X-Git-Url: https://git.whamcloud.com/?p=fs%2Flustre-release.git;a=blobdiff_plain;f=lustre%2Ftests%2Fsanity.sh;h=6dde8abe33ceafbf3d5db726a5d5658eb2b6cf9f;hp=4f783238c0f430ff99a7abe03e5955bb2da49fcd;hb=76c696d545677bab4ce9641b9eff338fe99f6396;hpb=703195acc2157969e66cb07946522b918dfd7dea diff --git a/lustre/tests/sanity.sh b/lustre/tests/sanity.sh old mode 100644 new mode 100755 index 4f78323..6dde8ab --- a/lustre/tests/sanity.sh +++ b/lustre/tests/sanity.sh @@ -8,13 +8,31 @@ set -e ONLY=${ONLY:-"$*"} -# bug number for skipped test: 13297 2108 9789 3637 9789 3561 12622 5188 -ALWAYS_EXCEPT=" 42a 42b 42c 42d 45 51d 68b $SANITY_EXCEPT" +# bug number for skipped test: 13297 2108 9789 3637 9789 3561 5188 +ALWAYS_EXCEPT=" 42a 42b 42c 42d 45 68b $SANITY_EXCEPT" # UPDATE THE COMMENT ABOVE WITH BUG NUMBERS WHEN CHANGING ALWAYS_EXCEPT! # with LOD/OSP landing -# bug number for skipped tests: LU-2036 -ALWAYS_EXCEPT=" 76 $ALWAYS_EXCEPT" +# bug number for skipped tests: LU-2036 LU-8139 +ALWAYS_EXCEPT=" 76 101g $ALWAYS_EXCEPT" + +is_sles11() # LU-4341 +{ + if [ -r /etc/SuSE-release ] + then + local vers=$(grep VERSION /etc/SuSE-release | awk '{print $3}') + local patchlev=$(grep PATCHLEVEL /etc/SuSE-release | + awk '{ print $3 }') + if [ $vers -eq 11 ] && [ $patchlev -ge 3 ]; then + return 0 + fi + fi + return 1 +} + +if is_sles11; then # LU-4341 + ALWAYS_EXCEPT="$ALWAYS_EXCEPT 170" +fi SRCDIR=$(cd $(dirname $0); echo $PWD) export PATH=$PATH:/sbin @@ -38,7 +56,6 @@ SOCKETCLIENT=${SOCKETCLIENT:-socketclient} MEMHOG=${MEMHOG:-memhog} DIRECTIO=${DIRECTIO:-directio} ACCEPTOR_PORT=${ACCEPTOR_PORT:-988} -UMOUNT=${UMOUNT:-"umount -d"} STRIPES_PER_OBJ=-1 CHECK_GRANT=${CHECK_GRANT:-"yes"} GRANT_CHECK_LIST=${GRANT_CHECK_LIST:-""} @@ -58,11 +75,13 @@ init_test_env $@ . ${CONFIG:=$LUSTRE/tests/cfg/${NAME}.sh} init_logging -[ "$SLOW" = "no" ] && EXCEPT_SLOW="24o 24D 27m 64b 68 71 77f 78 115 124b" +# 5 12 (min)" +[ "$SLOW" = "no" ] && EXCEPT_SLOW="24D 27m 64b 68 71 115 300o" if [ $(facet_fstype $SINGLEMDS) = "zfs" ]; then - # bug number for skipped test: LU-4536 LU-1957 LU-2805 - ALWAYS_EXCEPT="$ALWAYS_EXCEPT 65ic 180 184c" + # bug number for skipped test: LU-4536 LU-1957 + ALWAYS_EXCEPT="$ALWAYS_EXCEPT 65ic 180" + # 4 13 (min)" [ "$SLOW" = "no" ] && EXCEPT_SLOW="$EXCEPT_SLOW 51b 51ba" fi @@ -88,17 +107,11 @@ check_swap_layouts_support() } check_and_setup_lustre - DIR=${DIR:-$MOUNT} assert_DIR MDT0=$($LCTL get_param -n mdc.*.mds_server_uuid | awk '{ gsub(/_UUID/,""); print $1 }' | head -n1) -LOVNAME=$($LCTL get_param -n llite.*.lov.common_name | tail -n 1) -OSTCOUNT=$($LCTL get_param -n lov.$LOVNAME.numobd) -STRIPECOUNT=$($LCTL get_param -n lov.$LOVNAME.stripecount) -STRIPESIZE=$($LCTL get_param -n lov.$LOVNAME.stripesize) -ORIGFREE=$($LCTL get_param -n lov.$LOVNAME.kbytesavail) MAXFREE=${MAXFREE:-$((200000 * $OSTCOUNT))} [ -f $DIR/d52a/foo ] && chattr -a $DIR/d52a/foo @@ -539,11 +552,7 @@ run_test 17l "Ensure lgetxattr's returned xattr size is consistent ========" test_17m() { local short_sym="0123456789" local WDIR=$DIR/${tdir}m - local mds_index - local devname - local cmd local i - local rc=0 remote_mds_nodsh && skip "remote MDS with nodsh" && return [ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.2.0) ] && @@ -576,49 +585,42 @@ test_17m() { echo "recreate the 512 symlink files with a shorter string" for ((i = 0; i < 512; ++i)); do # rewrite the symlink file with a shorter string - ln -sf ${long_sym} $WDIR/long-$i - ln -sf ${short_sym} $WDIR/short-$i + ln -sf ${long_sym} $WDIR/long-$i || error "long_sym failed" + ln -sf ${short_sym} $WDIR/short-$i || error "short_sym failed" done - mds_index=$($LFS getstripe -M $WDIR) - mds_index=$((mds_index+1)) - devname=$(mdsdevname $mds_index) - cmd="$E2FSCK -fnvd $devname" + local mds_index=$(($($LFS getstripe -M $WDIR) + 1)) + local devname=$(mdsdevname $mds_index) - echo "stop and checking mds${mds_index}: $cmd" + echo "stop and checking mds${mds_index}:" # e2fsck should not return error stop mds${mds_index} - do_facet mds${mds_index} $cmd || rc=$? + run_e2fsck $(facet_active_host mds${mds_index}) $devname -n + rc=$? start mds${mds_index} $devname $MDS_MOUNT_OPTS || error "start failed" df $MOUNT > /dev/null 2>&1 - [ $rc -ne 0 ] && error "e2fsck should not report error upon "\ - "short/long symlink MDT: rc=$rc" - return $rc + [ $rc -eq 0 ] || + error "e2fsck detected error for short/long symlink: rc=$rc" } run_test 17m "run e2fsck against MDT which contains short/long symlink" check_fs_consistency_17n() { local mdt_index - local devname - local cmd local rc=0 # create/unlink in 17n only change 2 MDTs(MDT1/MDT2), # so it only check MDT1/MDT2 instead of all of MDTs. - for mdt_index in $(seq 1 2); do - devname=$(mdsdevname $mdt_index) - cmd="$E2FSCK -fnvd $devname" - - echo "stop and checking mds${mdt_index}: $cmd" + for mdt_index in 1 2; do + local devname=$(mdsdevname $mdt_index) # e2fsck should not return error stop mds${mdt_index} - do_facet mds${mdt_index} $cmd || rc=$? + run_e2fsck $(facet_active_host mds$mdt_index) $devname -n || + rc=$((rc + $?)) start mds${mdt_index} $devname $MDS_MOUNT_OPTS || - error "mount mds${mdt_index} failed" + error "mount mds$mdt_index failed" df $MOUNT > /dev/null 2>&1 - [ $rc -ne 0 ] && break done return $rc } @@ -1015,7 +1017,9 @@ test_24u() { # bug12192 run_test 24u "create stripe file" page_size() { - getconf PAGE_SIZE + local size + size=$(getconf PAGE_SIZE 2>/dev/null) + echo -n ${size:-4096} } simple_cleanup_common() { @@ -1080,6 +1084,10 @@ run_test 24w "Reading a file larger than 4Gb" test_24x() { [[ $MDSCOUNT -lt 2 ]] && skip "needs >= 2 MDTs" && return + + [[ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.7.56) ]] && + skip "Need MDS version at least 2.7.56" && return + [ $PARALLEL == "yes" ] && skip "skip parallel run" && return local MDTIDX=1 local remote_dir=$DIR/$tdir/remote_dir @@ -1425,6 +1433,9 @@ run_test 27l "check setstripe permissions (should return error)" test_27m() { [[ $OSTCOUNT -lt 2 ]] && skip_env "$OSTCOUNT < 2 OSTs -- skipping" && return + + ORIGFREE=$($LCTL get_param -n lov.$FSNAME-clilov-*.kbytesavail | + head -n1) if [[ $ORIGFREE -gt $MAXFREE ]]; then skip "$ORIGFREE > $MAXFREE skipping out-of-space test on OST0" return @@ -1844,7 +1855,7 @@ check_seq_oid() log "want: stripe:$stripe_nr ost:$obdidx oid:$oid/$hex seq:$seq" seq=$(echo $seq | sed -e "s/^0x//g") - if [ $seq == 0 ]; then + if [ $seq == 0 ] || [ $(facet_fstype ost$ost) == zfs ]; then oid_hex=$(echo $oid) else oid_hex=$(echo $hex | sed -e "s/^0x//g") @@ -1868,6 +1879,7 @@ check_seq_oid() $(facet_mntpt ost$ost)/$obj_file) unmount_fstype ost$ost start ost$ost $dev $OST_MOUNT_OPTS + clients_up fi [ -z "$ff" ] && error "$obj_file: no filter_fid info" @@ -2011,6 +2023,8 @@ run_test 27C "check full striping across all OSTs" test_27D() { [ $OSTCOUNT -lt 2 ] && skip "needs >= 2 OSTs" && return + [ -n "$FILESET" ] && skip "SKIP due to FILESET set" && return + remote_mds_nodsh && skip "remote MDS with nodsh" && return local POOL=${POOL:-testpool} local first_ost=0 local last_ost=$(($OSTCOUNT - 1)) @@ -2021,8 +2035,13 @@ test_27D() { test_mkdir -p $DIR/$tdir pool_add $POOL || error "pool_add failed" pool_add_targets $POOL $ost_range || error "pool_add_targets failed" - llapi_layout_test -d$DIR/$tdir -p$POOL -o$OSTCOUNT || + + local skip27D + [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.7.64) ] && + skip27D = "-s 29" + llapi_layout_test -d$DIR/$tdir -p$POOL -o$OSTCOUNT $skip27D || error "llapi_layout_test failed" + cleanup_pools || error "cleanup_pools failed" } run_test 27D "validate llapi_layout API" @@ -2031,6 +2050,8 @@ run_test 27D "validate llapi_layout API" # accessing a widely striped file. test_27E() { [ $OSTCOUNT -lt 2 ] && skip "needs >= 2 OSTs" && return + [ $(lustre_version_code client) -lt $(version_code 2.5.57) ] && + skip "client does not have LU-3338 fix" && return # 72 bytes is the minimum space required to store striping # information for a file striped across one OST: @@ -2056,6 +2077,46 @@ test_27E() { } run_test 27E "check that default extended attribute size properly increases" +test_27F() { # LU-5346/LU-7975 + + [ $PARALLEL == "yes" ] && skip "skip parallel run" && return + + [[ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.8.51) ]] && + skip "Need MDS version at least 2.8.51" && return + + test_mkdir -p $DIR/$tdir + rm -f $DIR/$tdir/f0 + $SETSTRIPE -c 2 $DIR/$tdir + + # stop all OSTs to reproduce situation for LU-7975 ticket + for num in $(seq $OSTCOUNT); do + stop ost$num + done + + # open/create f0 with O_LOV_DELAY_CREATE + # truncate f0 to a non-0 size + # close + multiop $DIR/$tdir/f0 oO_RDWR:O_CREAT:O_LOV_DELAY_CREATE:T1050000c + + $CHECKSTAT -s 1050000 $DIR/$tdir/f0 || error "checkstat failed" + # open/write it again to force delayed layout creation + cat /etc/hosts > $DIR/$tdir/f0 & + catpid=$! + + # restart OSTs + for num in $(seq $OSTCOUNT); do + start ost$num $(ostdevname $num) $OST_MOUNT_OPTS || + error "ost$num failed to start" + done + + wait $catpid || error "cat failed" + + cmp /etc/hosts $DIR/$tdir/f0 || error "cmp failed" + [[ $($GETSTRIPE -c $DIR/$tdir/f0) == 2 ]] || error "wrong stripecount" + +} +run_test 27F "Client resend delayed layout creation with non-zero size" + # createtest also checks that device nodes are created and # then visible correctly (#2091) test_28() { # bug 2091 @@ -2065,8 +2126,9 @@ test_28() { # bug 2091 run_test 28 "create/mknod/mkdir with bad file types ============" test_29() { - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return - cancel_lru_locks mdc + [ $PARALLEL == "yes" ] && skip "skip parallel run" && return 0 + sync; sleep 1; sync # flush out any dirty pages from previous tests + cancel_lru_locks test_mkdir $DIR/d29 touch $DIR/d29/foo log 'first d29' @@ -2076,7 +2138,7 @@ test_29() { for lock_count in $(lctl get_param -n ldlm.namespaces.*mdc*.lock_count); do let LOCKCOUNTORIG=$LOCKCOUNTORIG+$lock_count done - [ $LOCKCOUNTORIG -eq 0 ] && echo "No mdc lock count" && return 1 + [ $LOCKCOUNTORIG -eq 0 ] && error "No mdc lock count" && return 1 declare -i LOCKUNUSEDCOUNTORIG=0 for unused_count in $(lctl get_param -n ldlm.namespaces.*mdc*.lock_unused_count); do @@ -2347,8 +2409,13 @@ test_31p() { run_test 31p "remove of open striped directory" cleanup_test32_mount() { + local rc=0 trap 0 - $UMOUNT -d $DIR/$tdir/ext2-mountpoint + local loopdev=$(losetup -a | grep $EXT2_DEV | sed -ne 's/:.*$/p') + $UMOUNT $DIR/$tdir/ext2-mountpoint || rc=$? + losetup -d $loopdev || true + rm -rf $DIR/$tdir/ext2-mountpoint + return $rc } test_32a() { @@ -2554,8 +2621,13 @@ test_32p() { run_test 32p "open d32p/symlink->tmp/symlink->lustre-root/$tfile" cleanup_testdir_mount() { + local rc=0 trap 0 - $UMOUNT -d $DIR/$tdir + local loopdev=$(losetup -a | grep $EXT2_DEV | sed -ne 's/:.*$/p') + $UMOUNT $DIR/$tdir || rc=$? + losetup -d $loopdev || true + rm -rf $DIR/$tdir + return $rc } test_32q() { @@ -2698,8 +2770,7 @@ test_33d() { error "create" || true $RUNAS $OPENFILE -f O_RDWR:O_CREAT -m 0444 $remote_dir/f33 && error "open RDWR" || true - $RUNAS $OPENFILE -f 1286739555 $remote_dir/f33 && - error "create" || true + $RUNAS $OPENFILE -f 1286739555 $remote_dir/f33 || true } run_test 33d "openfile with 444 modes and malformed flags under remote dir" @@ -2756,6 +2827,7 @@ cleanup_33f() { test_33f() { [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return + remote_mds_nodsh && skip "remote MDS with nodsh" && return mkdir $DIR/$tdir chmod go+rwx $DIR/$tdir @@ -3324,29 +3396,27 @@ test_39l() { # test setting directory atime to future touch -a -d @$TEST_39_ATIME $DIR/$tdir local atime=$(stat -c %X $DIR/$tdir) - [ "$atime" = $TEST_39_ATIME ] || \ + [ "$atime" = $TEST_39_ATIME ] || error "atime is not set to future: $atime, $TEST_39_ATIME" # test setting directory atime from future to now - local d1=$(date +%s) - ls $DIR/$tdir - local d2=$(date +%s) + local now=$(date +%s) + touch -a -d @$now $DIR/$tdir - cancel_lru_locks mdc atime=$(stat -c %X $DIR/$tdir) - [ "$atime" -ge "$d1" -a "$atime" -le "$d2" ] || \ - error "atime is not updated from future: $atime, $d1 dir atime + atime_diff - d1=$(date +%s) + local d1=$(date +%s) ls $DIR/$tdir - d2=$(date +%s) + local d2=$(date +%s) cancel_lru_locks mdc atime=$(stat -c %X $DIR/$tdir) - [ "$atime" -ge "$d1" -a "$atime" -le "$d2" ] || \ + [ "$atime" -ge "$d1" -a "$atime" -le "$d2" ] || error "atime is not updated : $atime, should be $d2" do_facet $SINGLEMDS lctl set_param -n mdd.