X-Git-Url: https://git.whamcloud.com/?p=fs%2Flustre-release.git;a=blobdiff_plain;f=lustre%2Ftests%2Fsanity.sh;h=3c7e54eae68fb80121ef61e428609ba3c326cb9c;hp=a74c2eaf854ebcc8bf8ba505046558fd7dfd52be;hb=76b9eecdebf830606b021079148eaefa6aab99cc;hpb=e70a6fd8a6400c0460a6c66668103c23b7997d30 diff --git a/lustre/tests/sanity.sh b/lustre/tests/sanity.sh index a74c2ea..3c7e54e 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 .. 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 @@ -106,10 +117,8 @@ elif [ -r /etc/os-release ]; then if [[ $ubuntu_version -gt $(version_code 16.0.0) ]]; then # bug number for skipped test: - # LU-10334 LU-10335 LU-10335 LU-10335 - ALWAYS_EXCEPT+=" 103a 130a 130b 130c" - # LU-10335 LU-10335 LU-10366 - ALWAYS_EXCEPT+=" 130d 130e 410" + # LU-10334 LU-10366 + ALWAYS_EXCEPT+=" 103a 410" fi fi fi @@ -1135,12 +1144,6 @@ test_24u() { # bug12192 } run_test 24u "create stripe file" -page_size() { - local size - size=$(getconf PAGE_SIZE 2>/dev/null) - echo -n ${size:-4096} -} - simple_cleanup_common() { local rc=0 trap 0 @@ -1198,7 +1201,7 @@ test_24v() { # take into account of overhead in lu_dirpage header and end mark in # each page, plus one in rpc_num calculation. local dirent_size=$((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)) @@ -3205,7 +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 @@ -3991,7 +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]*" @@ -4037,7 +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 $proc_osc0/rpc_stats @@ -4216,10 +4218,10 @@ test_46() { f="$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=$PAGE_SIZE seek=511 count=1 sync start_writeback } @@ -4660,7 +4662,7 @@ test_54b() { f="$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 ======================" @@ -4699,17 +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 } @@ -5515,7 +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" @@ -5905,6 +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 @@ -6294,10 +6299,11 @@ test_60b() { # bug 6411 else print from_begin }") - [[ $LLOG_COUNT -gt 100 ]] && + + [[ $LLOG_COUNT -gt 120 ]] && error "CDEBUG_LIMIT not limiting messages ($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" @@ -6342,11 +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 @@ -6437,8 +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) @@ -6447,15 +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 | @@ -7804,7 +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 || @@ -10150,38 +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" @@ -10773,7 +10813,7 @@ test_133b() { #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.*.stats=clear + do_facet ost1 $LCTL set_param obdfilter.*.exports.*.stats=clear $LFS df || error "lfs failed" check_stats $SINGLEMDS "statfs" 1 @@ -10784,12 +10824,20 @@ test_133b() { return 0 sleep 2 do_facet $SINGLEMDS $LCTL set_param mdt.*.md_stats=clear - do_facet ost1 $LCTL set_param obdfilter.*.stats=clear + do_facet ost1 $LCTL set_param obdfilter.*.exports.*.stats=clear df $DIR check_stats $SINGLEMDS "statfs" 1 - res=$(do_facet ost1 \ - $LCTL get_param obdfilter.$FSNAME-OST0000.stats | grep "statfs") - [ "$res" ] && error "OST got STATFS" + + # We want to check that the client didn't send OST_STATFS to + # ost1 but the MDT also uses OST_STATFS for precreate. So some + # extra care is needed here. + if remote_mds; then + local nid=$($LCTL list_nids | head -1 | sed "s/\./\\\./g") + local param="obdfilter.$FSNAME-OST0000.exports.'$nid'.stats" + + res=$(do_facet ost1 $LCTL get_param $param | grep statfs) + [ "$res" ] && error "OST got STATFS" + fi return 0 } @@ -15403,12 +15451,79 @@ 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_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) @@ -15598,18 +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) @@ -15699,36 +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" @@ -16718,7 +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 @@ -16727,14 +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 $dom) == 0 ] || error "bad stripe count" @@ -16746,46 +16860,61 @@ test_270a() { local space_check=1 # Skip free space checks with ZFS - if [ "$(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 @@ -16793,22 +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 $tmp return 0 } @@ -17026,9 +17158,8 @@ test_271a() { lctl set_param -n mdc.