ run_test 270d "DoM: change striping from DoM to RAID0" test_270e() { [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.10.55) ] && - skip "Need MDS version at least 2.10.55" && return + skip "Need MDS version at least 2.10.55" mkdir -p $DIR/$tdir/dom mkdir -p $DIR/$tdir/norm @@ -16074,7 +16993,7 @@ run_test 270e "DoM: lfs find with DoM files test" test_270f() { [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.10.55) ] && - skip "Need MDS version at least 2.10.55" && return + skip "Need MDS version at least 2.10.55" local mdtname=${FSNAME}-MDT0000-mdtlov local dom=$DIR/$tdir/dom_file @@ -16094,7 +17013,9 @@ test_270f() { error "Can't set directory default striping" # exceed maximum stripe size - $LFS setstripe -E $(($dom_limit * 2)) -L mdt $dom && + $LFS setstripe -E $((dom_limit * 2)) -L mdt $dom || + error "Can't create file with $((dom_limit * 2)) DoM stripe" + [ $($LFS getstripe -S $dom) -eq $((dom_limit * 2)) ] && error "Able to create DoM component size more than LOD limit" do_facet mds1 $LCTL set_param -n lod.$mdtname.dom_stripesize=0 @@ -16102,6 +17023,19 @@ test_270f() { lod.$mdtname.dom_stripesize) [ 0 -eq ${dom_current} ] || error "Can't set zero DoM stripe limit" + rm $dom + + # attempt to create DoM file on server with disabled DoM should + # remove DoM entry from layout and be succeed + $LFS setstripe -E $dom_limit -L mdt -E -1 $dom || + error "Can't create DoM file (DoM is disabled)" + [ $($LFS getstripe -L $dom) == "mdt" ] && + error "File has DoM component while DoM is disabled" + rm $dom + + # attempt to create DoM file with only DoM stripe should return error + $LFS setstripe -E $dom_limit -L mdt $dom && + error "Able to create DoM-only file while DoM is disabled" # too low values to be aligned with smallest stripe size 64K do_facet mds1 $LCTL set_param -n lod.$mdtname.dom_stripesize=30000 @@ -16110,6 +17044,10 @@ test_270f() { [ 30000 -eq ${dom_current} ] && error "Can set too small DoM stripe limit" + # 64K is a minimal stripe size in Lustre, expect limit of that size + [ 65536 -eq ${dom_current} ] || + error "Limit is not set to 64K but ${dom_current}" + do_facet mds1 $LCTL set_param -n lod.$mdtname.dom_stripesize=2147483648 dom_current=$(do_facet mds1 $LCTL get_param -n \ lod.$mdtname.dom_stripesize) @@ -16123,7 +17061,9 @@ test_270f() { error "Can't create DoM component size after limit change" do_facet mds1 $LCTL set_param -n \ lod.$mdtname.dom_stripesize=$((dom_limit / 2)) - $LFS setstripe -E $dom_limit -L mdt ${dom}_big && + $LFS setstripe -E $dom_limit -L mdt ${dom}_big || + error "Can't create DoM file after limit decrease" + [ $($LFS getstripe -S ${dom}_big) -eq $((dom_limit / 2)) ] || error "Can create big DoM component after limit decrease" touch ${dom}_def || error "Can't create file with old default layout" @@ -16135,7 +17075,7 @@ run_test 270f "DoM: maximum DoM stripe size checks" test_271a() { [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.10.55) ] && - skip "Need MDS version at least 2.10.55" && return + skip "Need MDS version at least 2.10.55" local dom=$DIR/$tdir/dom @@ -16146,9 +17086,8 @@ test_271a() { lctl set_param -n mdc.*.stats=clear dd if=/dev/zero of=$dom bs=4096 count=1 || return 1 cat $dom > /dev/null - local reads=$(lctl get_param -n mdc.*.stats | - awk '/ost_read/ {print $2}') - [ -z $reads ] || error "Unexpected $reads READ RPCs" + local reads=$(lctl get_param -n mdc.*.stats | grep -c ost_read) + [ $reads -eq 0 ] || error "Unexpected $reads READ RPCs" ls $dom rm -f $dom } @@ -16156,7 +17095,7 @@ run_test 271a "DoM: data is cached for read after write" test_271b() { [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.10.55) ] && - skip "Need MDS version at least 2.10.55" && return + skip "Need MDS version at least 2.10.55" local dom=$DIR/$tdir/dom @@ -16170,16 +17109,15 @@ test_271b() { $CHECKSTAT -t file -s 4096 $dom || error "stat #1 fails" # second stat to check size is cached on client $CHECKSTAT -t file -s 4096 $dom || error "stat #2 fails" - local gls=$(lctl get_param -n mdc.*.stats | - awk '/ldlm_glimpse/ {print $2}') - [ -z $gls ] || error "Unexpected $gls glimpse RPCs" + local gls=$(lctl get_param -n mdc.*.stats | grep -c ldlm_glimpse) + [ $gls -eq 0 ] || error "Unexpected $gls glimpse RPCs" rm -f $dom } run_test 271b "DoM: no glimpse RPC for stat (DoM only file)" test_271ba() { [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.10.55) ] && - skip "Need MDS version at least 2.10.55" && return + skip "Need MDS version at least 2.10.55" local dom=$DIR/$tdir/dom @@ -16194,22 +17132,30 @@ test_271ba() { $CHECKSTAT -t file -s 2097152 $dom || error "stat" # second stat to check size is cached on client $CHECKSTAT -t file -s 2097152 $dom || error "stat" - local gls=$(lctl get_param -n mdc.*.stats | - awk '/ldlm_glimpse/ {print $2}') - [ -z $gls ] || error "Unexpected $gls glimpse RPCs" - local gls=$(lctl get_param -n osc.*.stats | - awk '/ldlm_glimpse/ {print $2}') - [ -z $gls ] || error "Unexpected $gls OSC glimpse RPCs" + local gls=$(lctl get_param -n mdc.*.stats | grep -c ldlm_glimpse) + [ $gls == 0 ] || error "Unexpected $gls glimpse RPCs" + local gls=$(lctl get_param -n osc.*.stats | grep -c ldlm_glimpse) + [ $gls == 0 ] || error "Unexpected $gls OSC glimpse RPCs" rm -f $dom } run_test 271ba "DoM: no glimpse RPC for stat (combined file)" -test_271c() { - # test to be enabled with lock_convert - skip "skipped until lock convert will be implemented" && return +get_mdc_stats() { + local mdtidx=$1 + local param=$2 + local mdt=MDT$(printf %04x $mdtidx) + + if [ -z $param ]; then + lctl get_param -n mdc.*$mdt*.stats + else + lctl get_param -n mdc.*$mdt*.stats | awk "/$param/"'{print $2}' + fi +} + +test_271c() { [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.10.55) ] && - skip "Need MDS version at least 2.10.55" && return + skip "Need MDS version at least 2.10.55" local dom=$DIR/$tdir/dom @@ -16217,7 +17163,7 @@ test_271c() { $LFS setstripe -E 1024K -L mdt $DIR/$tdir - local mdtidx=$($LFS getstripe -M $DIR/$tdir) + local mdtidx=$($LFS getstripe -m $DIR/$tdir) local facet=mds$((mdtidx + 1)) cancel_lru_locks mdc @@ -16225,9 +17171,8 @@ test_271c() { createmany -o $dom 1000 lctl set_param -n mdc.*.stats=clear smalliomany -w $dom 1000 200 - lctl get_param -n mdc.*.stats - local enq=$(lctl get_param -n mdc.*.stats | - awk '/ldlm_ibits_enqueue/ {print $2}') + get_mdc_stats $mdtidx + local enq=$(get_mdc_stats $mdtidx ldlm_ibits_enqueue) # Each file has 1 open, 1 IO enqueues, total 2000 # but now we have also +1 getxattr for security.capability, total 3000 [ $enq -ge 2000 ] || error "Too few enqueues $enq, expected > 2000" @@ -16238,9 +17183,7 @@ test_271c() { createmany -o $dom 1000 lctl set_param -n mdc.