*MDT0000*.atime_diff=60 @@ -3356,7 +3426,7 @@ test_39l() { ls $DIR/$tdir cancel_lru_locks mdc atime=$(stat -c %X $DIR/$tdir) - [ "$atime" -ge "$d1" -a "$atime" -le "$d2" ] || \ + [ "$atime" -ge "$d1" -a "$atime" -le "$d2" ] || error "atime is updated to $atime, should remain $d1 /dev/null && error "'ls .' worked after removing cwd" ls .. > /dev/null || error "'ls ..' failed after removing cwd" - is_patchless || ( cd . && error "'cd .' worked after removing cwd" ) test_mkdir . && error "'mkdir .' worked after removing cwd" rmdir . && error "'rmdir .' worked after removing cwd" ln -s . foo && error "'ln -s .' worked after removing cwd" @@ -3915,8 +3997,6 @@ test_48c() { # bug 2350 test_mkdir .foo && error "mkdir .foo worked after removing cwd" $TRACE ls . && error "'ls .' worked after removing cwd" $TRACE ls .. || error "'ls ..' failed after removing cwd" - is_patchless || ( $TRACE cd . && - error "'cd .' worked after removing cwd" ) $TRACE test_mkdir . && error "'mkdir .' worked after removing cwd" $TRACE rmdir . && error "'rmdir .' worked after removing cwd" $TRACE ln -s . foo && error "'ln -s .' worked after removing cwd" @@ -3938,13 +4018,10 @@ test_48d() { # bug 2350 test_mkdir .foo && error "mkdir .foo worked after removing parent" $TRACE ls . && error "'ls .' worked after removing parent" $TRACE ls .. && error "'ls ..' worked after removing parent" - is_patchless || ( $TRACE cd . && - error "'cd .' worked after recreate parent" ) $TRACE test_mkdir . && error "'mkdir .' worked after removing parent" $TRACE rmdir . && error "'rmdir .' worked after removing parent" $TRACE ln -s . foo && error "'ln -s .' worked after removing parent" - is_patchless || ( $TRACE cd .. && - error "'cd ..' worked after removing parent" || true ) + true } run_test 48d "Access removed parent subdir (should return errors)" @@ -3968,6 +4045,7 @@ run_test 48e "Access to recreated parent subdir (should return errors)" test_49() { # LU-1030 [ $PARALLEL == "yes" ] && skip "skip parallel run" && return + remote_ost_nodsh && skip "remote OST with nodsh" && return # get ost1 size - lustre-OST0000 ost1_size=$(do_facet ost1 $LFS df | grep ${ost1_svc} | awk '{ print $4 }') @@ -4019,87 +4097,62 @@ test_51a() { # was test_51 } run_test 51a "special situations: split htree with empty entry ==" -export NUMTEST=70000 test_51b() { [ $PARALLEL == "yes" ] && skip "skip parallel run" && return - local BASE=$DIR/d${base}.${TESTSUITE} + local dir=$DIR/$tdir + + local nrdirs=$((65536 + 100)) # cleanup the directory - rm -fr $BASE + rm -fr $dir - test_mkdir -p -c1 $BASE + test_mkdir -p -c1 $dir $LFS df $LFS df -i - local mdtidx=$(printf "%04x" $($LFS getstripe -M $BASE)) + local mdtidx=$(printf "%04x" $($LFS getstripe -M $dir)) local numfree=$(lctl get_param -n mdc.$FSNAME-MDT$mdtidx*.filesfree) - [[ $numfree -lt 21000 ]] && + [[ $numfree -lt $nrdirs ]] && skip "not enough free inodes ($numfree) on MDT$mdtidx" && return - [[ $numfree -lt $NUMTEST ]] && NUMTEST=$(($numfree - 50)) && - echo "reduced count to $NUMTEST due to inodes on MDT$mdtidx" - # need to check free space for the directories as well local blkfree=$(lctl get_param -n mdc.$FSNAME-MDT$mdtidx*.kbytesavail) numfree=$((blkfree / 4)) - [[ $numfree -lt $NUMTEST ]] && NUMTEST=$((numfree - 50)) && - echo "reduced count to $NUMTEST due to blocks on MDT$mdtidx" - - createmany -d $BASE/d $NUMTEST && echo $NUMTEST > $BASE/fnum || - { - $LFS df - $LFS df -i - echo "failed" > $BASE/fnum - error "failed to create $NUMTEST subdirs in MDT$mdtidx:$BASE" - } -} -run_test 51b "exceed 64k subdirectory nlink limit" + [[ $numfree -lt $nrdirs ]] && skip "not enough blocks ($numfree)" && + return -test_51ba() { # LU-993 - local BASE=$DIR/d${base}.${TESTSUITE} - # unlink all but 100 subdirectories, then check it still works - local LEFT=100 - [ -f $BASE/fnum ] && local NUMPREV=$(cat $BASE/fnum) && rm $BASE/fnum + # create files + createmany -d $dir/d $nrdirs || + error "failed to create $nrdirs subdirs in MDT$mdtidx:$dir" - [ "$NUMPREV" != "failed" ] && NUMTEST=$NUMPREV - local DELETE=$((NUMTEST - LEFT)) + # really created : + nrdirs=$(ls -U $dir | wc -l) - # continue on to run this test even if 51b didn't finish, - # just to delete the many subdirectories created. - [ ! -d "${BASE}/d1" ] && skip "test_51b() not run" && return 0 + # unlink all but 100 subdirectories, then check it still works + local left=100 + local delete=$((nrdirs - left)) # for ldiskfs the nlink count should be 1, but this is OSD specific # and so this is listed for informational purposes only - echo "nlink before: $(stat -c %h $BASE), created before: $NUMTEST" - unlinkmany -d $BASE/d $DELETE - RC=$? - - if [ $RC -ne 0 ]; then - if [ "$NUMPREV" == "failed" ]; then - skip "previous setup failed" - return 0 - else - error "unlink of first $DELETE subdirs failed" - return $RC - fi - fi + echo "nlink before: $(stat -c %h $dir), created before: $nrdirs" + unlinkmany -d $dir/d $delete || + error "unlink of first $delete subdirs failed" - echo "nlink between: $(stat -c %h $BASE)" - # trim the first line of ls output - local FOUND=$(($(ls -l ${BASE} | wc -l) - 1)) - [ $FOUND -ne $LEFT ] && - error "can't find subdirs: found only $FOUND/$LEFT" + echo "nlink between: $(stat -c %h $dir)" + local found=$(ls -U $dir | wc -l) + [ $found -ne $left ] && + error "can't find subdirs: found only $found, expected $left" - unlinkmany -d $BASE/d $DELETE $LEFT || - error "unlink of second $LEFT subdirs failed" + unlinkmany -d $dir/d $delete $left || + error "unlink of second $left subdirs failed" # regardless of whether the backing filesystem tracks nlink accurately # or not, the nlink count shouldn't be more than "." and ".." here - local AFTER=$(stat -c %h $BASE) - [[ $AFTER -gt 2 ]] && error "nlink after: $AFTER > 2" || - echo "nlink after: $AFTER" + local after=$(stat -c %h $dir) + [[ $after -gt 2 ]] && error "nlink after: $after > 2" || + echo "nlink after: $after" } -run_test 51ba "verify nlink for many subdirectory cleanup" +run_test 51b "exceed 64k subdirectory nlink limit on create, verify unlink" test_51d() { [ $PARALLEL == "yes" ] && skip "skip parallel run" && return @@ -4107,11 +4160,11 @@ test_51d() { skip_env "skipping test with few OSTs" && return test_mkdir -p $DIR/$tdir createmany -o $DIR/$tdir/t- 1000 - $GETSTRIPE $DIR/$tdir > $TMP/files + $GETSTRIPE $DIR/$tdir > $TMP/$tfile for N in $(seq 0 $((OSTCOUNT - 1))); do OBJS[$N]=$(awk -vobjs=0 '($1 == '$N') { objs += 1 } \ - END { printf("%0.0f", objs) }' $TMP/files) - OBJS0[$N]=$(grep -A 1 idx $TMP/files | awk -vobjs=0 \ + END { printf("%0.0f", objs) }' $TMP/$tfile) + OBJS0[$N]=$(grep -A 1 idx $TMP/$tfile | awk -vobjs=0 \ '($1 == '$N') { objs += 1 } \ END { printf("%0.0f", objs) }') log "OST$N has ${OBJS[$N]} objects, ${OBJS0[$N]} are index 0" @@ -4135,8 +4188,9 @@ test_51d() { " (${OBJS0[$N]} < ${OBJS0[$NLAST]}" NLAST=$N done + rm -f $TMP/$tfile } -run_test 51d "check object distribution ====================" +run_test 51d "check object distribution" test_51e() { if [ "$(facet_fstype $SINGLEMDS)" != ldiskfs ]; then @@ -4170,10 +4224,10 @@ test_52a() { lsattr $DIR/$tdir/foo | egrep -q "^-+a[-e]+ $DIR/$tdir/foo" || error "lsattr" chattr -a $DIR/$tdir/foo || error "chattr -a failed" - cp -r $DIR/$tdir /tmp/ - rm -fr $DIR/$tdir || error "cleanup rm failed" + cp -r $DIR/$tdir $TMP/ + rm -fr $DIR/$tdir $TMP/$tdir || error "cleanup rm failed" } -run_test 52a "append-only flag test (should return errors) =====" +run_test 52a "append-only flag test (should return errors)" test_52b() { [ -f $DIR/$tdir/foo ] && chattr -i $DIR/$tdir/foo @@ -4294,10 +4348,11 @@ find_loop_dev() { } cleanup_54c() { + local rc=0 loopdev="$DIR/loop54c" trap 0 - $UMOUNT -d $DIR/$tdir || rc=$? + $UMOUNT $DIR/$tdir || rc=$? losetup -d $loopdev || true losetup -d $LOOPDEV || true rm -rf $loopdev $DIR/$tfile $DIR/$tdir @@ -4404,6 +4459,26 @@ test_56a() { # was test_56 } run_test 56a "check $GETSTRIPE" +test_56b() { + test_mkdir $DIR/$tdir + NUMDIRS=3 + for i in $(seq 1 $NUMDIRS); do + test_mkdir $DIR/$tdir/dir$i + done + + # test lfs getdirstripe default mode is non-recursion, which is + # different from lfs getstripe + dircnt=$($LFS getdirstripe $DIR/$tdir | grep -c lmv_stripe_count) + [[ $dircnt -eq 1 ]] || + error "$LFS getdirstripe: found $dircnt, not 1" + dircnt=$($LFS getdirstripe --recursive $DIR/$tdir | + grep -c lmv_stripe_count) + [[ $dircnt -eq $((NUMDIRS + 1)) ]] || + error "$LFS getdirstripe --recursive: found $dircnt, \ + not $((NUMDIRS + 1))" +} +run_test 56b "check $LFS getdirstripe" + NUMFILES=3 NUMDIRS=3 setup_56() { @@ -4842,7 +4917,9 @@ test_56w() { local file_size=$((stripe_size * OSTCOUNT)) local file_num=$((NUMDIRS * NUMFILES + NUMFILES)) local required_space=$((file_num * file_size)) - local free_space=$($LCTL get_param -n lov.$LOVNAME.kbytesavail) + + local free_space=$($LCTL get_param -n lov.$FSNAME-clilov-*.kbytesavail | + head -n1) [[ $free_space -le $((required_space / 1024)) ]] && skip_env "need at least $required_space bytes free space," \ "have $free_space kbytes" && return @@ -4950,6 +5027,29 @@ test_56x() { } run_test 56x "lfs migration support" +test_56xa() { + check_swap_layouts_support && return 0 + [[ $OSTCOUNT -lt 2 ]] && + skip_env "need 2 OST, skipping test" && return + + local dir0=$DIR/$tdir/$testnum + test_mkdir -p $dir0 || error "creating dir $dir0" + + local ref1=/etc/passwd + local file1=$dir0/file1 + + $SETSTRIPE -c 2 $file1 + cp $ref1 $file1 + $LFS migrate --block -c 1 $file1 || error "migrate failed rc = $?" + local stripe=$($GETSTRIPE -c $file1) + [[ $stripe == 1 ]] || error "stripe of $file1 is $stripe != 1" + cmp $file1 $ref1 || error "content mismatch $file1 differs from $ref1" + + # clean up + rm -f $file1 +} +run_test 56xa "lfs migration --block support" + test_56y() { [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.4.53) ] && skip "No HSM $(lustre_build_version $SINGLEMDS) MDS < 2.4.53" && @@ -5213,19 +5313,19 @@ run_test 60a "llog_test run from kernel module and test llog_reader ==========" test_60b() { # bug 6411 [ $PARALLEL == "yes" ] && skip "skip parallel run" && return 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 - }"` - [[ $LLOG_COUNT -gt 50 ]] && + LLOG_COUNT=$(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 + }") + [[ $LLOG_COUNT -gt 100 ]] && error "CDEBUG_LIMIT not limiting messages ($LLOG_COUNT)" || true } run_test 60b "limit repeated messages from CERROR/CWARN ========" @@ -5261,6 +5361,16 @@ test_60d() { } run_test 60d "test printk console message masking" +test_60e() { + [ $PARALLEL == "yes" ] && skip "skip parallel run" && return + remote_mds_nodsh && skip "remote MDS with nodsh" && return + touch $DIR/$tfile +#define OBD_FAIL_MDS_LLOG_CREATE_FAILED2 0x15b + do_facet mds1 lctl set_param fail_loc=0x15b + rm $DIR/$tfile +} +run_test 60e "no space while new llog is being created" + test_61() { [ $PARALLEL == "yes" ] && skip "skip parallel run" && return f="$DIR/f61" @@ -5356,28 +5466,35 @@ run_test 65a "directory with no stripe info ====================" test_65b() { [ $PARALLEL == "yes" ] && skip "skip parallel run" && return test_mkdir -p $DIR/$tdir + local STRIPESIZE=$($GETSTRIPE -S $DIR/$tdir) + $SETSTRIPE -S $((STRIPESIZE * 2)) -i 0 -c 1 $DIR/$tdir || error "setstripe" touch $DIR/$tdir/f2 $LVERIFY $DIR/$tdir $DIR/$tdir/f2 || error "lverify failed" } -run_test 65b "directory setstripe -S $((STRIPESIZE * 2)) -i 0 -c 1" +run_test 65b "directory setstripe -S stripe_size*2 -i 0 -c 1" test_65c() { [ $PARALLEL == "yes" ] && skip "skip parallel run" && return if [[ $OSTCOUNT -gt 1 ]]; then test_mkdir -p $DIR/$tdir + local STRIPESIZE=$($GETSTRIPE -S $DIR/$tdir) + $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" fi } -run_test 65c "directory setstripe -S $((STRIPESIZE*4)) -i 1 -c $((OSTCOUNT-1))" +run_test 65c "directory setstripe -S stripe_size*4 -i 1 -c $((OSTCOUNT-1))" test_65d() { [ $PARALLEL == "yes" ] && skip "skip parallel run" && return test_mkdir -p $DIR/$tdir + local STRIPECOUNT=$($GETSTRIPE -c $DIR/$tdir) + local STRIPESIZE=$($GETSTRIPE -S $DIR/$tdir) + if [[ $STRIPECOUNT -le 0 ]]; then sc=1 elif [[ $STRIPECOUNT -gt 2000 ]]; then @@ -5391,7 +5508,7 @@ test_65d() { $LVERIFY $DIR/$tdir $DIR/$tdir/f4 $DIR/$tdir/f5 || error "lverify failed" } -run_test 65d "directory setstripe -S $STRIPESIZE -c stripe_count" +run_test 65d "directory setstripe -S stripe_size -c stripe_count" test_65e() { [ $PARALLEL == "yes" ] && skip "skip parallel run" && return @@ -5415,6 +5532,8 @@ run_test 65f "dir setstripe permission (should return error) ===" test_65g() { [ $PARALLEL == "yes" ] && skip "skip parallel run" && return test_mkdir -p $DIR/$tdir + local STRIPESIZE=$($GETSTRIPE -S $DIR/$tdir) + $SETSTRIPE -S $((STRIPESIZE * 2)) -i 0 -c 1 $DIR/$tdir || error "setstripe" $SETSTRIPE -d $DIR/$tdir || error "setstripe" @@ -5426,6 +5545,8 @@ run_test 65g "directory setstripe -d ===========================" test_65h() { [ $PARALLEL == "yes" ] && skip "skip parallel run" && return test_mkdir -p $DIR/$tdir + local STRIPESIZE=$($GETSTRIPE -S $DIR/$tdir) + $SETSTRIPE -S $((STRIPESIZE * 2)) -i 0 -c 1 $DIR/$tdir || error "setstripe" test_mkdir -p $DIR/$tdir/dd1 @@ -5475,34 +5596,59 @@ test_65k() { # bug11679 [[ $OSTCOUNT -lt 2 ]] && skip_env "too few OSTs" && return remote_mds_nodsh && skip "remote MDS with nodsh" && return - echo "Check OST status: " - local MDS_OSCS=`do_facet $SINGLEMDS lctl dl | - awk '/[oO][sS][cC].