X-Git-Url: https://git.whamcloud.com/?p=fs%2Flustre-release.git;a=blobdiff_plain;f=lustre%2Ftests%2Fsanity.sh;h=500d660e7d143ad4713e3fddd59b71c92bd031c1;hp=438f6aacbf6c3a222cb38c66460fd41e1cfc1a61;hb=c038909fbc2b3b14763dd731e9c877d11d338f04;hpb=9146d261f35b394e10afde3eec2d5895425261e0 diff --git a/lustre/tests/sanity.sh b/lustre/tests/sanity.sh index 438f6aa..500d660 100755 --- a/lustre/tests/sanity.sh +++ b/lustre/tests/sanity.sh @@ -8,16 +8,27 @@ 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 + +if [[ $(uname -m) = aarch64 ]]; then + # bug number: LU-11596 (all below) + ALWAYS_EXCEPT+=" 42d 42e 63a 63b 64a 64b 64c" + # bug number: LU-11671 LU-11665 LU-11594 LU-11667 + ALWAYS_EXCEPT+=" 45 101c 103a 317" +fi # Check Grants after these tests GRANT_CHECK_LIST="$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 +70,11 @@ 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" @@ -101,10 +117,8 @@ elif [ -r /etc/os-release ]; then if [[ $ubuntu_version -gt $(version_code 16.0.0) ]]; then # bug number for skipped test: - # LU-10334 LU-10335 LU-10335 LU-10335 - ALWAYS_EXCEPT+=" 103a 130a 130b 130c" - # LU-10335 LU-10335 LU-10365 LU-10366 - ALWAYS_EXCEPT+=" 130d 130e 400a 410" + # LU-10334 LU-10366 + ALWAYS_EXCEPT+=" 103a 410" fi fi fi @@ -901,6 +915,19 @@ test_23c() { } run_test 23c "O_APPEND size checks for tiny writes" +# LU-11069 file offset is correct after appending writes +test_23d() { + local file=$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' @@ -1117,12 +1144,6 @@ test_24u() { # bug12192 } run_test 24u "create stripe file" -page_size() { - local size - size=$(getconf PAGE_SIZE 2>/dev/null) - echo -n ${size:-4096} -} - simple_cleanup_common() { local rc=0 trap 0 @@ -1180,7 +1201,7 @@ test_24v() { # take into account of overhead in lu_dirpage header and end mark in # each page, plus one in rpc_num calculation. local dirent_size=$((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)) @@ -2277,6 +2298,32 @@ test_27G() { #LU-10629 } 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 @@ -3161,7 +3208,7 @@ 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 @@ -3947,7 +3994,6 @@ test_42e() { # bug22074 [ $PARALLEL == "yes" ] && skip "skip parallel run" local TDIR=$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]*" @@ -3993,7 +4039,7 @@ 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 $proc_osc0/rpc_stats @@ -4172,10 +4218,10 @@ test_46() { f="$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=$PAGE_SIZE seek=511 count=1 sync start_writeback } @@ -4616,7 +4662,7 @@ test_54b() { f="$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 ======================" @@ -4655,17 +4701,17 @@ test_54c() { 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 } @@ -5471,7 +5517,7 @@ 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" @@ -5861,6 +5907,9 @@ test_56ab() { # LU-10705 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 @@ -6211,6 +6260,10 @@ 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 @@ -6233,22 +6286,24 @@ test_60b() { # bug 6411 [ $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" @@ -6293,11 +6348,41 @@ 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() { [ $PARALLEL == "yes" ] && skip "skip parallel run" f="$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 @@ -6388,8 +6473,6 @@ 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) @@ -6398,15 +6481,15 @@ 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 | @@ -7163,29 +7246,6 @@ test_77g() { # bug 10889 } run_test 77g "checksum error on OST write, read" -test_77j() { # bug 13805 - [ $PARALLEL == "yes" ] && skip "skip parallel run" - $GSS && skip_env "could not run with gss" - - #define OBD_FAIL_OSC_CKSUM_ADLER_ONLY 0x40c - lctl set_param fail_loc=0x40c - remount_client $MOUNT - lctl set_param fail_loc=0 - # wait async osc connect to finish and reflect updated state value - local i - for (( i=0; i < OSTCOUNT; i++ )) ; do - wait_osc_import_state client ost$((i+1)) FULL - 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" - done - remount_client $MOUNT -} -run_test 77j "client only supporting ADLER32" - test_77k() { # LU-10906 [ $PARALLEL == "yes" ] && skip "skip parallel run" $GSS && skip_env "could not run with gss" @@ -7778,7 +7838,7 @@ 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 || @@ -8116,7 +8176,7 @@ test_102h() { # bug 15777 run_test 102h "grow xattr from inside inode to external block" test_102ha() { - large_xattr_enabled || skip_env "large_xattr disabled" + large_xattr_enabled || skip_env "ea_inode feature disabled" grow_xattr $(max_xattr_size) } @@ -8367,6 +8427,56 @@ 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 @@ -10074,38 +10184,44 @@ run_test 127a "verify the client stats are sane" test_127b() { # bug LU-333 [ $PARALLEL == "yes" ] && skip "skip parallel run" + local name count samp unit min max sum sumsq $LCTL set_param llite.*.stats=0 - FSIZE=65536 # sized fixed to match PAGE_SIZE for most clients # perform 2 reads and writes so MAX is different from SUM. - dd if=/dev/zero of=$DIR/$tfile bs=$FSIZE count=1 - dd if=/dev/zero of=$DIR/$tfile bs=$FSIZE count=1 + dd if=/dev/zero of=$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/$tfile of=/dev/null bs=$FSIZE count=1 - dd if=$DIR/$tfile of=/dev/null bs=$FSIZE 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 $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 + 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 [ "$read_bytes" ] || error "Missing read_bytes stats" @@ -10604,7 +10720,6 @@ check_stats() { ;; *) error "Wrong facet '$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 @@ -10663,6 +10778,8 @@ test_133a() { run_test 133a "Verifying MDT stats ========================================" test_133b() { + local res + [ $PARALLEL == "yes" ] && skip "skip parallel run" remote_ost_nodsh && skip "remote OST with nodsh" remote_mds_nodsh && skip "remote MDS with nodsh" @@ -10686,10 +10803,43 @@ 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 ==================================" @@ -10910,6 +11060,18 @@ 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) @@ -10918,7 +11080,7 @@ 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 \ @@ -10927,7 +11089,7 @@ test_133f() { -not -name force_lbug \ -not -name changelog_mask \ -exec badarea_io '{}' \; || - error "find $proc_dirs failed" + error_133 "find $proc_dirs failed" } run_test 133f "Check reads/writes of client lustre proc files with bad area io" @@ -10961,7 +11123,7 @@ test_133g() { -not -name force_lbug \ -not -name changelog_mask \ -exec badarea_io '{}' \\\; || - error "$facet find $facet_proc_dirs failed" + error_133 "$facet find $facet_proc_dirs failed" done # remount the FS in case writes/reads /proc break the FS @@ -11744,8 +11906,7 @@ test_155_big_load() { save_writethrough() { local facets=$(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() { @@ -12237,18 +12398,21 @@ test_160e() { run_test 160e "changelog negative testing (should return errors)" test_160f() { - remote_mds_nodsh && skip "remote MDS with nodsh" + remote_mds_nodsh && skip "remote MDS with nodsh" && return [[ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.10.56) ]] || - skip "Need MDS version at least 2.10.56" + { skip "Need MDS version at least 2.10.56"; return 