*.stats=clear smalliomany -w $dom 1000 200 - lctl get_param -n mdc.*.stats - local enq_2=$(lctl get_param -n mdc.*.stats | - awk '/ldlm_ibits_enqueue/ {print $2}') + local enq_2=$(get_mdc_stats $mdtidx ldlm_ibits_enqueue) # Expect to see reduced amount of RPCs by 1000 due to single enqueue # for OPEN and IO lock. [ $((enq - enq_2)) -ge 1000 ] || @@ -16250,10 +17193,314 @@ test_271c() { } run_test 271c "DoM: IO lock at open saves enqueue RPCs" +cleanup_271def_tests() { + trap 0 + rm -f $1 +} + +test_271d() { + [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.10.57) ] && + skip "Need MDS version at least 2.10.57" && return + + local dom=$DIR/$tdir/dom + local tmp=$TMP/$tfile + trap "cleanup_271def_tests $tmp" EXIT + + mkdir -p $DIR/$tdir + + $LFS setstripe -E 1024K -L mdt $DIR/$tdir + + local mdtidx=$($GETSTRIPE -M $DIR/$tdir) + + cancel_lru_locks mdc + dd if=/dev/urandom of=$tmp bs=1000 count=1 + dd if=$tmp of=$dom bs=1000 count=1 + cancel_lru_locks mdc + + cat /etc/hosts >> $tmp + lctl set_param -n mdc.*.stats=clear + + # append data to the same file it should update local page + echo "Append to the same page" + cat /etc/hosts >> $dom + local num=$(get_mdc_stats $mdtidx ost_read) + local ra=$(get_mdc_stats $mdtidx req_active) + local rw=$(get_mdc_stats $mdtidx req_waittime) + + [ -z $num ] || error "$num READ RPC occured" + [ $ra == $rw ] || error "$((ra - rw)) resend occured" + echo "... DONE" + + # compare content + cmp $tmp $dom || error "file miscompare" + + cancel_lru_locks mdc + lctl set_param -n mdc.*.stats=clear + + echo "Open and read file" + cat $dom > /dev/null + local num=$(get_mdc_stats $mdtidx ost_read) + local ra=$(get_mdc_stats $mdtidx req_active) + local rw=$(get_mdc_stats $mdtidx req_waittime) + + [ -z $num ] || error "$num READ RPC occured" + [ $ra == $rw ] || error "$((ra - rw)) resend occured" + echo "... DONE" + + # compare content + cmp $tmp $dom || error "file miscompare" + + return 0 +} +run_test 271d "DoM: read on open (1K file in reply buffer)" + +test_271e() { + [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.10.57) ] && + skip "Need MDS version at least 2.10.57" && return + + local dom=$DIR/$tdir/dom + local tmp=$TMP/${tfile}.data + trap "cleanup_271def_tests $tmp" EXIT + + mkdir -p $DIR/$tdir + + $LFS setstripe -E 1024K -L mdt $DIR/$tdir + + local mdtidx=$($GETSTRIPE -M $DIR/$tdir) + + cancel_lru_locks mdc + dd if=/dev/urandom of=$tmp bs=30K count=1 + dd if=$tmp of=$dom bs=30K count=1 + cancel_lru_locks mdc + cat /etc/hosts >> $tmp + lctl set_param -n mdc.*.stats=clear + + echo "Append to the same page" + cat /etc/hosts >> $dom + + local num=$(get_mdc_stats $mdtidx ost_read) + local ra=$(get_mdc_stats $mdtidx req_active) + local rw=$(get_mdc_stats $mdtidx req_waittime) + + [ -z $num ] || error "$num READ RPC occured" + # Reply buffer can be adjusted for larger buffer by resend + echo "... DONE with $((ra - rw)) resends" + + # compare content + cmp $tmp $dom || error "file miscompare" + + cancel_lru_locks mdc + lctl set_param -n mdc.*.stats=clear + + echo "Open and read file" + cat $dom > /dev/null + local num=$(get_mdc_stats $mdtidx ost_read) + local ra=$(get_mdc_stats $mdtidx req_active) + local rw=$(get_mdc_stats $mdtidx req_waittime) + + [ -z $num ] || error "$num READ RPC occured" + # Reply buffer can be adjusted for larger buffer by resend + echo "... DONE with $((ra - rw)) resends" + + # compare content + cmp $tmp $dom || error "file miscompare" + + return 0 +} +run_test 271e "DoM: read on open (30K file with reply buffer adjusting)" + +test_271f() { + [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.10.57) ] && + skip "Need MDS version at least 2.10.57" && return + + local dom=$DIR/$tdir/dom + local tmp=$TMP/$tfile + trap "cleanup_271def_tests $tmp" EXIT + + mkdir -p $DIR/$tdir + + $LFS setstripe -E 1024K -L mdt $DIR/$tdir + + local mdtidx=$($GETSTRIPE -M $DIR/$tdir) + + cancel_lru_locks mdc + dd if=/dev/urandom of=$tmp bs=200000 count=1 + dd if=$tmp of=$dom bs=200000 count=1 + cancel_lru_locks mdc + cat /etc/hosts >> $tmp + lctl set_param -n mdc.*.stats=clear + + echo "Append to the same page" + cat /etc/hosts >> $dom + local num=$(get_mdc_stats $mdtidx ost_read) + local ra=$(get_mdc_stats $mdtidx req_active) + local rw=$(get_mdc_stats $mdtidx req_waittime) + + [ -z $num ] || error "$num READ RPC occured" + [ $ra == $rw ] || error "$((ra - rw)) resend occured" + echo "... DONE" + + # compare content + cmp $tmp $dom || error "file miscompare" + + cancel_lru_locks mdc + lctl set_param -n mdc.*.stats=clear + + echo "Open and read file" + cat $dom > /dev/null + local num=$(get_mdc_stats $mdtidx ost_read) + local ra=$(get_mdc_stats $mdtidx req_active) + local rw=$(get_mdc_stats $mdtidx req_waittime) + + [ $num -eq 1 ] || error "expect 1 READ RPC, $num occured" + [ $ra == $rw ] || error "$((ra - rw)) resend occured" + echo "... DONE" + + # compare content + cmp $tmp $dom || error "file miscompare" + + return 0 +} +run_test 271f "DoM: read on open (200K file and read tail)" + +test_272a() { + [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.11.50) ] && + skip "Need MDS version at least 2.11.50" + + local dom=$DIR/$tdir/dom + mkdir -p $DIR/$tdir + + $LFS setstripe -E 256K -L mdt -E -1 -c1 $dom + dd if=/dev/urandom of=$dom bs=512K count=1 || + error "failed to write data into $dom" + local old_md5=$(md5sum $dom) + + $LFS migrate -E 256K -L mdt -E -1 -c2 $dom || + error "failed to migrate to the same DoM component" + + [ $($LFS getstripe -c $dom) -eq 2 ] || + error "layout was not changed silently" + + local new_md5=$(md5sum $dom) + + [ "$old_md5" != "$new_md5" ] && + error "md5sum differ: $old_md5, $new_md5" + return 0 +} +run_test 272a "DoM migration: new layout with the same DOM component" + +test_272b() { + [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.11.50) ] && + skip "Need MDS version at least 2.11.50" + + local dom=$DIR/$tdir/dom + mkdir -p $DIR/$tdir + $LFS setstripe -E 1M -L mdt -E -1 -c1 $dom + + local mdtidx=$($LFS getstripe -m $dom) + local mdtname=MDT$(printf %04x $mdtidx) + local facet=mds$((mdtidx + 1)) + + local mdtfree1=$(do_facet $facet \ + lctl get_param -n osd*.*$mdtname.kbytesfree) + dd if=/dev/urandom of=$dom bs=2M count=1 || + error "failed to write data into $dom" + local old_md5=$(md5sum $dom) + cancel_lru_locks mdc + local mdtfree1=$(do_facet $facet \ + lctl get_param -n osd*.