*md[ts]/ { print $4 }'` + local disable_precreate=true + [ $(lustre_version_code $SINGLEMDS) -le $(version_code 2.8.54) ] && + disable_precreate=false - for OSC in $MDS_OSCS; do - echo $OSC "is activate" - do_facet $SINGLEMDS lctl --device %$OSC activate - done + echo "Check OST status: " + local MDS_OSCS=$(do_facet $SINGLEMDS lctl dl | + awk '/[oO][sS][cC].*md[ts]/ { print $4 }') - mkdir -p $DIR/$tdir - for INACTIVE_OSC in $MDS_OSCS; do - echo "Deactivate: " $INACTIVE_OSC - do_facet $SINGLEMDS lctl --device %$INACTIVE_OSC deactivate - for STRIPE_OSC in $MDS_OSCS; do - OST=`osc_to_ost $STRIPE_OSC` - IDX=`do_facet $SINGLEMDS lctl get_param -n lov.*md*.target_obd | - awk -F: /$OST/'{ print $1 }' | head -n 1` - - [ -f $DIR/$tdir/$IDX ] && continue - echo "$SETSTRIPE -i $IDX -c 1 $DIR/$tdir/$IDX" - $SETSTRIPE -i $IDX -c 1 $DIR/$tdir/$IDX - RC=$? - [ $RC -ne 0 ] && error "setstripe should have succeeded" - done - rm -f $DIR/$tdir/* - echo $INACTIVE_OSC "is Activate." - do_facet $SINGLEMDS lctl --device %$INACTIVE_OSC activate - done + for OSC in $MDS_OSCS; do + echo $OSC "is active" + do_facet $SINGLEMDS lctl --device %$OSC activate + done + + for INACTIVE_OSC in $MDS_OSCS; do + local ost=$(osc_to_ost $INACTIVE_OSC) + local ostnum=$(do_facet $SINGLEMDS lctl get_param -n \ + lov.*md*.target_obd | + awk -F: /$ost/'{ print $1 }' | head -n 1) + + mkdir -p $DIR/$tdir + $SETSTRIPE -i $ostnum -c 1 $DIR/$tdir + createmany -o $DIR/$tdir/$tfile.$ostnum. 1000 + + echo "Deactivate: " $INACTIVE_OSC + do_facet $SINGLEMDS lctl --device %$INACTIVE_OSC deactivate + + local count=$(do_facet $SINGLEMDS "lctl get_param -n \ + osp.$ost*MDT0000.create_count") + local max_count=$(do_facet $SINGLEMDS "lctl get_param -n \ + osp.$ost*MDT0000.max_create_count") + $disable_precreate && + do_facet $SINGLEMDS "lctl set_param -n \ + osp.$ost*MDT0000.max_create_count=0" + + for idx in $(seq 0 $((OSTCOUNT - 1))); do + [ -f $DIR/$tdir/$idx ] && continue + echo "$SETSTRIPE -i $idx -c 1 $DIR/$tdir/$idx" + $SETSTRIPE -i $idx -c 1 $DIR/$tdir/$idx || + error "setstripe $idx should succeed" + rm -f $DIR/$tdir/$idx || error "rm $idx failed" + done + unlinkmany $DIR/$tdir/$tfile.$ostnum. 1000 + rmdir $DIR/$tdir + + do_facet $SINGLEMDS "lctl set_param -n \ + osp.$ost*MDT0000.max_create_count=$max_count" + do_facet $SINGLEMDS "lctl set_param -n \ + osp.$ost*MDT0000.create_count=$count" + do_facet $SINGLEMDS lctl --device %$INACTIVE_OSC activate + echo $INACTIVE_OSC "is Activate" + + wait_osc_import_state mds ost$ostnum FULL + done } run_test 65k "validate manual striping works properly with deactivated OSCs" @@ -5514,6 +5660,12 @@ test_65l() { # bug 12836 } run_test 65l "lfs find on -1 stripe dir ========================" +test_65m() { + $RUNAS $SETSTRIPE -c 2 $MOUNT && error "setstripe should fail" + true +} +run_test 65m "normal user can't set filesystem default stripe" + # bug 2543 - update blocks count on client test_66() { [ $PARALLEL == "yes" ] && skip "skip parallel run" && return @@ -5624,7 +5776,7 @@ test_68b() { # was test_68 run_test 68b "support swapping to Lustre ========================" # bug5265, obdfilter oa2dentry return -ENOENT -# #define OBD_FAIL_OST_ENOENT 0x217 +# #define OBD_FAIL_SRV_ENOENT 0x217 test_69() { [ $PARALLEL == "yes" ] && skip "skip parallel run" && return remote_ost_nodsh && skip "remote OST with nodsh" && return @@ -5799,8 +5951,8 @@ set_inode_slab_tunables() { test_76() { # Now for bug 20433, added originally in bug 1443 [ $PARALLEL == "yes" ] && skip "skip parallel run" && return - local SLAB_SETTINGS=`get_inode_slab_tunables` - local CPUS=`getconf _NPROCESSORS_ONLN` + local SLAB_SETTINGS=$(get_inode_slab_tunables) + local CPUS=$(getconf _NPROCESSORS_ONLN 2>/dev/null) # we cannot set limit below 1 which means 1 inode in each # per-cpu cache is still allowed set_inode_slab_tunables "1 1 0" @@ -5817,7 +5969,7 @@ test_76() { # Now for bug 20433, added originally in bug 1443 AFTER_INODES=$(num_inodes) echo "after inodes: $AFTER_INODES" local wait=0 - while [[ $((AFTER_INODES-1*CPUS)) -gt $BEFORE_INODES ]]; do + while [[ $((AFTER_INODES-1*${CPUS:-1})) -gt $BEFORE_INODES ]]; do sleep 2 AFTER_INODES=$(num_inodes) wait=$((wait+2)) @@ -6003,6 +6155,11 @@ run_test 77j "client only supporting ADLER32" rm -f $F77_TMP unset F77_TMP +cleanup_test_78() { + trap 0 + rm -f $DIR/$tfile +} + test_78() { # bug 10901 [ $PARALLEL == "yes" ] && skip "skip parallel run" && return remote_ost || { skip_env "local OST" && return; } @@ -6032,6 +6189,8 @@ test_78() { # bug 10901 [[ $SMALLESTOST -lt 10240 ]] && skip "too small OSTSIZE, useless to run large O_DIRECT test" && return 0 + trap cleanup_test_78 EXIT + [[ $F78SIZE -gt $((SMALLESTOST * $OSTCOUNT / 1024 - 80)) ]] && F78SIZE=$((SMALLESTOST * $OSTCOUNT / 1024 - 80)) @@ -6044,7 +6203,7 @@ test_78() { # bug 10901 $DIRECTIO rdwr $DIR/$tfile 0 $FSIZE 1048576||error "rdwr failed" done - rm -f $DIR/$tfile + cleanup_test_78 } run_test 78 "handle large O_DIRECT writes correctly ============" @@ -6186,7 +6345,7 @@ test_99b() { # some versions of cvs import exit(1) when asked to import links or # files they can't read. ignore those files. TOIGNORE=$(find . -type l -printf '-I %f\n' -o \ - ! -perm +4 -printf '-I %f\n') + ! -perm /4 -printf '-I %f\n') $RUNAS cvs -d $DIR/d99cvsroot import -m "nomesg" $TOIGNORE \ d99reposname vtag rtag } @@ -6523,41 +6682,88 @@ test_101e() { } run_test 101e "check read-ahead for small read(1k) for small files(500k)" -cleanup_test101f() { - trap 0 - $LCTL set_param -n llite.*.max_read_ahead_whole_mb $MAX_WHOLE_MB - rm -rf $DIR/$tfile 2>/dev/null +test_101f() { + which iozone || { skip "no iozone installed" && return; } + + # create a test file + iozone -i 0 -+n -r 1m -s 128m -w -f $DIR/$tfile > /dev/null 2>&1 + + echo Cancel LRU locks on lustre client to flush the client cache + cancel_lru_locks osc + + echo Reset readahead stats + $LCTL set_param -n llite.*.read_ahead_stats 0 + + echo mmap read the file with small block size + iozone -i 1 -+n -r 32k -s 128m -B -f $DIR/$tfile > /dev/null 2>&1 + + echo checking missing pages + local miss=$($LCTL get_param -n llite.*.read_ahead_stats | + get_named_value 'misses' | cut -d" " -f1 | calc_total) + + [ $miss -lt 3 ] || error "misses too much pages!" + rm -f $DIR/$tfile } +run_test 101f "check mmap read performance" -test_101f() { - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return - local file=$DIR/$tfile - local nreads=1000 +test_101g() { + local rpcs + local osts=$(get_facets OST) + local list=$(comma_list $(osts_nodes)) + local p="$TMP/$TESTSUITE-$TESTNAME.parameters" - MAX_WHOLE_MB=$($LCTL get_param -n llite.*.max_read_ahead_whole_mb) - $LCTL set_param -n llite.*.max_read_ahead_whole_mb 2 - dd if=/dev/zero of=${file} bs=2097152 count=1 2>/dev/null - trap cleanup_test101f EXIT + save_lustre_params $osts "obdfilter.*.brw_size" > $p - echo Cancel LRU locks on lustre client to flush the client cache - cancel_lru_locks osc + $LFS setstripe -c 1 $DIR/$tfile + + if [ $(lustre_version_code ost1) -ge $(version_code 2.8.52) ]; then + set_osd_param $list '' brw_size 16M + + echo "remount client to enable large RPC size" + remount_client $MOUNT || error "remount_client failed" - echo Reset readahead stats - $LCTL set_param -n llite.*.read_ahead_stats 0 - # Random read in a 2M file, because max_read_ahead_whole_mb = 2M, - # readahead should read in 2M file on second read, so only miss - # 2 pages. - echo Random 4K reads on 2M file for 1000 times - $READS -f $file -s 2097152 -b 4096 -n $nreads + for mp in $($LCTL get_param -n osc.*.max_pages_per_rpc); do + [ "$mp" -eq 4096 ] || + error "max_pages_per_rpc not correctly set" + done + + $LCTL set_param -n osc.*.rpc_stats=0 + + # 10*16 MiB should be enough for the test + dd if=/dev/zero of=$DIR/$tfile bs=16M count=10 + cancel_lru_locks osc + dd of=/dev/null if=$DIR/$tfile bs=16M count=10 + + # calculate 16 MiB RPCs + rpcs=$($LCTL get_param 'osc.*.rpc_stats' | + sed -n '/pages per rpc/,/^$/p' | + awk 'BEGIN { sum = 0 }; /4096:/ { sum += $2 }; + END { print sum }') + echo $rpcs RPCs + [ "$rpcs" -eq 10 ] || error "not all RPCs are 16 MiB BRW rpcs" + fi + + echo "set RPC size to 4MB" + + $LCTL set_param -n osc.*.max_pages_per_rpc=4M osc.*.rpc_stats=0 + dd if=/dev/zero of=$DIR/$tfile bs=4M count=25 + cancel_lru_locks osc + dd of=/dev/null if=$DIR/$tfile bs=4M count=25 + + # calculate 4 MiB RPCs + rpcs=$($LCTL get_param 'osc.*.rpc_stats' | + sed -n '/pages per rpc/,/^$/p' | + awk 'BEGIN { sum = 0 }; /1024:/ { sum += $2 }; + END { print sum }') + echo $rpcs RPCs + [ "$rpcs" -eq 25 ] || error "not all RPCs are 4 MiB BRW rpcs" - echo checking missing pages - local miss=$($LCTL get_param -n llite.*.read_ahead_stats | - get_named_value 'misses' | cut -d" " -f1 | calc_total) + restore_lustre_params < $p + remount_client $MOUNT || error "remount_client failed" - [ $miss -lt 3 ] || error "misses too much pages!" - cleanup_test101f + rm -f $p $DIR/$tfile } -run_test 101f "check read-ahead for max_read_ahead_whole_mb" +run_test 101g "Big bulk(4/16 MiB) readahead" setup_test102() { test_mkdir -p $DIR/$tdir @@ -6950,7 +7156,7 @@ test_102n() { # LU-4101 mdt: protect internal xattrs # Get 'before' xattrs of $file1. getfattr --absolute-names --dump --match=- $file1 > $xattr0 - for name in lov lma lmv link fid version som hsm lfsck_namespace; do + for name in lov lma lmv link fid version som hsm; do # Try to copy xattr from $file0 to $file1. value=$(getxattr $file0 trusted.$name 2> /dev/null) @@ -7074,6 +7280,7 @@ test_103a() { [ -z "$(which setfacl 2>/dev/null)" ] && skip_env "could not find setfacl" && return $GSS && skip "could not run under gss" && return + remote_mds_nodsh && skip "remote MDS with nodsh" && return gpasswd -a daemon bin # LU-5641 do_facet $SINGLEMDS gpasswd -a daemon bin # LU-5641 @@ -7134,6 +7341,7 @@ test_103a() { run_test 103a "acl test =========================================" test_103b() { + remote_mds_nodsh && skip "remote MDS with nodsh" && return local noacl=false local MDT_DEV=$(mdsdevname ${SINGLEMDS//mds/}) local mountopts=$MDS_MOUNT_OPTS @@ -7468,9 +7676,10 @@ test_116a() { # was previously test_116() echo "Wrote ${DIFF}KB to smaller OST $MINI1" DIFF2=$(($MAXV1 - ${AVAIL[$MAXI1]})) echo "Wrote ${DIFF2}KB to larger OST $MAXI1" - FILL=$(($DIFF2 * 100 / $DIFF - 100)) - [ $DIFF -gt 0 ] && + if [[ $DIFF -gt 0 ]]; then + FILL=$(($DIFF2 * 100 / $DIFF - 100)) echo "Wrote ${FILL}% more data to larger OST $MAXI1" + fi # Figure out which files were written where UUID=$(lctl get_param -n lov.