0; } local mdts=$(comma_list $(mdts_nodes)) # Create a user changelog_register || error "first changelog_register failed" changelog_register || error "second changelog_register failed" - local cl_users=(${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" @@ -12256,11 +12420,11 @@ 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) @@ -12271,44 +12435,79 @@ 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 $first_rec" + done } run_test 160f "changelog garbage collect (timestamped users)" @@ -12325,9 +12524,12 @@ test_160g() { # Create a user changelog_register || error "first changelog_register failed" changelog_register || error "second changelog_register failed" - local cl_users=(${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" @@ -12335,11 +12537,11 @@ 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) @@ -12355,41 +12557,235 @@ 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 $SINGLEMDS \ - "ps -e -o comm= | grep chlg_gc_thread" "" 20 - - # check user still registered - [ -n "$(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" - - # 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 160g "changelog garbage collect (old users)" 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) + [[ $nbcl -eq 0 ]] && error "no changelogs found" + + for param in "changelog_max_idle_time=10" \ + "changelog_gc=1" \ + "changelog_min_gc_interval=2"; do + local MDT0=$(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 $first_rec" + done +} +run_test 160h "changelog gc thread stop upon umount, orphan records delete " \ + "during mount" + +test_160i() { local mdts=$(comma_list $(mdts_nodes)) @@ -12447,7 +12843,7 @@ test_160h() { error "changelogs are off on mds$i" done } -run_test 160h "changelog user register/unregister race" +run_test 160i "changelog user register/unregister race" test_161a() { [ $PARALLEL == "yes" ] && skip "skip parallel run" @@ -13255,6 +13651,20 @@ test_184e() { } run_test 184e "Recreate layout after stripeless layout swaps" +test_184f() { + # Create a file with name longer than sizeof(struct stat) == + # 144 to see if we can get chars from the file name to appear + # in the returned striping. Note that 'f' == 0x66. + local file=$(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) ]] || @@ -13563,8 +13973,8 @@ 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" $new_jobenv } test_205() { # Job stats @@ -13580,9 +13990,13 @@ test_205() { # Job stats local old_jobenv=$($LCTL get_param -n jobid_var) [ $old_jobenv != $JOBENV ] && jobstats_set $JOBENV - stack_trap "do_facet mgs \ - $LCTL conf_param $FSNAME.sys.jobid_var=$old_jobenv" EXIT + 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 $SINGLEMDS lctl get_param -n \ @@ -14214,17 +14628,15 @@ test_224c() { # LU-6441 osc.*.max_pages_per_rpc) local at_max=$($LCTL get_param -n at_max) local timeout=$($LCTL get_param -n timeout) - local test_at="$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 @@ -14233,10 +14645,9 @@ 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 @@ -14554,6 +14965,8 @@ run_test 229 "getstripe/stat/rm/attr changes work on released files" test_230a() { [ $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 @@ -14580,6 +14993,8 @@ run_test 230a "Create remote directory and files under the remote directory" test_230b() { [ $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 @@ -14743,11 +15158,14 @@ test_230b() { run_test 230b "migrate directory" test_230c() { - [ $MDSCOUNT -lt 2 ] && skip_env "needs >= 2 MDTs" [ $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 @@ -14756,35 +15174,59 @@ 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 @@ -14795,20 +15237,37 @@ test_230c() { 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" [ $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 MDTIDX=1 - local mdt_index 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} @@ -14816,14 +15275,23 @@ 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}" + 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" @@ -14833,6 +15301,8 @@ run_test 230d "check migrate big directory" test_230e() { [ $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 @@ -14879,6 +15349,8 @@ run_test 230e "migrate mulitple local link files" test_230f() { [ $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 @@ -14928,6 +15400,8 @@ run_test 230f "migrate mulitple remote link files" test_230g() { [ $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 @@ -14940,8 +15414,8 @@ run_test 230g "migrate dir to non-exist MDT" test_230h() { [ $PARALLEL == "yes" ] && skip "skip parallel run" [ $MDSCOUNT -lt 2 ] && skip_env "needs >= 2 MDTs" - [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.7.64) ] && - skip "Need MDS version at least 2.7.64" + [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.11.52) ] && + skip "Need MDS version at least 2.11.52" local mdt_index @@ -14953,21 +15427,19 @@ 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" [ $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 @@ -14979,12 +15451,79 @@ test_230i() { } run_test 230i "lfs migrate -m tolerates trailing slashes" -test_231a() -{ +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_230k() { + [ $MDSCOUNT -lt 4 ] && skip "needs >= 4 MDTs" + [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.11.56) ] && + skip "Need MDS version at least 2.11.56" + + local total=20 + local files_on_starting_mdt=0 + + $LFS mkdir -i -1 -c 2 $DIR/$tdir || error "mkdir failed" + $LFS getdirstripe $DIR/$tdir + for i in $(seq $total); do + echo $((i*i - i)) > $DIR/$tdir/$tfile.$i || error "write failed" + [[ $($LFS getstripe -m $DIR/$tdir/$tfile.$i) -eq 0 ]] && + files_on_starting_mdt=$((files_on_starting_mdt + 1)) + done + + echo "$files_on_starting_mdt files on MDT0" + + $LFS migrate -m 1,3 $DIR/$tdir || error "migrate -m 1,3 failed" + $LFS getdirstripe $DIR/$tdir + + files_on_starting_mdt=0 + for i in $(seq $total); do + $(echo $((i*i - i)) | cmp $DIR/$tdir/$tfile.$i -) || + error "file $tfile.$i mismatch after migration" + [[ $($LFS getstripe -m $DIR/$tdir/$tfile.$i) -eq 1 ]] && + files_on_starting_mdt=$((files_on_starting_mdt + 1)) + done + + echo "$files_on_starting_mdt files on MDT1 after migration" + [[ $files_on_starting_mdt -eq $total ]] && error "all files on MDT1" + + $LFS migrate -m 0 -c 2 $DIR/$tdir || error "migrate -m 0 -c 2 failed" + $LFS getdirstripe $DIR/$tdir + + files_on_starting_mdt=0 + for i in $(seq $total); do + $(echo $((i*i - i)) | cmp $DIR/$tdir/$tfile.$i -) || + error "file $tfile.$i mismatch after 2nd migration" + [[ $($LFS getstripe -m $DIR/$tdir/$tfile.$i) -eq 0 ]] && + files_on_starting_mdt=$((files_on_starting_mdt + 1)) + done + + echo "$files_on_starting_mdt files on MDT0 after 2nd migration" + [[ $files_on_starting_mdt -eq $total ]] && error "all files on MDT0" + + true +} +run_test 230k "file data not changed after dir migration" + +test_231a() +{ # For simplicity this test assumes that max_pages_per_rpc # is the same across all OSCs local max_pages=$($LCTL get_param -n osc.