*$mdtname.kbytesfree) + + $LFS migrate -c2 $dom || + error "failed to migrate to the new composite layout" + [ $($LFS getstripe -L $dom) == 'mdt' ] && + error "MDT stripe was not removed" + + cancel_lru_locks mdc + local new_md5=$(md5sum $dom) + [ "$old_md5" != "$new_md5" ] && + error "$old_md5 != $new_md5" + + # Skip free space checks with ZFS + if [ "$(facet_fstype $facet)" != "zfs" ]; then + local mdtfree2=$(do_facet $facet \ + lctl get_param -n osd*.*$mdtname.kbytesfree) + [ $mdtfree2 -gt $mdtfree1 ] || + error "MDT space is not freed after migration" + fi + return 0 +} +run_test 272b "DoM migration: DOM file to the OST-striped file (plain)" + +test_272c() { + [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.11.50) ] && + skip "Need MDS version at least 2.11.50" + + local dom=$DIR/$tdir/$tfile + mkdir -p $DIR/$tdir + $LFS setstripe -E 1M -L mdt -E -1 -c1 $dom + + local mdtidx=$($LFS getstripe -m $dom) + local mdtname=MDT$(printf %04x $mdtidx) + local facet=mds$((mdtidx + 1)) + + dd if=/dev/urandom of=$dom bs=2M count=1 oflag=direct || + error "failed to write data into $dom" + local old_md5=$(md5sum $dom) + cancel_lru_locks mdc + local mdtfree1=$(do_facet $facet \ + lctl get_param -n osd*.*$mdtname.kbytesfree) + + $LFS migrate -E 2M -c1 -E -1 -c2 $dom || + error "failed to migrate to the new composite layout" + [ $($LFS getstripe -L $dom) == 'mdt' ] && + error "MDT stripe was not removed" + + cancel_lru_locks mdc + local new_md5=$(md5sum $dom) + [ "$old_md5" != "$new_md5" ] && + error "$old_md5 != $new_md5" + + # Skip free space checks with ZFS + if [ "$(facet_fstype $facet)" != "zfs" ]; then + local mdtfree2=$(do_facet $facet \ + lctl get_param -n osd*.*$mdtname.kbytesfree) + [ $mdtfree2 -gt $mdtfree1 ] || + error "MDS space is not freed after migration" + fi + return 0 +} +run_test 272c "DoM migration: DOM file to the OST-striped file (composite)" + +test_273a() { + [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.11.50) ] && + skip "Need MDS version at least 2.11.50" + + # Layout swap cannot be done if either file has DOM component, + # this will never be supported, migration should be used instead + + local dom=$DIR/$tdir/$tfile + mkdir -p $DIR/$tdir + + $LFS setstripe -c2 ${dom}_plain + $LFS setstripe -E 1M -L mdt -E -1 -c2 ${dom}_dom + $LFS swap_layouts ${dom}_plain ${dom}_dom && + error "can swap layout with DoM component" + $LFS swap_layouts ${dom}_dom ${dom}_plain && + error "can swap layout with DoM component" + + $LFS setstripe -E 1M -c1 -E -1 -c2 ${dom}_comp + $LFS swap_layouts ${dom}_comp ${dom}_dom && + error "can swap layout with DoM component" + $LFS swap_layouts ${dom}_dom ${dom}_comp && + error "can swap layout with DoM component" + return 0 +} +run_test 273a "DoM: layout swapping should fail with DOM" + test_275() { - remote_ost_nodsh && skip "remote OST with nodsh" && return + remote_ost_nodsh && skip "remote OST with nodsh" [ $(lustre_version_code ost1) -lt $(version_code 2.10.57) ] && - skip "Need OST version >= 2.10.57" && return 0 + skip "Need OST version >= 2.10.57" local file=$DIR/$tfile local oss @@ -16284,7 +17531,7 @@ test_275() { run_test 275 "Read on a canceled duplicate lock" test_276() { - remote_ost_nodsh && skip "remote OST with nodsh" && return + remote_ost_nodsh && skip "remote OST with nodsh" local pid do_facet ost1 "(while true; do \ @@ -16376,9 +17623,9 @@ test_striped_dir() { test_300a() { [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.7.0) ] && - skip "skipped for lustre < 2.7.0" && return - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return - [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return + skip "skipped for lustre < 2.7.0" + [ $PARALLEL == "yes" ] && skip "skip parallel run" + [ $MDSCOUNT -lt 2 ] && skip_env "needs >= 2 MDTs" test_striped_dir 0 || error "failed on striped dir on MDT0" test_striped_dir 1 || error "failed on striped dir on MDT0" @@ -16387,9 +17634,10 @@ run_test 300a "basic striped dir sanity test" test_300b() { [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.7.0) ] && - skip "skipped for lustre < 2.7.0" && return - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return - [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return + skip "skipped for lustre < 2.7.0" + [ $PARALLEL == "yes" ] && skip "skip parallel run" + [ $MDSCOUNT -lt 2 ] && skip_env "needs >= 2 MDTs" + local i local mtime1 local mtime2 @@ -16415,9 +17663,10 @@ run_test 300b "check ctime/mtime for striped dir" test_300c() { [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.7.0) ] && - skip "skipped for lustre < 2.7.0" && return - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return - [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return + skip "skipped for lustre < 2.7.0" + [ $PARALLEL == "yes" ] && skip "skip parallel run" + [ $MDSCOUNT -lt 2 ] && skip_env "needs >= 2 MDTs" + local file_count mkdir -p $DIR/$tdir @@ -16440,9 +17689,10 @@ run_test 300c "chown && check ls under striped directory" test_300d() { [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.7.0) ] && - skip "skipped for lustre < 2.7.0" && return - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return - [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return + skip "skipped for lustre < 2.7.0" + [ $PARALLEL == "yes" ] && skip "skip parallel run" + [ $MDSCOUNT -lt 2 ] && skip_env "needs >= 2 MDTs" + local stripe_count local file @@ -16462,7 +17712,7 @@ test_300d() { error "create 10 files failed" for file in $(find $DIR/$tdir); do - stripe_count=$($GETSTRIPE -c $file) + stripe_count=$($LFS getstripe -c $file) [ $stripe_count -eq 2 ] || error "wrong stripe $stripe_count for $file" done @@ -16473,9 +17723,10 @@ run_test 300d "check default stripe under striped directory" test_300e() { [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.7.55) ] && - skip "Need MDS version at least 2.7.55" && return - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return - [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return + skip "Need MDS version at least 2.7.55" + [ $PARALLEL == "yes" ] && skip "skip parallel run" + [ $MDSCOUNT -lt 2 ] && skip_env "needs >= 2 MDTs" + local stripe_count local file @@ -16518,10 +17769,11 @@ test_300e() { run_test 300e "check rename under striped directory" test_300f() { + [ $PARALLEL == "yes" ] && skip "skip parallel run" + [ $MDSCOUNT -lt 2 ] && skip_env "needs >= 2 MDTs" [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.7.55) ] && - skip "Need MDS version at least 2.7.55" && return - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return - [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return + skip "Need MDS version at least 2.7.55" + local stripe_count local file @@ -16607,9 +17859,10 @@ test_300_check_default_striped_dir() } test_300g() { + [ $MDSCOUNT -lt 2 ] && skip_env "needs >= 2 