${FSNAME}-clilov-*.target_obd | @@ -7483,9 +7692,10 @@ test_116a() { # was previously test_116() echo $UUID MAXC=$($GETSTRIPE --ost $UUID $DIR/$tdir | grep $DIR | wc -l) echo "$MAXC files created on larger OST $MAXI1" - FILL=$(($MAXC * 100 / $MINC - 100)) - [[ $MINC -gt 0 ]] && + if [[ $MINC -gt 0 ]]; then + FILL=$(($MAXC * 100 / $MINC - 100)) echo "Wrote ${FILL}% more files to larger OST $MAXI1" + fi [[ $MAXC -gt $MINC ]] || error_ignore LU-9 "stripe QOS didn't balance free space" simple_cleanup_common @@ -7572,7 +7782,7 @@ test_118b() reset_async - #define OBD_FAIL_OST_ENOENT 0x217 + #define OBD_FAIL_SRV_ENOENT 0x217 set_nodes_failloc "$(osts_nodes)" 0x217 $MULTIOP $DIR/$tfile oO_CREAT:O_RDWR:O_SYNC:w4096c RC=$? @@ -7991,6 +8201,7 @@ run_test 119d "The DIO path should try to send a new rpc once one is completed" test_120a() { [ $PARALLEL == "yes" ] && skip "skip parallel run" && return + remote_mds_nodsh && skip "remote MDS with nodsh" && return test_mkdir -p $DIR/$tdir [ -z "`lctl get_param -n mdc.*.connect_flags | grep early_lock_cancel`" ] && \ skip "no early lock cancel on server" && return 0 @@ -8022,6 +8233,7 @@ run_test 120a "Early Lock Cancel: mkdir test" test_120b() { [ $PARALLEL == "yes" ] && skip "skip parallel run" && return + remote_mds_nodsh && skip "remote MDS with nodsh" && return test_mkdir $DIR/$tdir [ -z "$(lctl get_param -n mdc.*.connect_flags | grep early_lock_cancel)" ] && \ skip "no early lock cancel on server" && return 0 @@ -8049,6 +8261,7 @@ run_test 120b "Early Lock Cancel: create test" test_120c() { [ $PARALLEL == "yes" ] && skip "skip parallel run" && return + remote_mds_nodsh && skip "remote MDS with nodsh" && return test_mkdir -c1 $DIR/$tdir [ -z "$(lctl get_param -n mdc.*.connect_flags | grep early_lock_cancel)" ] && \ skip "no early lock cancel on server" && return 0 @@ -8079,6 +8292,7 @@ run_test 120c "Early Lock Cancel: link test" test_120d() { [ $PARALLEL == "yes" ] && skip "skip parallel run" && return + remote_mds_nodsh && skip "remote MDS with nodsh" && return test_mkdir -p -c1 $DIR/$tdir [ -z "$(lctl get_param -n mdc.*.connect_flags | grep early_lock_cancel)" ] && \ skip "no early lock cancel on server" && return 0 @@ -8107,16 +8321,21 @@ run_test 120d "Early Lock Cancel: setattr test" test_120e() { [ $PARALLEL == "yes" ] && skip "skip parallel run" && return - test_mkdir -p -c1 $DIR/$tdir - [ -z "`lctl get_param -n mdc.*.connect_flags | grep early_lock_cancel`" ] && \ - skip "no early lock cancel on server" && return 0 - lru_resize_disable mdc - lru_resize_disable osc - dd if=/dev/zero of=$DIR/$tdir/f1 count=1 - cancel_lru_locks mdc - cancel_lru_locks osc - dd if=$DIR/$tdir/f1 of=/dev/null - stat $DIR/$tdir $DIR/$tdir/f1 > /dev/null + ! $($LCTL get_param -n mdc.*.connect_flags | grep -q early_lock_can) && + skip "no early lock cancel on server" && return 0 + remote_mds_nodsh && skip "remote MDS with nodsh" && return + local dlmtrace_set=false + + test_mkdir -p -c1 $DIR/$tdir + lru_resize_disable mdc + lru_resize_disable osc + ! $LCTL get_param debug | grep -q dlmtrace && + $LCTL set_param debug=+dlmtrace && dlmtrace_set=true + dd if=/dev/zero of=$DIR/$tdir/f1 count=1 + cancel_lru_locks mdc + cancel_lru_locks osc + dd if=$DIR/$tdir/f1 of=/dev/null + stat $DIR/$tdir $DIR/$tdir/f1 > /dev/null # XXX client can not do early lock cancel of OST lock # during unlink (LU-4206), so cancel osc lock now. cancel_lru_locks osc @@ -8132,8 +8351,11 @@ test_120e() { awk '/ldlm_cancel/ {print $2}') blk2=$($LCTL get_param -n ldlm.services.ldlm_cbd.stats | awk '/ldlm_bl_callback/ {print $2}') - [ $can1 -eq $can2 ] || error $((can2-can1)) "cancel RPC occured." - [ $blk1 -eq $blk2 ] || error $((blk2-blk1)) "blocking RPC occured." + [ $can1 -ne $can2 ] && error "$((can2 - can1)) cancel RPC occured" && + $LCTL dk $TMP/cancel.debug.txt + [ $blk1 -ne $blk2 ] && error "$((blk2 - blk1)) blocking RPC occured" && + $LCTL dk $TMP/blocking.debug.txt + $dlmtrace_set && $LCTL set_param debug=-dlmtrace lru_resize_enable mdc lru_resize_enable osc } @@ -8143,6 +8365,7 @@ test_120f() { [ $PARALLEL == "yes" ] && skip "skip parallel run" && return [ -z "`lctl get_param -n mdc.*.connect_flags | grep early_lock_cancel`" ] && \ skip "no early lock cancel on server" && return 0 + remote_mds_nodsh && skip "remote MDS with nodsh" && return test_mkdir -p -c1 $DIR/$tdir lru_resize_disable mdc lru_resize_disable osc @@ -8181,6 +8404,7 @@ test_120g() { [ $PARALLEL == "yes" ] && skip "skip parallel run" && return [ -z "`lctl get_param -n mdc.*.connect_flags | grep early_lock_cancel`" ] && \ skip "no early lock cancel on server" && return 0 + remote_mds_nodsh && skip "remote MDS with nodsh" && return lru_resize_disable mdc lru_resize_disable osc count=10000 @@ -8638,11 +8862,13 @@ test_127b() { # bug LU-333 esac done < $TMP/${tfile}.tmp - #check that we actually got some stats - [ "$read_bytes" ] || error "Missing read_bytes stats" - [ "$write_bytes" ] || error "Missing write_bytes stats" - [ "$read_bytes" != 0 ] || error "no read done" - [ "$write_bytes" != 0 ] || error "no write done" + #check that we actually got some stats + [ "$read_bytes" ] || error "Missing read_bytes stats" + [ "$write_bytes" ] || error "Missing write_bytes stats" + [ "$read_bytes" != 0 ] || error "no read done" + [ "$write_bytes" != 0 ] || error "no write done" + + rm -f $TMP/${tfile}.tmp } run_test 127b "verify the llite client stats are sane" @@ -8654,8 +8880,9 @@ test_128() { # bug 15212 EOF result=$(grep error $TMP/$tfile.log) - rm -f $DIR/$tfile - [ -z "$result" ] || error "consecutive find's under interactive lfs failed" + rm -f $DIR/$tfile $TMP/$tfile.log + [ -z "$result" ] || + error "consecutive find's under interactive lfs failed" } run_test 128 "interactive lfs for 2 consecutive find's" @@ -8673,7 +8900,18 @@ set_dir_limits () { do_facet $facet "test -e $LDPROC/$canondev/max_dir_size" || LDPROC=/sys/fs/ldiskfs do_facet $facet "echo $1 >$LDPROC/$canondev/max_dir_size" + do_facet $facet "test -e $LDPROC/$canondev/warning_dir_size" || + LDPROC=/sys/fs/ldiskfs + do_facet $facet "echo $2 >$LDPROC/$canondev/warning_dir_size" + done +} + +check_mds_dmesg() { + local facets=$(get_facets MDS) + for facet in ${facets//,/ }; do + do_facet $facet "dmesg | tail -3 | grep -q $1" && return 0 done + return 1 } test_129() { @@ -8683,33 +8921,42 @@ test_129() { return fi remote_mds_nodsh && skip "remote MDS with nodsh" && return - ENOSPC=28 - EFBIG=27 + local ENOSPC=28 + local EFBIG=27 + local has_warning=0 rm -rf $DIR/$tdir - test_mkdir -p $DIR/$tdir + mkdir -p $DIR/$tdir # block size of mds1 local MDT_DEV=$(mdsdevname ${SINGLEMDS//mds/}) local MDSBLOCKSIZE=$($LCTL get_param -n mdc.*MDT0000*.blocksize) - local MAX=$((MDSBLOCKSIZE * 3)) - set_dir_limits $MAX + local MAX=$((MDSBLOCKSIZE * 5)) + set_dir_limits $MAX $MAX local I=$(stat -c%s "$DIR/$tdir") local J=0 - local STRIPE_COUNT=1 - [[ $MDSCOUNT -ge 2 ]] && STRIPE_COUNT=$($LFS getdirstripe -c $DIR/$tdir) - MAX=$((MAX*STRIPE_COUNT)) while [[ $I -le $MAX ]]; do $MULTIOP $DIR/$tdir/$J Oc rc=$? + if [ $has_warning -eq 0 ]; then + check_mds_dmesg '"is approaching"' && + has_warning=1 + fi #check two errors ENOSPC for new version of ext4 max_dir_size patch #mainline kernel commit df981d03eeff7971ac7e6ff37000bfa702327ef1 #and EFBIG for previous versions if [ $rc -eq $EFBIG -o $rc -eq $ENOSPC ]; then - set_dir_limits 0 + set_dir_limits 0 0 echo "return code $rc received as expected" - multiop $DIR/$tdir/$J Oc || - error_exit "multiop failed w/o dir size limit" + + createmany -o $DIR/$tdir/$J_file_ 1000 || + error_exit "create failed w/o dir size limit" + + check_mds_dmesg '"has reached"' || + error_exit "has reached message should be output" + + [ $has_warning -eq 0 ] && + error_exit "warning message should be output" I=$(stat -c%s "$DIR/$tdir") @@ -8722,7 +8969,7 @@ test_129() { fi error_exit "current dir size $I, previous limit $MAX" elif [ $rc -ne 0 ]; then - set_dir_limits 0 + set_dir_limits 0 0 error_exit "return code $rc received instead of expected " \ "$EFBIG or $ENOSPC, files in dir $I" fi @@ -8730,7 +8977,7 @@ test_129() { I=$(stat -c%s "$DIR/$tdir") done - set_dir_limits 0 + set_dir_limits 0 0 error "exceeded dir size limit $MAX($MDSCOUNT) : $I bytes" } run_test 129 "test directory size limit ========================" @@ -8760,8 +9007,8 @@ test_130a() { skip "ORI-366/LU-1941: FIEMAP unimplemented on ZFS" && return [ $RC != 0 ] && error "filefrag $fm_file failed" - filefrag_op=$(filefrag -ve $fm_file | grep -A 100 "ext:" | - grep -v "ext:" | grep -v "found") + filefrag_op=$(filefrag -ve $fm_file | + sed -n '/ext:/,/found/{/ext:/d; /found/d; p}') lun=$($GETSTRIPE -i $fm_file) start_blk=`echo $filefrag_op | cut -d: -f2 | cut -d. -f1` @@ -8793,10 +9040,7 @@ run_test 130a "FIEMAP (1-stripe file)" test_130b() { [ "$OSTCOUNT" -lt "2" ] && - skip_env "skipping FIEMAP on 2-stripe file test" && return - - [ "$OSTCOUNT" -ge "10" ] && - skip_env "skipping FIEMAP with >= 10 OSTs" && return + skip_env "skipping FIEMAP on $OSTCOUNT-stripe file" && return local filefrag_op=$(filefrag -e 2>&1 | grep "invalid option") [ -n "$filefrag_op" ] && skip_env "filefrag does not support FIEMAP" && @@ -8805,30 +9049,34 @@ test_130b() { trap cleanup_130 EXIT RETURN local fm_file=$DIR/$tfile - $SETSTRIPE -S 65536 -c 2 $fm_file || error "setstripe on $fm_file" + $SETSTRIPE -S 65536 -c $OSTCOUNT $fm_file || + error "setstripe on $fm_file" [ "$(facet_fstype ost$(($($GETSTRIPE -i $fm_file) + 1)))" = "zfs" ] && skip "ORI-366/LU-1941: FIEMAP unimplemented on ZFS" && return - dd if=/dev/zero of=$fm_file bs=1M count=2 || + dd if=/dev/zero of=$fm_file bs=1M count=$OSTCOUNT || error "dd failed on $fm_file" filefrag -ves $fm_file || error "filefrag $fm_file failed" - filefrag_op=$(filefrag -ve $fm_file | grep -A 100 "ext:" | - grep -v "ext:" | grep -v "found") + filefrag_op=$(filefrag -ve $fm_file | + sed -n '/ext:/,/found/{/ext:/d; /found/d; p}') - last_lun=$(echo $filefrag_op | cut -d: -f5) + last_lun=$(echo $filefrag_op | cut -d: -f5 | + sed -e 's/^[ \t]*/0x/' | sed -e 's/0x0x/0x/') IFS=$'\n' tot_len=0 num_luns=1 for line in $filefrag_op do - frag_lun=`echo $line | cut -d: -f5` - ext_len=`echo $line | cut -d: -f4` + frag_lun=$(echo $line | cut -d: -f5 | + sed -e 's/^[ \t]*/0x/' | sed -e 's/0x0x/0x/') + ext_len=$(echo $line | cut -d: -f4) if (( $frag_lun != $last_lun )); then if (( tot_len != 1024 )); then cleanup_130 - error "FIEMAP on $fm_file failed; returned len $tot_len for OST $last_lun instead of 256" + error "FIEMAP on $fm_file failed; returned " \ + "len $tot_len for OST $last_lun instead of 1024" return else (( num_luns += 1 )) @@ -8838,25 +9086,23 @@ test_130b() { (( tot_len += ext_len )) last_lun=$frag_lun done - if (( num_luns != 2 || tot_len != 1024 )); then + if (( num_luns != $OSTCOUNT || tot_len != 1024 )); then cleanup_130 - error "FIEMAP on $fm_file failed; returned wrong number of luns or wrong len for OST $last_lun" + error "FIEMAP on $fm_file failed; returned wrong number of " \ + "luns or wrong len for OST $last_lun" return fi cleanup_130 - echo "FIEMAP on 2-stripe file succeeded" + echo "FIEMAP on $OSTCOUNT-stripe file succeeded" } -run_test 130b "FIEMAP (2-stripe file)" +run_test 130b "FIEMAP ($OSTCOUNT-stripe file)" test_130c() { [ "$OSTCOUNT" -lt "2" ] && skip_env "skipping FIEMAP on 2-stripe file" && return - [ "$OSTCOUNT" -ge "10" ] && - skip_env "skipping FIEMAP with >= 10 OSTs" && return - filefrag_op=$(filefrag -e 2>&1 | grep "invalid option") [ -n "$filefrag_op" ] && skip "filefrag does not support FIEMAP" && return @@ -8868,30 +9114,36 @@ test_130c() { [ "$(facet_fstype ost$(($($GETSTRIPE -i $fm_file) + 1)))" = "zfs" ] && skip "ORI-366/LU-1941: FIEMAP unimplemented on ZFS" && return - dd if=/dev/zero of=$fm_file seek=1 bs=1M count=1 || error "dd failed on $fm_file" + dd