*.max_pages_per_rpc | head -n1) - local bulk_size=$((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) @@ -15174,18 +15713,6 @@ 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) @@ -15275,36 +15802,48 @@ 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 $PID } run_test 241b "dio vs dio" @@ -16011,6 +16550,9 @@ test_255b() { dd if=/dev/zero of=$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: $total KiB" @@ -16241,6 +16783,45 @@ test_258b() { } run_test 258b "verify i_mutex security behavior" +test_259() { + local file=$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 @@ -16252,7 +16833,7 @@ run_test 260 "Check mdc_close fail" ### Data-on-MDT sanity tests ### test_270a() { [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.10.55) ] && - skip "Need MDS version at least 2.10.55" + skip "Need MDS version at least 2.10.55 for DoM" # create DoM file local dom=$DIR/$tdir/dom_file @@ -16261,14 +16842,13 @@ test_270a() { mkdir -p $DIR/$tdir # basic checks for DoM component creation - $LFS setstripe -E 1024K -E 1024K -L mdt $dom 2>/dev/null && + $LFS setstripe -E 1024K -E 2048K -L mdt $dom 2>/dev/null && error "Can set MDT layout to non-first entry" - $LFS setstripe -E 1024K -L mdt -E 1024K -L mdt $dom 2>/dev/null && + $LFS setstripe -E 1024K -L mdt -E 2048K -L mdt $dom 2>/dev/null && error "Can define multiple entries as MDT layout" - $LFS setstripe -E 1M -L mdt $dom || - error "Can't create DoM layout" + $LFS setstripe -E 1M -L mdt $dom || error "Can't create DoM layout" [ $($LFS getstripe -L $dom) == "mdt" ] || error "bad pattern" [ $($LFS getstripe -c $dom) == 0 ] || error "bad stripe count" @@ -16280,46 +16860,61 @@ test_270a() { local space_check=1 # Skip free space checks with ZFS - if [ "$(facet_fstype $facet)" == "zfs" ]; then - space_check=0 - fi + [ "$(facet_fstype $facet)" == "zfs" ] && space_check=0 # write sync + local size_tmp=$((65536 * 3)) local mdtfree1=$(do_facet $facet \ - lctl get_param -n osd*.*$mdtname.kbytesfree) - dd if=/dev/urandom of=$tmp bs=1024 count=100 + lctl get_param -n osd*.*$mdtname.kbytesfree) + + dd if=/dev/urandom of=$tmp bs=1024 count=$((size_tmp / 1024)) # check also direct IO along write - dd if=$tmp of=$dom bs=102400 count=1 oflag=direct + # IO size must be a multiple of PAGE_SIZE on all platforms (ARM=64KB) + dd if=$tmp of=$dom bs=65536 count=$((size_tmp / 65536)) oflag=direct sync cmp $tmp $dom || error "file data is different" - [ $(stat -c%s $dom) == 102400 ] || error "bad size after write" + [ $(stat -c%s $dom) == $size_tmp ] || + error "bad size after write: $(stat -c%s $dom) != $size_tmp" if [ $space_check == 1 ]; then local mdtfree2=$(do_facet $facet \ - lctl get_param -n osd*.*$mdtname.kbytesfree) - [ $(($mdtfree1 - $mdtfree2)) -ge 102 ] || - error "MDT free space is wrong after write" + lctl get_param -n osd*.*$mdtname.kbytesfree) + + # increase in usage from by $size_tmp + [ $(($mdtfree1 - $mdtfree2)) -ge $((size_tmp / 1024)) ] || + error "MDT free space wrong after write: " \ + "$mdtfree1 >= $mdtfree2 + $size_tmp/1024" fi # truncate - $TRUNCATE $dom 10000 - [ $(stat -c%s $dom) == 10000 ] || error "bad size after truncate" + local size_dom=10000 + + $TRUNCATE $dom $size_dom + [ $(stat -c%s $dom) == $size_dom ] || + error "bad size after truncate: $(stat -c%s $dom) != $size_dom" if [ $space_check == 1 ]; then mdtfree1=$(do_facet $facet \ lctl get_param -n osd*.*$mdtname.kbytesfree) - [ $(($mdtfree1 - $mdtfree2)) -ge 92 ] || - error "MDT free space is wrong after truncate" + # decrease in usage from $size_tmp to new $size_dom + [ $(($mdtfree1 - $mdtfree2)) -ge \ + $(((size_tmp - size_dom) / 1024)) ] || + error "MDT free space is wrong after truncate: " \ + "$mdtfree1 >= $mdtfree2 + ($size_tmp - $size_dom) / 1024" fi # append cat $tmp >> $dom sync - [ $(stat -c%s $dom) == 112400 ] || error "bad size after append" + size_dom=$((size_dom + size_tmp)) + [ $(stat -c%s $dom) == $size_dom ] || + error "bad size after append: $(stat -c%s $dom) != $size_dom" if [ $space_check == 1 ]; then mdtfree2=$(do_facet $facet \ lctl get_param -n osd*.