MDTs" [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.7.55) ] && - skip "Need MDS version at least 2.7.55" && return - [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return + skip "Need MDS version at least 2.7.55" + local dir local stripe_count local stripe_index @@ -16658,9 +17911,10 @@ test_300g() { run_test 300g "check default striped directory for normal directory" test_300h() { + [ $MDSCOUNT -lt 2 ] && skip_env "needs >= 2 MDTs" [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.7.55) ] && - skip "Need MDS version at least 2.7.55" && return - [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return + skip "Need MDS version at least 2.7.55" + local dir local stripe_count @@ -16687,10 +17941,11 @@ test_300h() { run_test 300h "check default striped directory for striped directory" test_300i() { + [ $PARALLEL == "yes" ] && skip "skip parallel run" + [ $MDSCOUNT -lt 2 ] && skip_env "needs >= 2 MDTs" [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.7.55) ] && - skip "Need MDS version at least 2.7.55" && return - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return - [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return + skip "Need MDS version at least 2.7.55" + local stripe_count local file @@ -16741,10 +17996,11 @@ test_300i() { run_test 300i "client handle unknown hash type striped directory" test_300j() { - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return + [ $MDSCOUNT -lt 2 ] && skip_env "needs >= 2 MDTs" + [ $PARALLEL == "yes" ] && skip "skip parallel run" [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.7.55) ] && - skip "Need MDS version at least 2.7.55" && return - [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return + skip "Need MDS version at least 2.7.55" + local stripe_count local file @@ -16767,10 +18023,11 @@ test_300j() { run_test 300j "test large update record" test_300k() { + [ $PARALLEL == "yes" ] && skip "skip parallel run" + [ $MDSCOUNT -lt 2 ] && skip_env "needs >= 2 MDTs" [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.7.55) ] && - skip "Need MDS version at least 2.7.55" && return - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return - [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return + skip "Need MDS version at least 2.7.55" + local stripe_count local file @@ -16778,7 +18035,7 @@ test_300k() { #define OBD_FAIL_LARGE_STRIPE 0x1703 $LCTL set_param fail_loc=0x1703 - $LFS setdirstripe -i 0 -c512 $DIR/$tdir/striped_dir || + $LFS setdirstripe -i 0 -c192 $DIR/$tdir/striped_dir || error "set striped dir error" $LCTL set_param fail_loc=0 @@ -16792,10 +18049,11 @@ test_300k() { run_test 300k "test large striped directory" test_300l() { + [ $PARALLEL == "yes" ] && skip "skip parallel run" + [ $MDSCOUNT -lt 2 ] && skip_env "needs >= 2 MDTs" [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.7.55) ] && - skip "Need MDS version at least 2.7.55" && return - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return - [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return + skip "Need MDS version at least 2.7.55" + local stripe_index test_mkdir -p $DIR/$tdir/striped_dir @@ -16815,10 +18073,10 @@ test_300l() { run_test 300l "non-root user to create dir under striped dir with stale layout" test_300m() { + [ $PARALLEL == "yes" ] && skip "skip parallel run" + [ $MDSCOUNT -ge 2 ] && skip_env "Only for single MDT" [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.7.55) ] && - skip "Need MDS version at least 2.7.55" && return - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return - [ $MDSCOUNT -ge 2 ] && skip "Only for single MDT" && return + skip "Need MDS version at least 2.7.55" mkdir -p $DIR/$tdir/striped_dir $LFS setdirstripe -D -c 1 $DIR/$tdir/striped_dir || @@ -16857,11 +18115,11 @@ cleanup_300n() { } test_300n() { + [ $PARALLEL == "yes" ] && skip "skip parallel run" + [ $MDSCOUNT -lt 2 ] && skip_env "needs >= 2 MDTs" [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.7.55) ] && - skip "Need MDS version at least 2.7.55" && return - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return - [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return - remote_mds_nodsh && skip "remote MDS with nodsh" && return + skip "Need MDS version at least 2.7.55" + remote_mds_nodsh && skip "remote MDS with nodsh" local stripe_index local list=$(comma_list $(mdts_nodes)) @@ -16900,10 +18158,11 @@ test_300n() { run_test 300n "non-root user to create dir under striped dir with default EA" test_300o() { + [ $PARALLEL == "yes" ] && skip "skip parallel run" + [ $MDSCOUNT -lt 2 ] && skip_env "needs >= 2 MDTs" [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.7.55) ] && - skip "Need MDS version at least 2.7.55" && return - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return - [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return + skip "Need MDS version at least 2.7.55" + local numfree1 local numfree2 @@ -16913,14 +18172,12 @@ test_300o() { numfree2=$(lctl get_param -n mdc.*MDT0001*.filesfree) if [ $numfree1 -lt 66000 -o $numfree2 -lt 66000 ]; then skip "not enough free inodes $numfree1 $numfree2" - return fi numfree1=$(lctl get_param -n mdc.*MDT0000-mdc-*.kbytesfree) numfree2=$(lctl get_param -n mdc.*MDT0001-mdc-*.kbytesfree) if [ $numfree1 -lt 300000 -o $numfree2 -lt 300000 ]; then skip "not enough free space $numfree1 $numfree2" - return fi $LFS setdirstripe -c2 $DIR/$tdir/striped_dir || @@ -16938,9 +18195,9 @@ test_300o() { run_test 300o "unlink big sub stripe(> 65000 subdirs)" test_300p() { - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return - [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return - remote_mds_nodsh && skip "remote MDS with nodsh" && return + [ $PARALLEL == "yes" ] && skip "skip parallel run" + [ $MDSCOUNT -lt 2 ] && skip_env "needs >= 2 MDTs" + remote_mds_nodsh && skip "remote MDS with nodsh" mkdir -p $DIR/$tdir @@ -16957,8 +18214,8 @@ test_300p() { run_test 300p "create striped directory without space" test_300q() { - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return - [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return + [ $PARALLEL == "yes" ] && skip "skip parallel run" + [ $MDSCOUNT -lt 2 ] && skip_env "needs >= 2 MDTs" local fd=$(free_fd) local cmd="exec $fd<$tdir" @@ -17003,8 +18260,9 @@ prepare_remote_file() { } test_310a() { - [[ $MDSCOUNT -lt 2 ]] && skip "needs >= 4 MDTs" && return - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return + [[ $MDSCOUNT -lt 2 ]] && skip_env "needs >= 4 MDTs" + [ $PARALLEL == "yes" ] && skip "skip parallel run" + local remote_file=$DIR/$tdir/tgt_dir/b mkdir -p $DIR/$tdir @@ -17019,8 +18277,9 @@ test_310a() { run_test 