if=/dev/zero of=$fm_file seek=1 bs=1M count=1 || + error "dd failed on $fm_file" filefrag -ves $fm_file || error "filefrag $fm_file failed" - filefrag_op=`filefrag -ve $fm_file | grep -A 100 "ext:" | grep -v "ext:" | grep -v "found"` + filefrag_op=$(filefrag -ve $fm_file | + sed -n '/ext:/,/found/{/ext:/d; /found/d; p}') - last_lun=`echo $filefrag_op | cut -d: -f5` + last_lun=$(echo $filefrag_op | cut -d: -f5 | + sed -e 's/^[ \t]*/0x/' | sed -e 's/0x0x/0x/') IFS=$'\n' tot_len=0 num_luns=1 for line in $filefrag_op do - frag_lun=`echo $line | cut -d: -f5` - ext_len=`echo $line | cut -d: -f4` + frag_lun=$(echo $line | cut -d: -f5 | + sed -e 's/^[ \t]*/0x/' | sed -e 's/0x0x/0x/') + ext_len=$(echo $line | cut -d: -f4) if (( $frag_lun != $last_lun )); then logical=`echo $line | cut -d: -f2 | cut -d. -f1` if (( logical != 512 )); then cleanup_130 - error "FIEMAP on $fm_file failed; returned logical start for lun $logical instead of 512" + error "FIEMAP on $fm_file failed; returned " \ + "logical start for lun $logical instead of 512" return fi if (( tot_len != 512 )); then cleanup_130 - error "FIEMAP on $fm_file failed; returned len $tot_len for OST $last_lun instead of 1024" + error "FIEMAP on $fm_file failed; returned " \ + "len $tot_len for OST $last_lun instead of 1024" return else (( num_luns += 1 )) @@ -8903,7 +9155,8 @@ test_130c() { done if (( num_luns != 2 || tot_len != 512 )); then cleanup_130 - error "FIEMAP on $fm_file failed; returned wrong number of luns or wrong len for OST $last_lun" + error "FIEMAP on $fm_file failed; returned wrong number of " \ + "luns or wrong len for OST $last_lun" return fi @@ -8914,18 +9167,18 @@ test_130c() { run_test 130c "FIEMAP (2-stripe file with hole)" test_130d() { - [ "$OSTCOUNT" -lt "3" ] && skip_env "skipping FIEMAP on N-stripe file test" && return - - [ "$OSTCOUNT" -ge "10" ] && - skip_env "skipping FIEMAP with >= 10 OSTs" && return + [ "$OSTCOUNT" -lt "3" ] && + skip_env "skipping FIEMAP on N-stripe file test" && return filefrag_op=$(filefrag -e 2>&1 | grep "invalid option") - [ -n "$filefrag_op" ] && skip "filefrag does not support FIEMAP" && return + [ -n "$filefrag_op" ] && skip "filefrag does not support FIEMAP" && + return trap cleanup_130 EXIT RETURN local fm_file=$DIR/$tfile - $SETSTRIPE -S 65536 -c $OSTCOUNT $fm_file||error "setstripe on $fm_file" + $SETSTRIPE -S 65536 -c $OSTCOUNT $fm_file || + error "setstripe on $fm_file" [ "$(facet_fstype ost$(($($GETSTRIPE -i $fm_file) + 1)))" = "zfs" ] && skip "ORI-366/LU-1941: FIEMAP unimplemented on ZFS" && return @@ -8934,22 +9187,25 @@ test_130d() { error "dd failed on $fm_file" filefrag -ves $fm_file || error "filefrag $fm_file failed" - filefrag_op=`filefrag -ve $fm_file | grep -A 100 "ext:" | - grep -v "ext:" | grep -v "found"` + filefrag_op=$(filefrag -ve $fm_file | + sed -n '/ext:/,/found/{/ext:/d; /found/d; p}') - last_lun=`echo $filefrag_op | cut -d: -f5` + last_lun=$(echo $filefrag_op | cut -d: -f5 | + sed -e 's/^[ \t]*/0x/' | sed -e 's/0x0x/0x/') IFS=$'\n' tot_len=0 num_luns=1 for line in $filefrag_op do - frag_lun=`echo $line | cut -d: -f5` - ext_len=`echo $line | cut -d: -f4` + frag_lun=$(echo $line | cut -d: -f5 | + sed -e 's/^[ \t]*/0x/' | sed -e 's/0x0x/0x/') + ext_len=$(echo $line | cut -d: -f4) if (( $frag_lun != $last_lun )); then if (( tot_len != 1024 )); then cleanup_130 - error "FIEMAP on $fm_file failed; returned len $tot_len for OST $last_lun instead of 1024" + error "FIEMAP on $fm_file failed; returned " \ + "len $tot_len for OST $last_lun instead of 1024" return else (( num_luns += 1 )) @@ -8961,7 +9217,8 @@ test_130d() { done if (( num_luns != actual_stripecnt || tot_len != 1024 )); then cleanup_130 - error "FIEMAP on $fm_file failed; returned wrong number of luns or wrong len for OST $last_lun" + error "FIEMAP on $fm_file failed; returned wrong number of " \ + "luns or wrong len for OST $last_lun" return fi @@ -8972,10 +9229,8 @@ test_130d() { run_test 130d "FIEMAP (N-stripe file)" test_130e() { - [ "$OSTCOUNT" -lt "2" ] && skip_env "skipping continuation FIEMAP test" && return - - [ "$OSTCOUNT" -ge "10" ] && - skip_env "skipping FIEMAP with >= 10 OSTs" && return + [ "$OSTCOUNT" -lt "2" ] && + skip_env "skipping continuation FIEMAP test" && return filefrag_op=$(filefrag -e 2>&1 | grep "invalid option") [ -n "$filefrag_op" ] && skip "filefrag does not support FIEMAP" && return @@ -8995,21 +9250,26 @@ test_130e() { done filefrag -ves $fm_file || error "filefrag $fm_file failed" - filefrag_op=`filefrag -ve $fm_file | grep -A 12000 "ext:" | grep -v "ext:" | grep -v "found"` + filefrag_op=$(filefrag -ve $fm_file | + sed -n '/ext:/,/found/{/ext:/d; /found/d; p}') - last_lun=`echo $filefrag_op | cut -d: -f5` + last_lun=$(echo $filefrag_op | cut -d: -f5 | + sed -e 's/^[ \t]*/0x/' | sed -e 's/0x0x/0x/') IFS=$'\n' tot_len=0 num_luns=1 for line in $filefrag_op do - frag_lun=`echo $line | cut -d: -f5` - ext_len=`echo $line | cut -d: -f4` + frag_lun=$(echo $line | cut -d: -f5 | + sed -e 's/^[ \t]*/0x/' | sed -e 's/0x0x/0x/') + ext_len=$(echo $line | cut -d: -f4) if (( $frag_lun != $last_lun )); then if (( tot_len != $EXPECTED_LEN )); then cleanup_130 - error "FIEMAP on $fm_file failed; returned len $tot_len for OST $last_lun instead of $EXPECTED_LEN" + error "FIEMAP on $fm_file failed; returned " \ + "len $tot_len for OST $last_lun instead " \ + "of $EXPECTED_LEN" return else (( num_luns += 1 )) @@ -9021,7 +9281,8 @@ test_130e() { done if (( num_luns != 2 || tot_len != $EXPECTED_LEN )); then cleanup_130 - error "FIEMAP on $fm_file failed; returned wrong number of luns or wrong len for OST $last_lun" + error "FIEMAP on $fm_file failed; returned wrong number " \ + "of luns or wrong len for OST $last_lun" return fi @@ -9111,12 +9372,14 @@ test_133a() { # verify mdt stats first. mkdir ${testdir} || error "mkdir failed" check_stats $SINGLEMDS "mkdir" 1 - touch ${testdir}/${tfile} || "touch failed" + touch ${testdir}/${tfile} || error "touch failed" check_stats $SINGLEMDS "open" 1 check_stats $SINGLEMDS "close" 1 - mknod ${testdir}/${tfile}-pipe p || "mknod failed" - check_stats $SINGLEMDS "mknod" 1 - rm -f ${testdir}/${tfile}-pipe || "pipe remove failed" + [ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.8.54) ] && { + mknod ${testdir}/${tfile}-pipe p || error "mknod failed" + check_stats $SINGLEMDS "mknod" 2 + } + rm -f ${testdir}/${tfile}-pipe || error "pipe remove failed" check_stats $SINGLEMDS "unlink" 1 rm -f ${testdir}/${tfile} || error "file remove failed" check_stats $SINGLEMDS "unlink" 2 @@ -9145,7 +9408,7 @@ test_133b() { remote_mds_nodsh && skip "remote MDS with nodsh" && return local testdir=$DIR/${tdir}/stats_testdir mkdir -p ${testdir} || error "mkdir failed" - touch ${testdir}/${tfile} || "touch failed" + touch ${testdir}/${tfile} || error "touch failed" cancel_lru_locks mdc # clear stats. @@ -9375,31 +9638,25 @@ test_133e() { } run_test 133e "Verifying OST {read,write}_bytes nid stats =================" -test_133f() { - local proc_dirs - - local dirs="/proc/fs/lustre/ /proc/sys/lnet/ /proc/sys/lustre/ \ -/sys/fs/lustre/ /sys/fs/lnet/" - local dir - for dir in $dirs; do - if [ -d $dir ]; then - proc_dirs="$proc_dirs $dir" - fi - done - - local facet +proc_dirs="" +for dir in /proc/fs/lustre/ /proc/sys/lnet/ /proc/sys/lustre/ \ + /sys/fs/lustre/ /sys/fs/lnet/ /sys/kernel/debug/lnet/ \ + /sys/kernel/debug/lustre/; do + [[ -d $dir ]] && proc_dirs+=" $dir" +done +test_133f() { remote_mds_nodsh && skip "remote MDS with nodsh" && return remote_ost_nodsh && skip "remote OST with nodsh" && return # First without trusting modes. find $proc_dirs -exec cat '{}' \; &> /dev/null # Second verifying readability. - find $proc_dirs \ - -type f \ - -exec cat '{}' \; &> /dev/null || - error "proc file read failed" + $LCTL get_param -R '*' &> /dev/null || error "proc file read failed" + # eventually, this can also be replaced with "lctl get_param -R", + # but not until that option is always available on the server + local facet for facet in $SINGLEMDS ost1; do do_facet $facet find $proc_dirs \ ! -name req_history \ @@ -9415,20 +9672,9 @@ test_133f() { run_test 133f "Check for LBUGs/Oopses/unreadable files in /proc" test_133g() { - local proc_dirs - - local dirs="/proc/fs/lustre/ /proc/sys/lnet/ /proc/sys/lustre/ \ -/sys/fs/lustre/ /sys/fs/lnet/" - local dir - for dir in $dirs; do - if [ -d $dir ]; then - proc_dirs="$proc_dirs $dir" - fi - done - - local facet - - # Second verifying readability. + remote_mds_nodsh && skip "remote MDS with nodsh" && return + remote_ost_nodsh && skip "remote OST with nodsh" && return + # Second verifying writability. find $proc_dirs \ -type f \ -not -name force_lbug \ @@ -9442,6 +9688,7 @@ test_133g() { [ $(lustre_version_code ost1) -le $(version_code 2.5.54) ] && skip "Too old lustre on ost1" && return + local facet for facet in $SINGLEMDS ost1; do do_facet $facet find $proc_dirs \ -type f \ @@ -9460,6 +9707,7 @@ test_133g() { run_test 133g "Check for Oopses on bad io area writes/reads in /proc" test_134a() { + remote_mds_nodsh && skip "remote MDS with nodsh" && return [[ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.7.54) ]] && skip "Need MDS version at least 2.7.54" && return @@ -9468,7 +9716,7 @@ test_134a() { local nsdir="ldlm.namespaces.*-MDT0000-mdc-*" local unused=$($LCTL get_param -n $nsdir.lock_unused_count) - [ $unused -eq 0 ] || "$unused locks are not cleared" + [ $unused -eq 0 ] || error "$unused locks are not cleared" local nr=1000 createmany -o $DIR/$tdir/f $nr || @@ -9495,6 +9743,7 @@ test_134a() { run_test 134a "Server reclaims locks when reaching lock_reclaim_threshold" test_134b() { + remote_mds_nodsh && skip "remote MDS with nodsh" && return [[ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.7.54) ]] && skip "Need MDS version at least 2.7.54" && return @@ -9536,9 +9785,9 @@ run_test 134b "Server rejects lock request when reaching lock_limit_mb" test_140() { #bug-17379 [ $PARALLEL == "yes" ] && skip "skip parallel run" && return - test_mkdir -p $DIR/$tdir || error "Creating dir $DIR/$tdir" - cd $DIR/$tdir || error "Changing to $DIR/$tdir" - cp /usr/bin/stat . || error "Copying stat to $DIR/$tdir" + test_mkdir -p $DIR/$tdir || error "Creating dir $DIR/$tdir" + cd $DIR/$tdir || error "Changing to $DIR/$tdir" + cp $(which stat) . || error "Copying stat to $DIR/$tdir" # VFS limits max symlink depth to 5(4KSTACK) or 7(8KSTACK) or 8 # For kernel > 3.5, bellow only tests consecutive symlink (MAX 40) @@ -9905,6 +10154,7 @@ test_154a() { [[ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.2.51) ]] || { skip "Need MDS version at least 2.2.51"; return 0; } [ -z "$(which setfacl)" ] && skip "must have setfacl tool" && return + [ -n "$FILESET" ] && skip "SKIP due to FILESET set" && return cp /etc/hosts $DIR/$tfile @@ -9931,6 +10181,7 @@ test_154b() { [ $PARALLEL == "yes" ] && skip "skip parallel run" && return [[ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.2.51) ]] || { skip "Need MDS version at least 2.2.51"; return 0; } + [ -n "$FILESET" ] && skip "SKIP due to FILESET set" && return [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return @@ -9983,6 +10234,7 @@ test_154c() { run_test 154c "lfs path2fid and fid2path multiple arguments" test_154d() { + remote_mds_nodsh && skip "remote MDS with nodsh" && return [[ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.5.53) ]] && skip "Need MDS version at least 2.5.53" && return @@ -10027,6 +10279,7 @@ test_154e() run_test 154e ".lustre is not returned by readdir" test_154f() { + [ -n "$FILESET" ] && skip "SKIP due to FILESET set" && return # create parent directory on a single MDT to avoid cross-MDT hardlinks test_mkdir -p -c1 $DIR/$tdir/d # test dirs inherit from its stripe @@ -10099,6 +10352,7 @@ test_154f() { rm -f $DIR/f restore_lustre_params < $save + rm -f $save } run_test 154f "get parent fids by reading link ea" @@ -10106,6 +10360,7 @@ test_154g() { [[ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.6.92) ]] || { skip "Need MDS version at least 2.6.92"; return 0; } + [ -n "$FILESET" ] && skip "SKIP due to FILESET set" && return mkdir -p $DIR/$tdir llapi_fid_test -d $DIR/$tdir @@ -10377,6 +10632,12 @@ test_156() { run_test 156 "Verification of tunables" #Changelogs +cleanup_changelog () { + trap 0 + echo "Deregistering changelog client $CL_USER" + do_facet $SINGLEMDS $LCTL --device $MDT0 changelog_deregister $CL_USER +} + err17935 () { if [[ $MDSCOUNT -gt 1 ]]; then error_ignore bz17935 $* @@ -10417,10 +10678,12 @@ test_160a() { local CL_USERS="mdd.