*$mdtname.kbytesfree) - [ $(($mdtfree1 - $mdtfree2)) -ge 102 ] || - error "MDT free space is wrong after append" + # increase in usage by $size_tmp from previous + [ $(($mdtfree1 - $mdtfree2)) -ge $((size_tmp / 1024)) ] || + error "MDT free space is wrong after append: " \ + "$mdtfree1 >= $mdtfree2 + $size_tmp/1024" fi # delete @@ -16327,22 +16922,25 @@ test_270a() { if [ $space_check == 1 ]; then mdtfree1=$(do_facet $facet \ lctl get_param -n osd*.*$mdtname.kbytesfree) - [ $(($mdtfree1 - $mdtfree2)) -ge 112 ] || - error "MDT free space is wrong after removal" + # decrease in usage by $size_dom from previous + [ $(($mdtfree1 - $mdtfree2)) -ge $((size_dom / 1024)) ] || + error "MDT free space is wrong after removal: " \ + "$mdtfree1 >= $mdtfree2 + $size_dom/1024" fi # combined striping $LFS setstripe -E 1024K -L mdt -E EOF $dom || error "Can't create DoM + OST striping" - dd if=/dev/urandom of=$tmp bs=1024 count=2000 + size_tmp=2031616 # must be a multiple of PAGE_SIZE=65536 on ARM + dd if=/dev/urandom of=$tmp bs=1024 count=$((size_tmp / 1024)) # check also direct IO along write - dd if=$tmp of=$dom bs=102400 count=20 oflag=direct + dd if=$tmp of=$dom bs=65536 count=$((size_tmp / 65536)) oflag=direct sync cmp $tmp $dom || error "file data is different" - [ $(stat -c%s $dom) == 2048000 ] || error "bad size after write" - rm $dom - rm $tmp + [ $(stat -c%s $dom) == $size_tmp ] || + error "bad size after write: $(stat -c%s $dom) != $size_tmp" + rm $dom $tmp return 0 } @@ -16487,7 +17085,8 @@ 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 $((dom_limit * 2)) ] && error "Able to create DoM component size more than LOD limit" @@ -16496,6 +17095,19 @@ test_270f() { lod.$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 @@ -16521,7 +17133,8 @@ 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 ${dom}_def || @@ -16545,9 +17158,8 @@ test_271a() { lctl set_param -n mdc.*.stats=clear dd if=/dev/zero of=$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 } @@ -16569,9 +17181,8 @@ 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)" @@ -16593,16 +17204,27 @@ 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)" + +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" @@ -16621,9 +17243,8 @@ 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" @@ -16634,9 +17255,7 @@ 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 ] || @@ -16646,6 +17265,310 @@ 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" [ $(lustre_version_code ost1) -lt $(version_code 2.10.57) ] && @@ -17184,7 +18107,7 @@ 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" $LCTL set_param fail_loc=0 @@ -17557,7 +18480,6 @@ test_312() { # LU-4856 local max_blksz=$(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 @@ -17569,30 +18491,31 @@ 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 $tf @@ -17603,8 +18526,8 @@ test_312() { # LU-4856 dd if=/dev/zero of=$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 $zfs_objid) @@ -17620,12 +18543,13 @@ test_313() { remote_ost_nodsh && skip "remote OST with nodsh" local file=$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 $file @@ -17647,8 +18571,9 @@ test_315() { # LU-618 local file=$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 @@ -17663,7 +18588,7 @@ run_test 315 "read should be accounted" test_316() { [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" - large_xattr_enabled || skip_env "large_xattr