310a "open unlink remote file" test_310b() { - [[ $MDSCOUNT -lt 2 ]] && skip "needs >= 4 MDTs" && return - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return + [[ $MDSCOUNT -lt 2 ]] && skip_env "needs >= 4 MDTs" + [ $PARALLEL == "yes" ] && skip "skip parallel run" + local remote_file=$DIR/$tdir/tgt_dir/b mkdir -p $DIR/$tdir @@ -17034,8 +18293,9 @@ test_310b() { run_test 310b "unlink remote file with multiple links while open" test_310c() { - [[ $MDSCOUNT -lt 4 ]] && skip "needs >= 4 MDTs" && return - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return + [ $PARALLEL == "yes" ] && skip "skip parallel run" + [[ $MDSCOUNT -lt 4 ]] && skip_env "needs >= 4 MDTs" + local remote_file=$DIR/$tdir/tgt_dir/b mkdir -p $DIR/$tdir @@ -17054,10 +18314,11 @@ run_test 310c "open-unlink remote file with multiple links" #LU-4825 test_311() { + [ $PARALLEL == "yes" ] && skip "skip parallel run" + [ $OSTCOUNT -lt 2 ] && skip "needs >= 2 OSTs" [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.8.54) ] && - skip "lustre < 2.8.54 does not contain LU-4825 fix" && return - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return - remote_mds_nodsh && skip "remote MDS with nodsh" && return + skip "lustre < 2.8.54 does not contain LU-4825 fix" + remote_mds_nodsh && skip "remote MDS with nodsh" local old_iused=$($LFS df -i | grep OST0000 | awk '{ print $3 }') @@ -17140,15 +18401,13 @@ zfs_object_blksz() { } test_312() { # LU-4856 - remote_ost_nodsh && skip "remote OST with nodsh" && return - + remote_ost_nodsh && skip "remote OST with nodsh" [ $(facet_fstype ost1) = "zfs" ] || - { skip "the test only applies to zfs" && return; } + skip_env "the test only applies to zfs" local max_blksz=$(do_facet ost1 \ $ZFS get -p recordsize $(facet_device ost1) | awk '!/VALUE/{print $3}') - local min_blksz=$(getconf PAGE_SIZE) # to make life a little bit easier $LFS mkdir -c 1 -i 0 $DIR/$tdir @@ -17160,30 +18419,31 @@ test_312() { # LU-4856 # Get ZFS object id local zfs_objid=$(zfs_oid_to_objid ost1 $oid) + # block size change by sequential overwrite + local bs - # block size change by sequential over write - local blksz - for ((bs=$min_blksz; bs <= max_blksz; bs <<= 2)); do + for ((bs=$PAGE_SIZE; bs <= max_blksz; bs *= 4)) ; do dd if=/dev/zero of=$tf bs=$bs count=1 oflag=sync conv=notrunc - blksz=$(zfs_object_blksz ost1 $zfs_objid) + local blksz=$(zfs_object_blksz ost1 $zfs_objid) [ $blksz -eq $bs ] || error "blksz error: $blksz, expected: $bs" done rm -f $tf # block size change by sequential append write - dd if=/dev/zero of=$tf bs=$min_blksz count=1 oflag=sync conv=notrunc + dd if=/dev/zero of=$tf bs=$PAGE_SIZE count=1 oflag=sync conv=notrunc oid=$($LFS getstripe $tf | awk '/obdidx/{getline; print $2}') zfs_objid=$(zfs_oid_to_objid ost1 $oid) + local count - for ((count = 1; count < $((max_blksz / min_blksz)); count *= 2)); do - dd if=/dev/zero of=$tf bs=$min_blksz count=$count seek=$count \ + for ((count = 1; count < $((max_blksz / PAGE_SIZE)); count *= 2)); do + dd if=/dev/zero of=$tf bs=$PAGE_SIZE count=$count seek=$count \ oflag=sync conv=notrunc blksz=$(zfs_object_blksz ost1 $zfs_objid) - [ $blksz -eq $((2 * count * min_blksz)) ] || - error "blksz error, actual $blksz, " \ - "expected: 2 * $count * $min_blksz" + [ $blksz -eq $((2 * count * PAGE_SIZE)) ] || + error "blksz error, actual $blksz, " \ + "expected: 2 * $count * $PAGE_SIZE" done rm -f $tf @@ -17194,8 +18454,8 @@ test_312() { # LU-4856 dd if=/dev/zero of=$tf bs=1K count=1 oflag=sync conv=notrunc blksz=$(zfs_object_blksz ost1 $zfs_objid) - [ $blksz -eq $min_blksz ] || - error "blksz error: $blksz, expected: $min_blksz" + [ $blksz -eq $PAGE_SIZE ] || + error "blksz error: $blksz, expected: $PAGE_SIZE" dd if=/dev/zero of=$tf bs=64K count=1 oflag=sync conv=notrunc seek=128 blksz=$(zfs_object_blksz ost1 $zfs_objid) @@ -17208,15 +18468,16 @@ test_312() { # LU-4856 run_test 312 "make sure ZFS adjusts its block size by write pattern" test_313() { - remote_ost_nodsh && skip "remote OST with nodsh" && return + remote_ost_nodsh && skip "remote OST with nodsh" local file=$DIR/$tfile + rm -f $file $SETSTRIPE -c 1 -i 0 $file || error "setstripe failed" # define OBD_FAIL_TGT_RCVD_EIO 0x720 do_facet ost1 "$LCTL set_param fail_loc=0x720" - dd if=/dev/zero of=$file bs=4096 oflag=direct count=1 && + dd if=/dev/zero of=$file bs=$PAGE_SIZE oflag=direct count=1 && error "write should failed" do_facet ost1 "$LCTL set_param fail_loc=0" rm -f $file @@ -17224,6 +18485,8 @@ test_313() { run_test 313 "io should fail after last_rcvd update fail" test_314() { + [ $OSTCOUNT -lt 2 ] && skip "needs >= 2 OSTs" + $SETSTRIPE -c 2 -i 0 $DIR/$tfile || error "setstripe failed" do_facet ost1 "$LCTL set_param fail_loc=0x720" rm -f $DIR/$tfile @@ -17236,8 +18499,9 @@ test_315() { # LU-618 local file=$DIR/$tfile rm -f $file - $MULTIOP $file oO_CREAT:O_DIRECT:O_RDWR:w4096000c - $MULTIOP $file oO_RDONLY:r4096000_c & + $MULTIOP $file oO_CREAT:O_DIRECT:O_RDWR:w4063232c || + error "multiop file write failed" + $MULTIOP $file oO_RDONLY:r4063232_c & PID=$! sleep 2 @@ -17251,8 +18515,8 @@ test_315() { # LU-618 run_test 315 "read should be accounted" test_316() { - [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0 - large_xattr_enabled || { skip "large_xattr disabled" && return; } + [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" + large_xattr_enabled || skip_env "ea_inode feature disabled" rm -rf $DIR/$tdir/d mkdir -p $DIR/$tdir/d @@ -17263,6 +18527,73 @@ test_316() { } run_test 316 "lfs mv" +test_317() { + local trunc_sz + local grant_blk_size + + if [ "$(facet_fstype $facet)" == "zfs" ]; then + skip "LU-10370: no implementation for ZFS" && return + fi + + stack_trap "rm -f $DIR/$tfile" EXIT + grant_blk_size=$($LCTL get_param osc.