$MDT0.changelog_users" local GET_CL_USERS="do_facet $SINGLEMDS $LCTL get_param -n $CL_USERS" - USER=$(do_facet $SINGLEMDS $LCTL --device $MDT0 changelog_register -n) - echo "Registered as changelog user $USER" - $GET_CL_USERS | grep -q $USER || - error "User $USER not found in changelog_users" + CL_USER=$(do_facet $SINGLEMDS $LCTL --device $MDT0 \ + changelog_register -n) + echo "Registered as changelog user $CL_USER" + trap cleanup_changelog EXIT + $GET_CL_USERS | grep -q $CL_USER || + error "User $CL_USER not found in changelog_users" # change something test_mkdir -p $DIR/$tdir/pics/2008/zachy @@ -10464,9 +10727,9 @@ test_160a() { [ "$fidc" == "$fidf" ] || err17935 "pfid in changelog $fidc != dir fid $fidf" - USER_REC1=$($GET_CL_USERS | awk "\$1 == \"$USER\" {print \$2}") - $LFS changelog_clear $MDT0 $USER $(($USER_REC1 + 5)) - USER_REC2=$($GET_CL_USERS | awk "\$1 == \"$USER\" {print \$2}") + USER_REC1=$($GET_CL_USERS | awk "\$1 == \"$CL_USER\" {print \$2}") + $LFS changelog_clear $MDT0 $CL_USER $(($USER_REC1 + 5)) + USER_REC2=$($GET_CL_USERS | awk "\$1 == \"$CL_USER\" {print \$2}") echo "verifying user clear: $(( $USER_REC1 + 5 )) == $USER_REC2" [ $USER_REC2 == $(($USER_REC1 + 5)) ] || err17935 "user index expected $(($USER_REC1 + 5)) is $USER_REC2" @@ -10481,7 +10744,7 @@ test_160a() { # LU-3446 changelog index reset on MDT restart local MDT_DEV=$(mdsdevname ${SINGLEMDS//mds/}) CUR_REC1=$($GET_CL_USERS | head -n1 | cut -f3 -d' ') - $LFS changelog_clear $MDT0 $USER 0 + $LFS changelog_clear $MDT0 $CL_USER 0 stop $SINGLEMDS || error "Fail to stop MDT." start $SINGLEMDS $MDT_DEV $MDS_MOUNT_OPTS || error "Fail to start MDT." CUR_REC2=$($GET_CL_USERS | head -n1 | cut -f3 -d' ') @@ -10490,19 +10753,19 @@ test_160a() { err17935 "current index should be $CUR_REC1 is $CUR_REC2" echo "verifying user deregister" - do_facet $SINGLEMDS $LCTL --device $MDT0 changelog_deregister $USER - $GET_CL_USERS | grep -q $USER && - error "User $USER still in changelog_users" + cleanup_changelog + $GET_CL_USERS | grep -q $CL_USER && + error "User $CL_USER still in changelog_users" USERS=$(( $($GET_CL_USERS | wc -l) - 2 )) - if [ $USERS -eq 0 ]; then + if [ $CL_USER -eq 0 ]; then LAST_REC1=$($GET_CL_USERS | head -n1 | cut -f3 -d' ') touch $DIR/$tdir/chloe LAST_REC2=$($GET_CL_USERS | head -n1 | cut -f3 -d' ') echo "verify changelogs are off: $LAST_REC1 == $LAST_REC2" [ $LAST_REC1 == $LAST_REC2 ] || error "changelogs not off" else - echo "$USERS other changelog users; can't verify off" + echo "$CL_USER other changelog users; can't verify off" fi } run_test 160a "changelog sanity" @@ -10515,10 +10778,12 @@ test_160b() { # LU-3587 local CL_USERS="mdd.$MDT0.changelog_users" local GET_CL_USERS="do_facet $SINGLEMDS $LCTL get_param -n $CL_USERS" - USER=$(do_facet $SINGLEMDS $LCTL --device $MDT0 changelog_register -n) - echo "Registered as changelog user $USER" - $GET_CL_USERS | grep -q $USER || - error "User $USER not found in changelog_users" + CL_USER=$(do_facet $SINGLEMDS $LCTL --device $MDT0 \ + changelog_register -n) + echo "Registered as changelog user $CL_USER" + trap cleanup_changelog EXIT + $GET_CL_USERS | grep -q $CL_USER || + error "User $CL_USER not found in changelog_users" local LONGNAME1=$(str_repeat a 255) local LONGNAME2=$(str_repeat b 255) @@ -10530,11 +10795,8 @@ test_160b() { # LU-3587 mv $LONGNAME1 $LONGNAME2 $LFS changelog $MDT0 | grep RENME - - echo "deregistering $USER" - do_facet $SINGLEMDS $LCTL --device $MDT0 changelog_deregister $USER - rm -f $LONGNAME2 + cleanup_changelog } run_test 160b "Verify that very long rename doesn't crash in changelog" @@ -10551,8 +10813,9 @@ test_160c() { [ $PARALLEL == "yes" ] && skip "skip parallel run" && return # Registration step - local USER=$(do_facet $SINGLEMDS $LCTL --device $MDT0 \ + CL_USER=$(do_facet $SINGLEMDS $LCTL --device $MDT0 \ changelog_register -n) + trap cleanup_changelog EXIT rm -rf $DIR/$tdir mkdir -p $DIR/$tdir @@ -10564,14 +10827,44 @@ test_160c() { $LFS changelog $MDT0 TRUNCS=$($LFS changelog $MDT0 | tail -5 | grep -c "TRUNC") [ $TRUNCS -eq 1 ] || err17935 "TRUNC changelog mask count $TRUNCS != 1" - $LFS changelog_clear $MDT0 $USER 0 + $LFS changelog_clear $MDT0 $CL_USER 0 # Deregistration step - echo "deregistering $USER" - do_facet $SINGLEMDS $LCTL --device $MDT0 changelog_deregister $USER + cleanup_changelog } run_test 160c "verify that changelog log catch the truncate event" +test_160d() { + remote_mds_nodsh && skip "remote MDS with nodsh" && return + [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return + + local server_version=$(lustre_version_code mds1) + local CL_MASK_PARAM="mdd.$MDT0.changelog_mask" + + [[ $server_version -ge $(version_code 2.7.60) ]] || + { skip "Need MDS version at least 2.7.60+"; return; } + [ $PARALLEL == "yes" ] && skip "skip parallel run" && return + + # Registration step + CL_USER=$(do_facet mds1 $LCTL --device $MDT0 \ + changelog_register -n) + + trap cleanup_changelog EXIT + mkdir -p $DIR/$tdir/migrate_dir + $LFS changelog_clear $MDT0 $CL_USER 0 + + $LFS migrate -m 1 $DIR/$tdir/migrate_dir || error "migrate fails" + $LFS changelog $MDT0 + MIGRATES=$($LFS changelog $MDT0 | tail -5 | grep -c "MIGRT") + $LFS changelog_clear $MDT0 $CL_USER 0 + [ $MIGRATES -eq 1 ] || + error "MIGRATE changelog mask count $MIGRATES != 1" + + # Deregistration step + cleanup_changelog +} +run_test 160d "verify that changelog log catch the migrate event" + test_161a() { [ $PARALLEL == "yes" ] && skip "skip parallel run" && return test_mkdir -p -c1 $DIR/$tdir @@ -10617,8 +10910,7 @@ run_test 161a "link ea sanity" test_161b() { [ $PARALLEL == "yes" ] && skip "skip parallel run" && return - [ $MDSCOUNT -lt 2 ] && - skip "skipping remote directory test" && return + [ $MDSCOUNT -lt 2 ] && skip "skipping remote directory test" && return local MDTIDX=1 local remote_dir=$DIR/$tdir/remote_dir @@ -10676,8 +10968,10 @@ test_161c() { # define CLF_RENAME_LAST 0x0001 # rename overwrite a target having nlink = 1 (changelog flag 0x1) - local USER=$(do_facet $SINGLEMDS $LCTL --device $MDT0 \ + CL_USER=$(do_facet $SINGLEMDS $LCTL --device $MDT0 \ changelog_register -n) + + trap cleanup_changelog EXIT rm -rf $DIR/$tdir mkdir -p $DIR/$tdir touch $DIR/$tdir/foo_161c @@ -10686,10 +10980,8 @@ test_161c() { $LFS changelog $MDT0 | grep RENME local flags=$($LFS changelog $MDT0 | grep RENME | tail -1 | \ cut -f5 -d' ') - $LFS changelog_clear $MDT0 $USER 0 + $LFS changelog_clear $MDT0 $CL_USER 0 if [ x$flags != "x0x1" ]; then - do_facet $SINGLEMDS $LCTL --device $MDT0 changelog_deregister \ - $USER error "flag $flags is not 0x1" fi echo "rename overwrite a target having nlink = 1," \ @@ -10702,10 +10994,8 @@ test_161c() { mv -f $DIR/$tdir/foo_161c $DIR/$tdir/bar_161c $LFS changelog $MDT0 | grep RENME flags=$($LFS changelog $MDT0 | grep RENME | tail -1 | cut -f5 -d' ') - $LFS changelog_clear $MDT0 $USER 0 + $LFS changelog_clear $MDT0 $CL_USER 0 if [ x$flags != "x0x0" ]; then - do_facet $SINGLEMDS $LCTL --device $MDT0 changelog_deregister \ - $USER error "flag $flags is not 0x0" fi echo "rename overwrite a target having nlink > 1," \ @@ -10716,10 +11006,8 @@ test_161c() { mv -f $DIR/$tdir/foo_161c $DIR/$tdir/foo2_161c $LFS changelog $MDT0 | grep RENME flags=$($LFS changelog $MDT0 | grep RENME | tail -1 | cut -f5 -d' ') - $LFS changelog_clear $MDT0 $USER 0 + $LFS changelog_clear $MDT0 $CL_USER 0 if [ x$flags != "x0x0" ]; then - do_facet $SINGLEMDS $LCTL --device $MDT0 changelog_deregister \ - $USER error "flag $flags is not 0x0" fi echo "rename doesn't overwrite a target," \ @@ -10730,10 +11018,8 @@ test_161c() { rm -f $DIR/$tdir/foo2_161c $LFS changelog $MDT0 | grep UNLNK flags=$($LFS changelog $MDT0 | grep UNLNK | tail -1 | cut -f5 -d' ') - $LFS changelog_clear $MDT0 $USER 0 + $LFS changelog_clear $MDT0 $CL_USER 0 if [ x$flags != "x0x1" ]; then - do_facet $SINGLEMDS $LCTL --device $MDT0 changelog_deregister \ - $USER error "flag $flags is not 0x1" fi echo "unlink a file having nlink = 1," \ @@ -10744,15 +11030,13 @@ test_161c() { rm -f $DIR/$tdir/foobar_161c $LFS changelog $MDT0 | grep UNLNK flags=$($LFS changelog $MDT0 | grep UNLNK | tail -1 | cut -f5 -d' ') - $LFS changelog_clear $MDT0 $USER 0 + $LFS changelog_clear $MDT0 $CL_USER 0 if [ x$flags != "x0x0" ]; then - do_facet $SINGLEMDS $LCTL --device $MDT0 changelog_deregister \ - $USER error "flag $flags is not 0x0" fi echo "unlink a file having nlink > 1," \ "changelog record has flags of $flags" - do_facet $SINGLEMDS $LCTL --device $MDT0 changelog_deregister $USER + cleanup_changelog } run_test 161c "check CL_RENME[UNLINK] changelog record flags" @@ -10776,7 +11060,7 @@ check_path() { echo "fid $fid resolves to path $path (expected $expected)" } -test_162() { +test_162a() { # was test_162 # Make changes to filesystem [ $PARALLEL == "yes" ] && skip "skip parallel run" && return test_mkdir -p -c1 $DIR/$tdir/d2 @@ -10824,7 +11108,7 @@ test_162() { return 0 } -run_test 162 "path lookup sanity" +run_test 162a "path lookup sanity" test_162b() { [ $PARALLEL == "yes" ] && skip "skip parallel run" && return @@ -11240,6 +11524,8 @@ test_184b() { run_test 184b "Forbidden layout swap (will generate errors)" test_184c() { + local cmpn_arg=$(cmp -n 2>&1 | grep "invalid option") + [ -n "$cmpn_arg" ] && skip_env "cmp does not support -n" && return check_swap_layouts_support && return 0 local dir0=$DIR/$tdir/$testnum @@ -11364,6 +11650,7 @@ test_185() { # LU-2441 local mtime1=$(stat -c "%Y" $DIR/$tdir) local fid=$($MULTIOP $DIR/$tdir VFw4096c) || error "cannot create/write a volatile file" + [ "$FILESET" == "" ] && $CHECKSTAT -t file $MOUNT/.lustre/fid/$fid 2>/dev/null && error "FID is still valid after close" @@ -11378,8 +11665,10 @@ test_185() { # LU-2441 # is unfortunately eaten by multiop_bg_pause local n=$((${fidv[1]} + 1)) local next_fid="${fidv[0]}:$(printf "0x%x" $n):${fidv[2]}" - $CHECKSTAT -t file $MOUNT/.lustre/fid/$next_fid || - error "FID is missing before close" + if [ "$FILESET" == "" ]; then + $CHECKSTAT -t file $MOUNT/.lustre/fid/$next_fid || + error "FID is missing before close" + fi kill -USR1 $multi_pid # 1 second delay, so if mtime change we will see it sleep 1 @@ -11430,6 +11719,7 @@ run_test 187b "Test data version change on volatile file" test_200() { [ $PARALLEL == "yes" ] && skip "skip parallel run" && return remote_mgs_nodsh && skip "remote MGS with nodsh" && return + [ -n "$FILESET" ] && skip "SKIP due to FILESET set" && return local POOL=${POOL:-cea1} local POOL_ROOT=${POOL_ROOT:-$DIR/d200.pools} @@ -11607,7 +11897,12 @@ elif [ -n "$LOADL_STEPID" ]; then # LoadLeveller elif [ -n "$JOB_ID" ]; then # Sun Grid Engine JOBENV=JOB_ID else - JOBENV=FAKE_JOBID + $LCTL list_param jobid_name > /dev/null 2>&1 + if [ $? -eq 0 ]; then + JOBENV=nodelocal + else + JOBENV=FAKE_JOBID + fi fi verify_jobstats() { @@ -11628,6 +11923,13 @@ verify_jobstats() { FAKE_JOBID=id.$testnum.$(basename ${cmd[0]}).$RANDOM JOBVAL=${!JOBENV} + + [ "$JOBENV" = "nodelocal" ] && { + FAKE_JOBID=id.$testnum.$(basename ${cmd[0]}).$RANDOM + $LCTL set_param jobid_name=$FAKE_JOBID + JOBVAL=$FAKE_JOBID + } + log "Test: ${cmd[*]}" log "Using JobID environment variable $JOBENV=$JOBVAL" @@ -11656,10 +11958,11 @@ jobstats_set() { } cleanup_205() { + trap 0 do_facet $SINGLEMDS \ $LCTL set_param mdt.