disabled" + large_xattr_enabled || skip_env "ea_inode feature disabled" rm -rf $DIR/$tdir/d mkdir -p $DIR/$tdir/d @@ -17674,6 +18599,73 @@ 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 @@ -17761,12 +18753,12 @@ test_400a() { # LU-1606, was conf-sanity test_74 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 $out @@ -18087,7 +19079,7 @@ test_407() { run_test 407 "transaction fail should cause operation fail" test_408() { - dd if=/dev/zero of=$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 @@ -18183,9 +19175,9 @@ 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 @@ -18254,6 +19246,98 @@ test_414() { } 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" + +cleanup_417() { + trap 0 + do_nodes $(comma_list $(mdts_nodes)) \ + "$LCTL set_param -n mdt.*MDT*.enable_dir_migration=1" + do_nodes $(comma_list $(mdts_nodes)) \ + "$LCTL set_param -n mdt.*MDT*.enable_remote_dir=1" + do_nodes $(comma_list $(mdts_nodes)) \ + "$LCTL set_param -n mdt.*MDT*.enable_striped_dir=1" +} + +test_417() { + [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return + [[ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.11.56) ]] && + skip "Need MDS version at least 2.11.56" && return + + trap cleanup_417 RETURN EXIT + + $LFS mkdir -i 1 $DIR/$tdir.1 || error "create remote dir $tdir.1 failed" + do_nodes $(comma_list $(mdts_nodes)) \ + "$LCTL set_param -n mdt.*MDT*.enable_dir_migration=0" + $LFS migrate -m 0 $DIR/$tdir.1 && + error "migrate dir $tdir.1 should fail" + + do_nodes $(comma_list $(mdts_nodes)) \ + "$LCTL set_param -n mdt.*MDT*.enable_remote_dir=0" + $LFS mkdir -i 1 $DIR/$tdir.2 && + error "create remote dir $tdir.2 should fail" + + do_nodes $(comma_list $(mdts_nodes)) \ + "$LCTL set_param -n mdt.*MDT*.enable_striped_dir=0" + $LFS mkdir -c 2 $DIR/$tdir.3 && + error "create striped dir $tdir.3 should fail" + true +} +run_test 417 "disable remote dir, striped dir and dir migration" + prep_801() { [[ $(lustre_version_code mds1) -lt $(version_code 2.9.55) ]] || [[ $(lustre_version_code ost1) -lt $(version_code 2.9.55) ]] && @@ -18485,6 +19569,8 @@ test_802() { [[ $(lustre_version_code ost1) -lt $(version_code 2.9.55) ]] && 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" cp $LUSTRE/tests/test-framework.sh $DIR/$tdir/ || @@ -18700,6 +19786,259 @@ test_805() { } 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" + +check_som_nodata() +{ + $LFS getsom $1 + [[ $? -eq 61 ]] || error "DoM-only file $1 has SOM xattr" +} + +test_809() { + [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.11.56) ] && + skip "Need MDS version at least 2.11.56" && return + + $LFS setstripe -E 1M -L mdt $DIR/$tfile || + error "failed to create DoM-only file $DIR/$tfile" + touch $DIR/$tfile || error "touch $tfile failed" + check_som_nodata $DIR/$tfile + + dd if=/dev/zero of=$DIR/$tfile bs=2048 count=1 || + error "write $tfile failed" + check_som_nodata $DIR/$tfile + + $TRUNCATE $DIR/$tfile 1234 + check_som_nodata $DIR/$tfile + + $TRUNCATE $DIR/$tfile 4097 + check_som_nodata $DIR/$file +} +run_test 809 "Verify no SOM xattr store for DoM-only files" + +test_810() { + local ORIG + local CSUM + + # t10 seem to dislike partial pages + lctl set_param osc.*.checksum_type=adler + lctl set_param fail_loc=0x411 + dd if=/dev/urandom of=$DIR/$tfile bs=10240 count=2 + ORIG=$(md5sum $DIR/$tfile) + lctl set_param ldlm.namespaces.*osc*.lru_size=clear + CSUM=$(md5sum $DIR/$tfile) + set_checksum_type adler + if [ "$ORIG" != "$CSUM" ]; then + error "$ORIG != $CSUM" + fi +} +run_test 810 "partial page writes on ZFS (LU-11663)" + # # tests that do cleanup/setup should be run at the end #