$FSNAME*.import | + awk '/grant_block_size:/ { print $2; exit; }') + # + # Create File of size 5M. Truncate it to below size's and verify + # blocks count. + # + dd if=/dev/zero of=$DIR/$tfile bs=5M count=1 conv=fsync || + error "Create file : $DIR/$tfile" + + for trunc_sz in 2097152 4097 4000 509 0; do + $TRUNCATE $DIR/$tfile $trunc_sz || + error "truncate $tfile to $trunc_sz failed" + local sz=$(stat --format=%s $DIR/$tfile) + local blk=$(stat --format=%b $DIR/$tfile) + local trunc_blk=$((((trunc_sz + (grant_blk_size - 1) ) / + grant_blk_size) * 8)) + + if [[ $blk -ne $trunc_blk ]]; then + $(which stat) $DIR/$tfile + error "Expected Block $trunc_blk got $blk for $tfile" + fi + + $CHECKSTAT -s $trunc_sz $DIR/$tfile || + error "Expected Size $trunc_sz got $sz for $tfile" + done + + # + # sparse file test + # Create file with a hole and write actual two blocks. Block count + # must be 16. + # + dd if=/dev/zero of=$DIR/$tfile bs=$grant_blk_size count=2 seek=5 \ + conv=fsync || error "Create file : $DIR/$tfile" + + # Calculate the final truncate size. + trunc_sz=$(($(stat --format=%s $DIR/$tfile) - (grant_blk_size + 1))) + + # + # truncate to size $trunc_sz bytes. Strip the last block + # The block count must drop to 8 + # + $TRUNCATE $DIR/$tfile $trunc_sz || + error "truncate $tfile to $trunc_sz failed" + + local trunc_bsz=$((grant_blk_size / $(stat --format=%B $DIR/$tfile))) + sz=$(stat --format=%s $DIR/$tfile) + blk=$(stat --format=%b $DIR/$tfile) + + if [[ $blk -ne $trunc_bsz ]]; then + $(which stat) $DIR/$tfile + error "Expected Block $trunc_bsz got $blk for $tfile" + fi + + $CHECKSTAT -s $trunc_sz $DIR/$tfile || + error "Expected Size $trunc_sz got $sz for $tfile" +} +run_test 317 "Verify blocks get correctly update after truncate" + test_fake_rw() { local read_write=$1 if [ "$read_write" = "write" ]; then @@ -17322,41 +18653,40 @@ test_fake_rw() { rm -f $DIR/$tfile } test_399a() { # LU-7655 for OST fake write - remote_ost_nodsh && skip "remote OST with nodsh" && return + remote_ost_nodsh && skip "remote OST with nodsh" test_fake_rw write } run_test 399a "fake write should not be slower than normal write" test_399b() { # LU-8726 for OST fake read - remote_ost_nodsh && skip "remote OST with nodsh" && return - + remote_ost_nodsh && skip "remote OST with nodsh" if [ "$(facet_fstype ost1)" != "ldiskfs" ]; then - skip "ldiskfs only test" && return 0 + skip_env "ldiskfs only test" fi + test_fake_rw read } run_test 399b "fake read should not be slower than normal read" test_400a() { # LU-1606, was conf-sanity test_74 + if ! which $CC > /dev/null 2>&1; then + skip_env "$CC is not installed" + fi + local extra_flags='' local out=$TMP/$tfile local prefix=/usr/include/lustre local prog - if ! which $CC > /dev/null 2>&1; then - skip_env "$CC is not installed" - return 0 - fi - if ! [[ -d $prefix ]]; then # Assume we're running in tree and fixup the include path. - extra_flags+=" -I$LUSTRE/include" - extra_flags+=" -L$LUSTRE/utils" + extra_flags+=" -I$LUSTRE/../lnet/include/uapi -I$LUSTRE/include/uapi -I$LUSTRE/include" + extra_flags+=" -L$LUSTRE/utils/.lib" fi for prog in $LUSTRE_TESTS_API_DIR/*.c; do - $CC -Wall -Werror $extra_flags -llustreapi -o $out $prog || + $CC -Wall -Werror $extra_flags -o $out $prog -llustreapi || error "client api broken" done rm -f $out @@ -17377,7 +18707,6 @@ test_400b() { # LU-1606, LU-5011 if ! which $CC > /dev/null 2>&1; then skip_env "$CC is not installed" - return 0 fi for header in $prefix/*.h; do @@ -17398,8 +18727,8 @@ run_test 400b "packaged headers can be compiled" test_401a() { #LU-7437 local printf_arg=$(find -printf 2>&1 | grep "unrecognized:") - [ -n "$printf_arg" ] && skip_env "find does not support -printf" && - return + [ -n "$printf_arg" ] && skip_env "find does not support -printf" + #count the number of parameters by "list_param -R" local params=$($LCTL list_param -R '*' 2>/dev/null | wc -l) #count the number of parameters by listing proc files @@ -17498,9 +18827,9 @@ test_402() { $server_version -lt $(version_code 2.7.50) ]] || [[ $server_version -ge $(version_code 2.7.2) && $server_version -lt $(version_code 2.7.11) ]] || - { skip "Need MDS version 2.7.2+ or 2.7.18.4+ or 2.7.66+"; - return; } - remote_mds_nodsh && skip "remote MDS with nodsh" && return + skip "Need MDS version 2.7.2+ or 2.7.18.4+ or 2.7.66+" + remote_mds_nodsh && skip "remote MDS with nodsh" + $LFS setdirstripe -i 0 $DIR/$tdir || error "setdirstripe -i 0 failed" #define OBD_FAIL_MDS_FLD_LOOKUP 0x15c do_facet mds1 "lctl set_param fail_loc=0x8000015c" @@ -17541,9 +18870,9 @@ run_test 403 "i_nlink should not drop to zero due to aliasing" test_404() { # LU-6601 local server_version=$(lustre_version_code $SINGLEMDS) [[ $server_version -ge $(version_code 2.8.53) ]] || - { skip "Need server version newer than 2.8.52"; return 0; } + skip "Need server version newer than 2.8.52" + remote_mds_nodsh && skip "remote MDS with nodsh" - remote_mds_nodsh && skip "remote MDS with nodsh" && return local mosps=$(do_facet $SINGLEMDS $LCTL dl | awk '/osp .*-osc-MDT/ { print $4}') @@ -17572,9 +18901,8 @@ run_test 404 "validate manual {de}activated works properly for OSPs" test_405() { [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.6.92) -o \ [ $(lustre_version_code client) -lt $(version_code 2.6.99) ] && - skip "Layout swap lock is not supported" && return - - check_swap_layouts_support && return 0 + skip "Layout swap lock is not supported" + check_swap_layouts_support test_mkdir $DIR/$tdir swap_lock_test -d $DIR/$tdir || @@ -17583,12 +18911,12 @@ test_405() { run_test 405 "Various layout swap lock tests" test_406() { - [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return - [ $OSTCOUNT -lt 2 ] && skip "needs >= 2 OSTs" && return - [ -n "$FILESET" ] && skip "SKIP due to FILESET set" && return - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return + [ $MDSCOUNT -lt 2 ] && skip_env "needs >= 2 MDTs" + [ $OSTCOUNT -lt 2 ] && skip_env "needs >= 2 OSTs" + [ -n "$FILESET" ] && skip "SKIP due to FILESET set" + [ $PARALLEL == "yes" ] && skip "skip parallel run" [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.8.50) ] && - skip "Need MDS version at least 2.8.50" && return + skip "Need MDS version at least 2.8.50" local def_stripe_size=$($LFS getstripe -S $MOUNT) local test_pool=$TESTNAME @@ -17656,10 +18984,10 @@ test_406() { run_test 406 "DNE support fs default striping" test_407() { - [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return + [ $MDSCOUNT -lt 2 ] && skip_env "needs >= 2 MDTs" [[ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.8.55) ]] && - skip "Need MDS version at least 2.8.55" && return - remote_mds_nodsh && skip "remote MDS with nodsh" && return + skip "Need MDS version at least 2.8.55" + remote_mds_nodsh && skip "remote MDS with nodsh" $LFS mkdir -i 0 -c 1 $DIR/$tdir.0 || error "$LFS mkdir -i 0 -c 1 $tdir.0 failed" @@ -17679,7 +19007,7 @@ test_407() { run_test 407 "transaction fail should cause operation fail" test_408() { - dd if=/dev/zero of=$DIR/$tfile bs=4096 count=1 oflag=direct + dd if=/dev/zero of=$DIR/$tfile bs=$PAGE_SIZE count=1 oflag=direct #define OBD_FAIL_OSC_BRW_PREP_REQ2 0x40a lctl set_param fail_loc=0x8000040a @@ -17700,9 +19028,7 @@ run_test 408 "drop_caches should not hang due to page leaks" test_409() { - [ $MDSCOUNT -lt 2 ] && - skip "We need at least 2 MDTs for this test" && return - + [ $MDSCOUNT -lt 2 ] && skip_env "needs >= 2 MDTs" check_mount_and_prep mkdir -p $DIR/$tdir || error "(0) Fail to mkdir" @@ -17733,7 +19059,7 @@ run_test 409 "Large