*.job_cleanup_interval=$OLD_INTERVAL [ $OLD_JOBENV != $JOBENV ] && jobstats_set $OLD_JOBENV - do_facet $SINGLEMDS lctl --device $MDT0 changelog_deregister $CL_USER + cleanup_changelog } test_205() { # Job stats @@ -11717,6 +12020,7 @@ test_205() { # Job stats cmd="dd if=/dev/zero of=$DIR/$tfile bs=1M count=1 oflag=sync" verify_jobstats "$cmd" "ost1" # read + cancel_lru_locks osc cmd="dd if=$DIR/$tfile of=/dev/null bs=1M count=1 iflag=direct" verify_jobstats "$cmd" "ost1" # truncate @@ -11726,7 +12030,7 @@ test_205() { # Job stats cmd="mv -f $DIR/$tfile $DIR/$tdir.rename" verify_jobstats "$cmd" "$SINGLEMDS" # jobstats expiry - sleep until old stats should be expired - local left=$((interval_new - (SECONDS - start))) + local left=$((interval_new + 2 - (SECONDS - start))) [ $left -ge 0 ] && echo "sleep $left for expiry" && sleep $((left + 1)) cmd="mkdir $DIR/$tdir.expire" verify_jobstats "$cmd" "$SINGLEMDS" @@ -12222,16 +12526,16 @@ run_test 222a "AGL for ls should not trigger CLIO lock failure ================" test_222b () { [ $PARALLEL == "yes" ] && skip "skip parallel run" && return - rm -rf $DIR/$tdir - test_mkdir -p $DIR/$tdir - $SETSTRIPE -c 1 -i 0 $DIR/$tdir - createmany -o $DIR/$tdir/$tfile 10 - cancel_lru_locks mdc - cancel_lru_locks osc - #define OBD_FAIL_LDLM_AGL_DELAY 0x31a - $LCTL set_param fail_loc=0x31a - rm -r $DIR/$tdir || "AGL for rmdir failed" - $LCTL set_param fail_loc=0 + rm -rf $DIR/$tdir + test_mkdir -p $DIR/$tdir + $SETSTRIPE -c 1 -i 0 $DIR/$tdir + createmany -o $DIR/$tdir/$tfile 10 + cancel_lru_locks mdc + cancel_lru_locks osc + #define OBD_FAIL_LDLM_AGL_DELAY 0x31a + $LCTL set_param fail_loc=0x31a + rm -r $DIR/$tdir || error "AGL for rmdir failed" + $LCTL set_param fail_loc=0 } run_test 222b "AGL for rmdir should not trigger CLIO lock failure =============" @@ -12275,6 +12579,7 @@ run_test 224b "Don't panic on bulk IO failure" test_224c() { # LU-6441 [ $PARALLEL == "yes" ] && skip "skip parallel run" && return + remote_mds_nodsh && skip "remote MDS with nodsh" && return local pages_per_rpc=$($LCTL get_param \ osc.*.max_pages_per_rpc) local at_max=$($LCTL get_param -n at_max) @@ -12908,6 +13213,7 @@ test_230e() { mkdir $DIR/$tdir/other_dir touch $DIR/$tdir/migrate_dir/a ln $DIR/$tdir/migrate_dir/a $DIR/$tdir/other_dir/b + ls $DIR/$tdir/other_dir $LFS migrate -m 1 $DIR/$tdir/migrate_dir || error "migrate dir fails" @@ -12937,7 +13243,108 @@ test_230e() { rm -rf $DIR/$tdir || error "rm dir failed after migration" } -run_test 230e "migrate mulitple link files" +run_test 230e "migrate mulitple local link files" + +test_230f() { + [ $PARALLEL == "yes" ] && skip "skip parallel run" && return + [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return + local a_fid + local ln_fid + + mkdir -p $DIR/$tdir + mkdir $DIR/$tdir/migrate_dir + $LFS mkdir -i1 $DIR/$tdir/other_dir + touch $DIR/$tdir/migrate_dir/a + ln $DIR/$tdir/migrate_dir/a $DIR/$tdir/other_dir/ln1 + ln $DIR/$tdir/migrate_dir/a $DIR/$tdir/other_dir/ln2 + ls $DIR/$tdir/other_dir + + # a should be migrated to MDT1, since no other links on MDT0 + $LFS migrate -m 1 $DIR/$tdir/migrate_dir || + error "#1 migrate dir fails" + mdt_index=$($LFS getstripe -M $DIR/$tdir/migrate_dir) + [ $mdt_index == 1 ] || error "migrate_dir is not on MDT1" + mdt_index=$($LFS getstripe -M $DIR/$tdir/migrate_dir/a) + [ $mdt_index == 1 ] || error "a is not on MDT1" + + # a should stay on MDT1, because it is a mulitple link file + $LFS migrate -m 0 $DIR/$tdir/migrate_dir || + error "#2 migrate dir fails" + mdt_index=$($LFS getstripe -M $DIR/$tdir/migrate_dir/a) + [ $mdt_index == 1 ] || error "a is not on MDT1" + + $LFS migrate -m 1 $DIR/$tdir/migrate_dir || + error "#3 migrate dir fails" + + a_fid=$($LFS path2fid $DIR/$tdir/migrate_dir/a) + ln_fid=$($LFS path2fid $DIR/$tdir/other_dir/ln1) + [ "$a_fid" = "$ln_fid" ] || error "different fid after migrate to MDT1" + + rm -rf $DIR/$tdir/other_dir/ln1 || error "unlink ln1 fails" + rm -rf $DIR/$tdir/other_dir/ln2 || error "unlink ln2 fails" + + # a should be migrated to MDT0, since no other links on MDT1 + $LFS migrate -m 0 $DIR/$tdir/migrate_dir || + error "#4 migrate dir fails" + mdt_index=$($LFS getstripe -M $DIR/$tdir/migrate_dir/a) + [ $mdt_index == 0 ] || error "a is not on MDT0" + + rm -rf $DIR/$tdir || error "rm dir failed after migration" +} +run_test 230f "migrate mulitple remote link files" + +test_230g() { + [ $PARALLEL == "yes" ] && skip "skip parallel run" && return + [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return + + mkdir -p $DIR/$tdir/migrate_dir + + $LFS migrate -m 1000 $DIR/$tdir/migrate_dir && + error "migrating dir to non-exist MDT succeeds" + true +} +run_test 230g "migrate dir to non-exist MDT" + +test_230h() { + [ $PARALLEL == "yes" ] && skip "skip parallel run" && return + [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return + [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.7.64) ] && + skip "Need MDS version at least 2.7.64" && return + local mdt_index + + mkdir -p $DIR/$tdir/migrate_dir + + $LFS migrate -m1 $DIR && + error "migrating mountpoint1 should fail" + + $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" + +} +run_test 230h "migrate .. and root" + +test_230i() { + [ $PARALLEL == "yes" ] && skip "skip parallel run" && return + [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return + + mkdir -p $DIR/$tdir/migrate_dir + + $LFS migrate -m 1 $DIR/$tdir/migrate_dir/ || + error "migration fails with a tailing slash" + + $LFS migrate -m 0 $DIR/$tdir/migrate_dir// || + error "migration fails with two tailing slashes" +} +run_test 230i "lfs migrate -m tolerates trailing slashes" test_231a() { @@ -12950,16 +13357,20 @@ test_231a() # clear the OSC stats $LCTL set_param osc.*.stats=0 &>/dev/null + stop_writeback # Client writes $bulk_size - there must be 1 rpc for $max_pages. dd if=/dev/zero of=$DIR/$tdir/$tfile bs=$bulk_size count=1 \ oflag=direct &>/dev/null || error "dd failed" + sync; sleep 1; sync # just to be safe local nrpcs=$($LCTL get_param osc.*.stats |awk '/ost_write/ {print $2}') if [ x$nrpcs != "x1" ]; then - error "found $nrpc ost_write RPCs, not 1 as expected" + $LCTL get_param osc.*.stats + error "found $nrpcs ost_write RPCs, not 1 as expected" fi + start_writeback # Drop the OSC cache, otherwise we will read from it cancel_lru_locks osc @@ -12972,7 +13383,8 @@ test_231a() nrpcs=$($LCTL get_param osc.*.stats | awk '/ost_read/ { print $2 }') if [ x$nrpcs != "x1" ]; then - error "found $nrpc ost_read RPCs, not 1 as expected" + $LCTL get_param osc.*.stats + error "found $nrpcs ost_read RPCs, not 1 as expected" fi } run_test 231a "checking that reading/writing of BRW RPC size results in one RPC" @@ -13006,6 +13418,7 @@ run_test 232 "failed lock should not block umount" test_233a() { [ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.3.64) ] || { skip "Need MDS version at least 2.3.64"; return; } + [ -n "$FILESET" ] && skip "SKIP due to FILESET set" && return local fid=$($LFS path2fid $MOUNT) stat $MOUNT/.lustre/fid/$fid > /dev/null || @@ -13016,6 +13429,7 @@ run_test 233a "checking that OBF of the FS root succeeds" test_233b() { [ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.5.90) ] || { skip "Need MDS version at least 2.5.90"; return; } + [ -n "$FILESET" ] && skip "SKIP due to FILESET set" && return local fid=$($LFS path2fid $MOUNT/.lustre) stat $MOUNT/.lustre/fid/$fid > /dev/null || @@ -13152,8 +13566,33 @@ test_239() { } run_test 239 "osp_sync test" +test_239a() { #LU-5297 + remote_mds_nodsh && skip "remote MDS with nodsh" && return + touch $DIR/$tfile + #define OBD_FAIL_OSP_CHECK_INVALID_REC 0x2100 + do_facet $SINGLEMDS $LCTL set_param fail_loc=0x2100 + chgrp $RUNAS_GID $DIR/$tfile + wait_delete_completed +} +run_test 239a "process invalid osp sync record correctly" + +test_239b() { #LU-5297 + remote_mds_nodsh && skip "remote MDS with nodsh" && return + touch $DIR/$tfile1 + #define OBD_FAIL_OSP_CHECK_ENOMEM 0x2101 + do_facet $SINGLEMDS $LCTL set_param fail_loc=0x2101 + chgrp $RUNAS_GID $DIR/$tfile1 + wait_delete_completed + do_facet $SINGLEMDS $LCTL set_param fail_loc=0 + touch $DIR/$tfile2 + chgrp $RUNAS_GID $DIR/$tfile2 + wait_delete_completed +} +run_test 239b "process osp sync record with ENOMEM error correctly" + test_240() { [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return + remote_mds_nodsh && skip "remote MDS with nodsh" && return mkdir -p $DIR/$tdir @@ -13175,7 +13614,7 @@ 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 - cancel_lru_locks osc + cancel_lru_locks osc || true done } @@ -13208,6 +13647,7 @@ test_241b() { run_test 241b "dio vs dio" test_242() { + remote_mds_nodsh && skip "remote MDS with nodsh" && return mkdir -p $DIR/$tdir touch $DIR/$tdir/$tfile @@ -13262,6 +13702,175 @@ test_245() { } run_test 245 "check mdc connection flag/data: multiple modify RPCs" +test_246() { # LU-7371 + remote_ost_nodsh && skip "remote OST with nodsh" && return + [ $(lustre_version_code ost1) -lt $(version_code 2.7.62) ] && + skip "Need OST version >= 2.7.62" && return 0 + do_facet ost1 $LCTL set_param fail_val=4095 +#define OBD_FAIL_OST_READ_SIZE 0x234 + do_facet ost1 $LCTL set_param fail_loc=0x234 + $LFS setstripe $DIR/$tfile -i 0 -c 1 + dd if=/dev/zero of=$DIR/$tfile bs=4095 count=1 > /dev/null 2>&1 + cancel_lru_locks $FSNAME-OST0000 + dd if=$DIR/$tfile of=/dev/null bs=1048576 || error "Read failed" +} +run_test 246 "Read file of size 4095 should return right length" + +test_247a() { + lctl get_param -n mdc.$FSNAME-MDT0000*.import | + grep -q subtree || + { skip "Fileset feature is not supported"; return; } + + local submount=${MOUNT}_$tdir + + mkdir $MOUNT/$tdir + mkdir -p $submount || error "mkdir $submount failed" + FILESET="$FILESET/$tdir" mount_client $submount || + error "mount $submount failed" + echo foo > $submount/$tfile || error "write $submount/$tfile failed" + [ $(cat $MOUNT/$tdir/$tfile) = "foo" ] || + error "read $MOUNT/$tdir/$tfile failed" + umount_client $submount || error "umount $submount failed" + rmdir $submount +} +run_test 247a "mount subdir as fileset" + +test_247b() { + lctl get_param -n mdc.$FSNAME-MDT0000*.import | grep -q subtree || + { skip "Fileset feature is not supported"; return; } + + local submount=${MOUNT}_$tdir + + rm -rf $MOUNT/$tdir + mkdir -p $submount || error "mkdir $submount failed" + SKIP_FILESET=1 + FILESET="$FILESET/$tdir" mount_client $submount && + error "mount $submount should fail" + rmdir $submount +} +run_test 247b "mount subdir that dose not exist" + +test_247c() { + lctl get_param -n mdc.$FSNAME-MDT0000*.import | grep -q subtree || + { skip "Fileset feature is not supported"; return; } + + local submount=${MOUNT}_$tdir + + mkdir -p $MOUNT/$tdir/dir1 + mkdir -p $submount || error "mkdir $submount failed" + FILESET="$FILESET/$tdir" mount_client $submount || + error "mount $submount failed" + local fid=$($LFS path2fid $MOUNT/) + $LFS fid2path $submount $fid && error "fid2path should fail" + umount_client $submount || error "umount $submount failed" + rmdir $submount +} +run_test 247c "running fid2path outside root" + +test_247d() { + lctl get_param -n mdc.$FSNAME-MDT0000*.import | grep -q subtree || + { skip "Fileset feature is not supported"; return; } + + local submount=${MOUNT}_$tdir + + mkdir -p $MOUNT/$tdir/dir1 + mkdir -p $submount || error "mkdir $submount failed" + FILESET="$FILESET/$tdir" mount_client $submount || + error "mount $submount failed" + local fid=$($LFS path2fid $submount/dir1) + $LFS fid2path $submount $fid || error "fid2path should succeed" + umount_client $submount || error "umount $submount failed" + rmdir $submount +} +run_test 247d "running fid2path inside root" + +# LU-8037 +test_247e() { + lctl get_param -n mdc.$FSNAME-MDT0000*.import | + grep -q subtree || + { skip "Fileset feature is not supported"; return; } + + local submount=${MOUNT}_$tdir + + mkdir $MOUNT/$tdir + mkdir -p $submount || error "mkdir $submount failed" + FILESET="$FILESET/.." mount_client $submount && + error "mount $submount should fail" + rmdir $submount +} +run_test 247e "mount .. as fileset" + +test_248() { + local my_error=error + + local fast_read_sav=$($LCTL get_param -n llite.