*.stats=clear dd if=/dev/zero of=$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 } @@ -17050,9 +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)" @@ -17074,16 +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" @@ -17102,9 +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" @@ -17115,9 +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 ] || @@ -17145,7 +17283,6 @@ test_271d() { $LFS setstripe -E 1024K -L mdt $DIR/$tdir local mdtidx=$($GETSTRIPE -M $DIR/$tdir) - local facet=mds$((mdtidx + 1)) cancel_lru_locks mdc dd if=/dev/urandom of=$tmp bs=1000 count=1 @@ -17158,12 +17295,9 @@ test_271d() { # append data to the same file it should update local page echo "Append to the same page" cat /etc/hosts >> $dom - local num=$(lctl get_param -n mdc.*.stats | - awk '/ost_read/ {print $2}') - local ra=$(lctl get_param -n mdc.*.stats | - awk '/req_active/ {print $2}') - local rw=$(lctl get_param -n mdc.*.stats | - awk '/req_waittime/ {print $2}') + local num=$(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" @@ -17177,12 +17311,9 @@ test_271d() { echo "Open and read file" cat $dom > /dev/null - local num=$(lctl get_param -n mdc.*.stats | - awk '/ost_read/ {print $2}') - local ra=$(lctl get_param -n mdc.*.stats | - awk '/req_active/ {print $2}') - local rw=$(lctl get_param -n mdc.*.stats | - awk '/req_waittime/ {print $2}') + local num=$(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" @@ -17208,7 +17339,6 @@ test_271e() { $LFS setstripe -E 1024K -L mdt $DIR/$tdir local mdtidx=$($GETSTRIPE -M $DIR/$tdir) - local facet=mds$((mdtidx + 1)) cancel_lru_locks mdc dd if=/dev/urandom of=$tmp bs=30K count=1 @@ -17220,12 +17350,9 @@ test_271e() { echo "Append to the same page" cat /etc/hosts >> $dom - local num=$(lctl get_param -n mdc.*.stats | \ - awk '/ost_read/ {print $2}') - local ra=$(lctl get_param -n mdc.*.stats | \ - awk '/req_active/ {print $2}') - local rw=$(lctl get_param -n mdc.*.stats | \ - awk '/req_waittime/ {print $2}') + local num=$(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 @@ -17239,12 +17366,9 @@ test_271e() { echo "Open and read file" cat $dom > /dev/null - local num=$(lctl get_param -n mdc.*.stats | \ - awk '/ost_read/ {print $2}') - local ra=$(lctl get_param -n mdc.*.stats | \ - awk '/req_active/ {print $2}') - local rw=$(lctl get_param -n mdc.*.stats | \ - awk '/req_waittime/ {print $2}') + local num=$(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 @@ -17270,7 +17394,6 @@ test_271f() { $LFS setstripe -E 1024K -L mdt $DIR/$tdir local mdtidx=$($GETSTRIPE -M $DIR/$tdir) - local facet=mds$((mdtidx + 1)) cancel_lru_locks mdc dd if=/dev/urandom of=$tmp bs=200000 count=1 @@ -17281,12 +17404,9 @@ test_271f() { echo "Append to the same page" cat /etc/hosts >> $dom - local num=$(lctl get_param -n mdc.*.stats | \ - awk '/ost_read/ {print $2}') - local ra=$(lctl get_param -n mdc.*.stats | \ - awk '/req_active/ {print $2}') - local rw=$(lctl get_param -n mdc.*.stats | \ - awk '/req_waittime/ {print $2}') + local num=$(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" @@ -17300,12 +17420,9 @@ test_271f() { echo "Open and read file" cat $dom > /dev/null - local num=$(lctl get_param -n mdc.*.stats | \ - awk '/ost_read/ {print $2}') - local ra=$(lctl get_param -n mdc.*.stats | \ - awk '/req_active/ {print $2}') - local rw=$(lctl get_param -n mdc.*.stats | \ - awk '/req_waittime/ {print $2}') + local num=$(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" @@ -18363,7 +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 @@ -18375,30 +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 $tf @@ -18409,8 +18526,8 @@ test_312() { # LU-4856 dd if=/dev/zero of=$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 $zfs_objid) @@ -18426,12 +18543,13 @@ test_313() { remote_ost_nodsh && skip "remote OST with nodsh" local file=$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 $file @@ -18453,8 +18571,9 @@ test_315() { # LU-618 local file=$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 @@ -18960,7 +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 @@ -19170,6 +19289,55 @@ test_415() { } 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 $(version_code 2.9.55) ]] && @@ -19794,6 +19962,65 @@ test_807() { } run_test 807 "verify LSOM syncing tool" +check_som_nologged() +{ + local lines=$($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" + # # tests that do cleanup/setup should be run at the end #