amount of cross-MDTs hard links on the same file" test_410() { [[ $(lustre_version_code client) -lt $(version_code 2.9.59) ]] && - skip "Need client version at least 2.9.59" && return + skip "Need client version at least 2.9.59" # Create a file, and stat it from the kernel local testfile=$DIR/$tfile @@ -17763,7 +19089,7 @@ test_411() { local cg_basedir=/sys/fs/cgroup/memory # LU-9966 test -f "$cg_basedir/memory.kmem.limit_in_bytes" || - { skip "no setup for cgroup"; return; } + skip "no setup for cgroup" dd if=/dev/zero of=$DIR/$tfile bs=1M count=100 conv=fsync || error "test file creation failed" @@ -17777,9 +19103,9 @@ test_411() { echo 1M > $cgdir/memory.limit_in_bytes # Should not LBUG, just be killed by oom-killer - sh -c "echo \$$ > $cgdir/tasks && dd if=$DIR/$tfile of=/dev/null" && - error "fail to trigger a memory allocation error" - + # dd will return 0 even allocation failure in some environment. + # So don't check return value + sh -c "echo \$$ > $cgdir/tasks && dd if=$DIR/$tfile of=/dev/null" cleanup_test411_cgroup $cgdir return 0 @@ -17787,11 +19113,9 @@ test_411() { run_test 411 "Slab allocation error with cgroup does not LBUG" test_412() { - [ $MDSCOUNT -lt 2 ] && - skip "We need at least 2 MDTs for this test" && return - + [ $MDSCOUNT -lt 2 ] && skip_env "needs >= 2 MDTs" if [ $(lustre_version_code mds1) -lt $(version_code 2.10.55) ]; then - skip "Need server version at least 2.10.55" && exit 0 + skip "Need server version at least 2.10.55" fi $LFS mkdir -i $((MDSCOUNT - 1)),$((MDSCOUNT - 2)) $DIR/$tdir || @@ -17808,10 +19132,10 @@ run_test 412 "mkdir on specific MDTs" test_413() { [ $MDSCOUNT -lt 2 ] && - skip "We need at least 2 MDTs for this test" && return + skip "We need at least 2 MDTs for this test" if [ $(lustre_version_code mds1) -lt $(version_code 2.10.55) ]; then - skip "Need server version at least 2.10.55" && exit 0 + skip "Need server version at least 2.10.55" fi mkdir $DIR/$tdir || error "mkdir failed" @@ -17842,10 +19166,76 @@ test_413() { } run_test 413 "mkdir on less full MDTs" +test_414() { +#define OBD_FAIL_PTLRPC_BULK_ATTACH 0x521 + $LCTL set_param fail_loc=0x80000521 + dd if=/dev/zero of=$DIR/$tfile bs=2M count=1 oflag=sync + rm -f $DIR/$tfile +} +run_test 414 "simulate ENOMEM in ptlrpc_register_bulk()" + +test_415() { + [ $PARALLEL == "yes" ] && skip "skip parallel run" + [ $(lustre_version_code mds1) -lt $(version_code 2.11.52) ] && + skip "Need server version at least 2.11.52" + + # LU-11102 + local total + local setattr_pid + local start_time + local end_time + local duration + + total=500 + # this test may be slow on ZFS + [ "$(facet_fstype mds1)" == "zfs" ] && total=100 + + # though this test is designed for striped directory, let's test normal + # directory too since lock is always saved as CoS lock. + test_mkdir $DIR/$tdir || error "mkdir $tdir" + createmany -o $DIR/$tdir/$tfile. $total || error "createmany" + + ( + while true; do + touch $DIR/$tdir + done + ) & + setattr_pid=$! + + start_time=$(date +%s) + for i in $(seq $total); do + mrename $DIR/$tdir/$tfile.$i $DIR/$tdir/$tfile-new.$i \ + > /dev/null + done + end_time=$(date +%s) + duration=$((end_time - start_time)) + + kill -9 $setattr_pid + + echo "rename $total files took $duration sec" + [ $duration -lt 100 ] || error "rename took $duration sec" +} +run_test 415 "lock revoke is not missing" + + +test_416() { + [ $(lustre_version_code mds1) -lt $(version_code 2.11.55) ] && + skip "Need server version at least 2.11.55" + + # define OBD_FAIL_OSD_TXN_START 0x19a + do_facet mds1 lctl set_param fail_loc=0x19a + + lfs mkdir -c $MDSCOUNT $DIR/$tdir + + true +} +run_test 416 "transaction start failure won't cause system hung" + prep_801() { [[ $(lustre_version_code mds1) -lt $(version_code 2.9.55) ]] || [[ $(lustre_version_code ost1) -lt $(version_code 2.9.55) ]] && - skip "Need server version at least 2.9.55" & exit 0 + skip "Need server version at least 2.9.55" + start_full_debug_logging } @@ -18013,7 +19403,7 @@ test_801b() { run_test 801b "modification will be blocked by write barrier" test_801c() { - [[ $MDSCOUNT -lt 2 ]] && skip "needs >= 2 MDTs" && return + [[ $MDSCOUNT -lt 2 ]] && skip_env "needs >= 2 MDTs" prep_801 @@ -18070,7 +19460,9 @@ test_802() { [[ $(lustre_version_code mds1) -lt $(version_code 2.9.55) ]] || [[ $(lustre_version_code ost1) -lt $(version_code 2.9.55) ]] && - skip "Need server version at least 2.9.55" & exit 0 + skip "Need server version at least 2.9.55" + + [[ $ENABLE_QUOTA ]] && skip "Quota enabled for read-only test" mkdir $DIR/$tdir || error "(1) fail to mkdir" @@ -18111,9 +19503,9 @@ test_802() { run_test 802 "simulate readonly device" test_803() { - [[ $MDSCOUNT -lt 2 ]] && skip "needs >= 2 MDTs" && return + [[ $MDSCOUNT -lt 2 ]] && skip_env "needs >= 2 MDTs" [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.10.54) ] && - skip "MDS needs to be newer than 2.10.54" && return + skip "MDS needs to be newer than 2.10.54" mkdir -p $DIR/$tdir # Create some objects on all MDTs to trigger related logs objects @@ -18159,12 +19551,11 @@ test_803() { run_test 803 "verify agent object for remote object" test_804() { - [[ $MDSCOUNT -lt 2 ]] && skip "needs >= 2 MDTs" && return + [[ $MDSCOUNT -lt 2 ]] && skip_env "needs >= 2 MDTs" [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.10.54) ] && - skip "MDS needs to be newer than 2.10.54" && return - + skip "MDS needs to be newer than 2.10.54" [ "$(facet_fstype $SINGLEMDS)" != "ldiskfs" ] && - skip "ldiskfs only test" && return 0 + skip_env "ldiskfs only test" mkdir -p $DIR/$tdir $LFS mkdir -c 1 -i 1 $DIR/$tdir/dir0 || @@ -18242,17 +19633,272 @@ test_804() { } run_test 804 "verify agent entry for remote entry" +cleanup_805() { + do_facet $SINGLEMDS zfs set quota=$old $fsset + unlinkmany $DIR/$tdir/f- 1000000 + trap 0 +} + +test_805() { + local zfs_version=$(do_node $SINGLEMDS cat /sys/module/zfs/version) + [ "$(facet_fstype mds1)" != "zfs" ] && + skip "ZFS specific test" + [ $(version_code $zfs_version) -lt $(version_code 0.7.2) ] && + skip "netfree not implemented before 0.7" + [[ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.10.57) ]] || + skip "Need MDS version at least 2.10.57" + + local fsset + local freekb + local usedkb + local old + local quota + local pref="osd-zfs.lustre-MDT0000." + + # limit available space on MDS dataset to meet nospace issue + # quickly. then ZFS 0.7.2 can use reserved space if asked + # properly (using netfree flag in osd_declare_destroy() + fsset=$(do_facet $SINGLEMDS lctl get_param -n $pref.mntdev) + old=$(do_facet $SINGLEMDS zfs get -H quota $fsset | \ + gawk '{print $3}') + freekb=$(do_facet $SINGLEMDS lctl get_param -n $pref.kbytesfree) + usedkb=$(do_facet $SINGLEMDS lctl get_param -n $pref.kbytestotal) + let "usedkb=usedkb-freekb" + let "freekb=freekb/2" + if let "freekb > 5000"; then + let "freekb=5000" + fi + do_facet $SINGLEMDS zfs set quota=$(((usedkb+freekb)*1024)) $fsset + trap cleanup_805 EXIT + mkdir $DIR/$tdir + $LFS setstripe -E 1M -L mdt $DIR/$tdir || error "DoM not working" + createmany -m $DIR/$tdir/f- 1000000 && error "ENOSPC wasn't met" + rm -rf $DIR/$tdir || error "not able to remove" + do_facet $SINGLEMDS zfs set quota=$old $fsset + trap 0 +} +run_test 805 "ZFS can remove from full fs" + +# Size-on-MDS test +check_lsom_data() +{ + local file=$1 + local size=$($LFS getsom -s $file) + local expect=$(stat -c %s $file) + + [[ $size == $expect ]] || + error "$file expected size: $expect, got: $size" + + local blocks=$($LFS getsom -b $file) + expect=$(stat -c %b $file) + [[ $blocks == $expect ]] || + error "$file expected blocks: $expect, got: $blocks" +} + +check_lsom_size() +{ + local size=$($LFS getsom -s $1) + local expect=$2 + + [[ $size == $expect ]] || + error "$file expected size: $expect, got: $size" +} + +test_806() { + [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.11.52) ] && + skip "Need MDS version at least 2.11.52" && return + + local bs=1048576 + + touch $DIR/$tfile || error "touch $tfile failed" + + local save="$TMP/$TESTSUITE-$TESTNAME.parameters" + save_lustre_params client "llite.*.xattr_cache" > $save + lctl set_param llite.*.xattr_cache=0 + stack_trap "restore_lustre_params < $save" EXIT + + # single-threaded write + echo "Test SOM for single-threaded write" + dd if=/dev/zero of=$DIR/$tfile bs=$bs count=1 || + error "write $tfile failed" + check_lsom_size $DIR/$tfile $bs + + local num=32 + local size=$(($num * $bs)) + local offset=0 + local i + + echo "Test SOM for single client muti-threaded($num) write" + $TRUNCATE $DIR/$tfile 0 + for ((i = 0; i < $num; i++)); do + $MULTIOP $DIR/$tfile Oz${offset}w${bs}c & + local pids[$i]=$! + offset=$((offset + $bs)) + done + for (( i=0; i < $num; i++ )); do + wait ${pids[$i]} + done + check_lsom_size $DIR/$tfile $size + + $TRUNCATE $DIR/$tfile 0 + for ((i = 0; i < $num; i++)); do + offset=$((offset - $bs)) + $MULTIOP $DIR/$tfile Oz${offset}w${bs}c & + local pids[$i]=$! + done + for (( i=0; i < $num; i++ )); do + wait ${pids[$i]} + done + check_lsom_size $DIR/$tfile $size + + # multi-client wirtes + num=$(get_node_count ${CLIENTS//,/ }) + size=$(($num * $bs)) + offset=0 + i=0 + + echo "Test SOM for muti-client ($num) writes" + $TRUNCATE $DIR/$tfile 0 + for client in ${CLIENTS//,/ }; do + do_node $client $MULTIOP $DIR/$tfile Oz${offset}w${bs}c & + local pids[$i]=$! + i=$((i + 1)) + offset=$((offset + $bs)) + done + for (( i=0; i < $num; i++ )); do + wait ${pids[$i]} + done + check_lsom_size $DIR/$tfile $offset + + i=0 + $TRUNCATE $DIR/$tfile 0 + for client in ${CLIENTS//,/ }; do + offset=$((offset - $bs)) + do_node $client $MULTIOP $DIR/$tfile Oz${offset}w${bs}c & + local pids[$i]=$! + i=$((i + 1)) + done + for (( i=0; i < $num; i++ )); do + wait ${pids[$i]} + done + check_lsom_size $DIR/$tfile $size + + # verify truncate + echo "Test SOM for truncate" + $TRUNCATE $DIR/$tfile 1048576 + check_lsom_size $DIR/$tfile 1048576 + $TRUNCATE $DIR/$tfile 1234 + check_lsom_size $DIR/$tfile 1234 + + # verify SOM blocks count + echo "Verify SOM block count" + $TRUNCATE $DIR/$tfile 0 + $MULTIOP $DIR/$tfile oO_TRUNC:O_RDWR:w1048576YSc || + error "failed to write file $tfile" + check_lsom_data $DIR/$tfile +} +run_test 806 "Verify Lazy Size on MDS" + +test_807() { + [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.11.52) ] && + skip "Need MDS version at least 2.11.52" && return + + # Registration step + changelog_register || error "changelog_register failed" + local cl_user="${CL_USERS[$SINGLEMDS]%% *}" + changelog_users $SINGLEMDS | grep -q $cl_user || + error "User $cl_user not found in changelog_users" + + local save="$TMP/$TESTSUITE-$TESTNAME.parameters" + save_lustre_params client "llite.*.xattr_cache" > $save + lctl set_param llite.*.xattr_cache=0 + stack_trap "restore_lustre_params < $save" EXIT + + rm -rf $DIR/$tdir || error "rm $tdir failed" + mkdir -p $DIR/$tdir || error "mkdir $tdir failed" + touch $DIR/$tdir/trunc || error "touch $tdir/trunc failed" + $TRUNCATE $DIR/$tdir/trunc 1024 || error "truncate $tdir/trunc failed" + $TRUNCATE $DIR/$tdir/trunc 1048576 || + error "truncate $tdir/trunc failed" + + local bs=1048576 + dd if=/dev/zero of=$DIR/$tdir/single_dd bs=$bs count=1 || + error "write $tfile failed" + + # multi-client wirtes + local num=$(get_node_count ${CLIENTS//,/ }) + local offset=0 + local i=0 + + echo "Test SOM for muti-client ($num) writes" + touch $DIR/$tfile || error "touch $tfile failed" + $TRUNCATE $DIR/$tfile 0 + for client in ${CLIENTS//,/ }; do + do_node $client $MULTIOP $DIR/$tfile Oz${offset}w${bs}c & + local pids[$i]=$! + i=$((i + 1)) + offset=$((offset + $bs)) + done + for (( i=0; i < $num; i++ )); do + wait ${pids[$i]} + done + + sleep 5 + $LSOM_SYNC -u $cl_user -m $FSNAME-MDT0000 $MOUNT + check_lsom_data $DIR/$tdir/trunc + check_lsom_data $DIR/$tdir/single_dd + check_lsom_data $DIR/$tfile + + rm -rf $DIR/$tdir + # Deregistration step + changelog_deregister || error "changelog_deregister failed" +} +run_test 807 "verify LSOM syncing tool" + +check_som_nologged() +{ + local lines=$($LFS changelog $FSNAME-MDT0000 | + grep 'x=trusted.som' | wc -l) + [ $lines -ne 0 ] && error "trusted.som xattr is logged in Changelogs" +} + +test_808() { + [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.11.55) ] && + skip "Need MDS version at least 2.11.55" && return + + # Registration step + changelog_register || error "changelog_register failed" + + touch $DIR/$tfile || error "touch $tfile failed" + check_som_nologged + + dd if=/dev/zero of=$DIR/$tfile bs=1048576 count=1 || + error "write $tfile failed" + check_som_nologged + + $TRUNCATE $DIR/$tfile 1234 + check_som_nologged + + $TRUNCATE $DIR/$tfile 1048576 + check_som_nologged + + # Deregistration step + changelog_deregister || error "changelog_deregister failed" +} +run_test 808 "Check trusted.som xattr not logged in Changelogs" + # # tests that do cleanup/setup should be run at the end # test_900() { - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return - local ls - #define OBD_FAIL_MGC_PAUSE_PROCESS_LOG 0x903 - $LCTL set_param fail_loc=0x903 + [ $PARALLEL == "yes" ] && skip "skip parallel run" + local ls + + #define OBD_FAIL_MGC_PAUSE_PROCESS_LOG 0x903 + $LCTL set_param fail_loc=0x903 - cancel_lru_locks MGC + cancel_lru_locks MGC FAIL_ON_ERROR=true cleanup FAIL_ON_ERROR=true setup