*.fast_read 2>/dev/null) + [ -z "$fast_read_sav" ] && skip "no fast read support" && return + + # This test case is time sensitive and Maloo uses KVM to run autotest. + # Therefore the complete time of I/O task is unreliable and depends on + # the workload on the host machine when the task is running. + local virt=$(running_in_vm) + [ -n "$virt" ] && echo "running in VM '$virt', ignore error" && + my_error="error_ignore env=$virt" + + # create a large file for fast read verification + dd if=/dev/zero of=$DIR/$tfile bs=1M count=128 > /dev/null 2>&1 + + # make sure the file is created correctly + $CHECKSTAT -s $((128*1024*1024)) $DIR/$tfile || + { rm -f $DIR/$tfile; skip "file creation error" && return; } + + echo "Test 1: verify that fast read is 4 times faster on cache read" + + # small read with fast read enabled + $LCTL set_param -n llite.*.fast_read=1 + local t_fast=$(dd if=$DIR/$tfile of=/dev/null bs=4k 2>&1 | + awk '/copied/ { print $6 }') + + # small read with fast read disabled + $LCTL set_param -n llite.*.fast_read=0 + local t_slow=$(dd if=$DIR/$tfile of=/dev/null bs=4k 2>&1 | + awk '/copied/ { print $6 }') + + # verify that fast read is 4 times faster for cache read + [ $(bc <<< "4 * $t_fast < $t_slow") -eq 1 ] || + $my_error "fast read was not 4 times faster: $t_fast vs $t_slow" + + echo "Test 2: verify the performance between big and small read" + $LCTL set_param -n llite.*.fast_read=1 + + # 1k non-cache read + cancel_lru_locks osc + local t_1k=$(dd if=$DIR/$tfile of=/dev/null bs=1k 2>&1 | + awk '/copied/ { print $6 }') + + # 1M non-cache read + cancel_lru_locks osc + local t_1m=$(dd if=$DIR/$tfile of=/dev/null bs=1k 2>&1 | + awk '/copied/ { print $6 }') + + # verify that big IO is not 4 times faster than small IO + [ $(bc <<< "4 * $t_1k >= $t_1m") -eq 1 ] || + $my_error "bigger IO is way too fast: $t_1k vs $t_1m" + + $LCTL set_param -n llite.*.fast_read=$fast_read_sav + rm -f $DIR/$tfile +} +run_test 248 "fast read verification" + +test_249() { # LU-7890 + rm -f $DIR/$tfile + $SETSTRIPE -c 1 $DIR/$tfile + + [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.8.53) ] && + skip "Need at least version 2.8.54" + + # Offset 2T == 4k * 512M + dd if=/dev/zero of=$DIR/$tfile bs=4k count=1 seek=512M || + error "dd to 2T offset failed" +} +run_test 249 "Write above 2T file size" + test_250() { [ "$(facet_fstype ost$(($($GETSTRIPE -i $DIR/$tfile) + 1)))" = "zfs" ] \ && skip "no 16TB file size limit on ZFS" && return @@ -13301,6 +13910,8 @@ test_252() { local num local gen + remote_mds_nodsh && skip "remote MDS with nodsh" && return + remote_ost_nodsh && skip "remote OST with nodsh" && return if [ "$(facet_fstype ost1)" != "ldiskfs" -o \ "$(facet_fstype mds1)" != "ldiskfs" ]; then skip "can only run lr_reader on ldiskfs target" @@ -13344,6 +13955,287 @@ test_252() { } run_test 252 "check lr_reader tool" +test_254() { + local cl_user + + [ $PARALLEL == "yes" ] && skip "skip parallel run" && return + remote_mds_nodsh && skip "remote MDS with nodsh" && return + do_facet mds1 $LCTL get_param -n mdd.$MDT0.changelog_size || + { skip "MDS does not support changelog_size" && return; } + + cl_user=$(do_facet mds1 $LCTL --device $MDT0 changelog_register -n) + echo "Registered as changelog user $cl_user" + + $LFS changelog_clear $MDT0 $cl_user 0 + + local size1=$(do_facet mds1 \ + $LCTL get_param -n mdd.$MDT0.changelog_size) + echo "Changelog size $size1" + + rm -rf $DIR/$tdir + $LFS mkdir -i 0 $DIR/$tdir + # change something + mkdir -p $DIR/$tdir/pics/2008/zachy + touch $DIR/$tdir/pics/2008/zachy/timestamp + cp /etc/hosts $DIR/$tdir/pics/2008/zachy/pic1.jpg + mv $DIR/$tdir/pics/2008/zachy $DIR/$tdir/pics/zach + ln $DIR/$tdir/pics/zach/pic1.jpg $DIR/$tdir/pics/2008/portland.jpg + ln -s $DIR/$tdir/pics/2008/portland.jpg $DIR/$tdir/pics/desktop.jpg + rm $DIR/$tdir/pics/desktop.jpg + + local size2=$(do_facet mds1 \ + $LCTL get_param -n mdd.$MDT0.changelog_size) + echo "Changelog size after work $size2" + + do_facet mds1 $LCTL --device $MDT0 changelog_deregister $cl_user + + if (( size2 <= size1 )); then + error "Changelog size after work should be greater than original" + fi + return 0 +} +run_test 254 "Check changelog size" + +ladvise_no_type() +{ + local type=$1 + local file=$2 + + lfs ladvise -a invalid $file 2>&1 | grep "Valid types" | + awk -F: '{print $2}' | grep $type > /dev/null + if [ $? -ne 0 ]; then + return 0 + fi + return 1 +} + +ladvise_no_ioctl() +{ + local file=$1 + + lfs ladvise -a willread $file > /dev/null 2>&1 + if [ $? -eq 0 ]; then + return 1 + fi + + lfs ladvise -a willread $file 2>&1 | + grep "Inappropriate ioctl for device" > /dev/null + if [ $? -eq 0 ]; then + return 0 + fi + return 1 +} + +ladvise_willread_performance() +{ + local repeat=10 + local average_cache=0 + local average_ladvise=0 + for ((i = 1; i <= $repeat; i++)); do + echo "Iter $i/$repeat: reading without willread hint" + cancel_lru_locks osc + do_nodes $(comma_list $(osts_nodes)) \ + "echo 3 > /proc/sys/vm/drop_caches" + local speed_origin=$($READS -f $DIR/$tfile -s $size \ + -b 4096 -n $((size / 4096)) -t 60 | + sed -e '/^$/d' -e 's#.*s, ##' -e 's#MB/s##') + + echo "Iter $i/$repeat: Reading again without willread hint" + cancel_lru_locks osc + local speed_cache=$($READS -f $DIR/$tfile -s $size \ + -b 4096 -n $((size / 4096)) -t 60 | + sed -e '/^$/d' -e 's#.*s, ##' -e 's#MB/s##') + + echo "Iter $i/$repeat: reading with willread hint" + cancel_lru_locks osc + do_nodes $(comma_list $(osts_nodes)) \ + "echo 3 > /proc/sys/vm/drop_caches" + lfs ladvise -a willread $DIR/$tfile || + error "Ladvise failed" + local speed_ladvise=$($READS -f $DIR/$tfile -s $size \ + -b 4096 -n $((size / 4096)) -t 60 | + sed -e '/^$/d' -e 's#.*s, ##' -e 's#MB/s##') + + local cache_speedup=$(echo "scale=2; \ + ($speed_cache-$speed_origin)/$speed_origin*100" | bc) + cache_speedup=$(echo ${cache_speedup%.*}) + echo "Iter $i/$repeat: cache speedup: $cache_speedup%" + average_cache=$((average_cache + cache_speedup)) + + local ladvise_speedup=$(echo "scale=2; \ + ($speed_ladvise-$speed_origin)/$speed_origin*100" | bc) + ladvise_speedup=$(echo ${ladvise_speedup%.*}) + echo "Iter $i/$repeat: ladvise speedup: $ladvise_speedup%" + average_ladvise=$((average_ladvise + ladvise_speedup)) + done + average_cache=$((average_cache / repeat)) + average_ladvise=$((average_ladvise / repeat)) + + if [ $average_cache -lt 20 ]; then + echo "Speedup with cache is less than 20% ($average_cache%),"\ + "skipping check of speedup with willread:"\ + "$average_ladvise%" + return 0 + fi + + local lowest_speedup=$((average_cache / 2)) + [ $average_ladvise -gt $lowest_speedup ] || + error "Speedup with willread is less than $lowest_speedup%,"\ + "got $average_ladvise%" + echo "Speedup with willread ladvise: $average_ladvise%" + echo "Speedup with cache: $average_cache%" +} + +test_255a() { + lfs setstripe -c -1 -i 0 $DIR/$tfile || error "$tfile failed" + + ladvise_no_type willread $DIR/$tfile && + skip "willread ladvise is not supported" && return + + ladvise_no_ioctl $DIR/$tfile && + skip "ladvise ioctl is not supported" && return + + [ $(lustre_version_code ost1) -lt $(version_code 2.8.54) ] && + skip "lustre < 2.8.54 does not support ladvise " && return + + local size_mb=100 + local size=$((size_mb * 1048576)) + dd if=/dev/zero of=$DIR/$tfile bs=1048576 count=$size_mb || + error "dd to $DIR/$tfile failed" + + lfs ladvise -a willread $DIR/$tfile || + error "Ladvise failed with no range argument" + + lfs ladvise -a willread -s 0 $DIR/$tfile || + error "Ladvise failed with no -l or -e argument" + + lfs ladvise -a willread -e 1 $DIR/$tfile || + error "Ladvise failed with only -e argument" + + lfs ladvise -a willread -l 1 $DIR/$tfile || + error "Ladvise failed with only -l argument" + + lfs ladvise -a willread -s 2 -e 1 $DIR/$tfile && + error "End offset should not be smaller than start offset" + + lfs ladvise -a willread -s 2 -e 2 $DIR/$tfile && + error "End offset should not be equal to start offset" + + lfs ladvise -a willread -s $size -l 1 $DIR/$tfile || + error "Ladvise failed with overflowing -s argument" + + lfs ladvise -a willread -s 1 -e $((size + 1)) $DIR/$tfile || + error "Ladvise failed with overflowing -e argument" + + lfs ladvise -a willread -s 1 -l $size $DIR/$tfile || + error "Ladvise failed with overflowing -l argument" + + lfs ladvise -a willread -l 1 -e 2 $DIR/$tfile && + error "Ladvise succeeded with conflicting -l and -e arguments" + + echo "Synchronous ladvise should wait" + local delay=4 +#define OBD_FAIL_OST_LADVISE_PAUSE 0x237 + do_nodes $(comma_list $(osts_nodes)) \ + $LCTL set_param fail_val=$delay fail_loc=0x237 + + local start_ts=$SECONDS + lfs ladvise -a willread $DIR/$tfile || + error "Ladvise failed with no range argument" + local end_ts=$SECONDS + local inteval_ts=$((end_ts - start_ts)) + + if [ $inteval_ts -lt $(($delay - 1)) ]; then + error "Synchronous advice didn't wait reply" + fi + + echo "Asynchronous ladvise shouldn't wait" + local start_ts=$SECONDS + lfs ladvise -a willread -b $DIR/$tfile || + error "Ladvise failed with no range argument" + local end_ts=$SECONDS + local inteval_ts=$((end_ts - start_ts)) + + if [ $inteval_ts -gt $(($delay / 2)) ]; then + error "Asynchronous advice blocked" + fi + + do_nodes $(comma_list $(osts_nodes)) $LCTL set_param fail_loc=0 + ladvise_willread_performance +} +run_test 255a "check 'lfs ladvise -a willread'" + +test_256() { + local cl_user + local cat_sl + local mdt_dev + + [ $PARALLEL == "yes" ] && skip "skip parallel run" && return + remote_mds_nodsh && skip "remote MDS with nodsh" && return + [ "$(facet_fstype mds1)" != "ldiskfs" ] && + skip "non-ldiskfs backend" && return + + mdt_dev=$(mdsdevname 1) + echo $mdt_dev + cl_user=$(do_facet mds1 \ + "$LCTL get_param -n mdd.$MDT0.changelog_users | grep cl") + if [[ -n $cl_user ]]; then + skip "active changelog user" + return + fi + + cl_user=$(do_facet mds1 $LCTL --device $MDT0 changelog_register -n) + echo "Registered as changelog user $cl_user" + + rm -rf $DIR/$tdir + mkdir -p $DIR/$tdir + + $LFS changelog_clear $MDT0 $cl_user 0 + + # change something + touch $DIR/$tdir/{1..10} + + # stop the MDT + stop mds1 || error "Fail to stop MDT." + + # remount the MDT + start mds1 $mdt_dev $MDS_MOUNT_OPTS || error "Fail to start MDT." + + #after mount new plainllog is used + touch $DIR/$tdir/{11..19} + cat_sl=$(do_facet mds1 \ + "$DEBUGFS -R \\\"dump changelog_catalog cat.dmp\\\" $mdt_dev; \ + llog_reader cat.dmp | grep \\\"type=1064553b\\\" | wc -l") + + if (( cat_sl != 2 )); then + do_facet mds1 $LCTL --device $MDT0 changelog_deregister $cl_user + error "Changelog catalog has wrong number of slots $cat_sl" + fi + + $LFS changelog_clear $MDT0 $cl_user 0 + + cat_sl=$(do_facet mds1 \ + "$DEBUGFS -R \\\"dump changelog_catalog cat.dmp\\\" $mdt_dev; \ + llog_reader cat.dmp | grep \\\"type=1064553b\\\" | wc -l") + + do_facet mds1 $LCTL --device $MDT0 changelog_deregister $cl_user + + if (( cat_sl == 2 )); then + error "Empty plain llog was not deleted from changelog catalog" + fi + if (( cat_sl != 1 )); then + error "Active plain llog shouldn\`t be deleted from catalog" + fi +} +run_test 256 "Check llog delete for empty and not full state" + +test_260() { +#define OBD_FAIL_MDC_CLOSE 0x806 + $LCTL set_param fail_loc=0x80000806 + touch $DIR/$tfile + +} +run_test 260 "Check mdc_close fail" cleanup_test_300() { trap 0 @@ -13414,6 +14306,8 @@ test_striped_dir() { } test_300a() { + [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.7.0) ] && + skip "skipped for lustre < 2.7.0" && return [ $PARALLEL == "yes" ] && skip "skip parallel run" && return [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return @@ -13423,6 +14317,8 @@ test_300a() { run_test 300a "basic striped dir sanity test" test_300b() { + [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.7.0) ] && + skip "skipped for lustre < 2.7.0" && return [ $PARALLEL == "yes" ] && skip "skip parallel run" && return [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return local i @@ -13453,6 +14349,8 @@ test_300b() { run_test 300b "check ctime/mtime for striped dir" test_300c() { + [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.7.0) ] && + skip "skipped for lustre < 2.7.0" && return [ $PARALLEL == "yes" ] && skip "skip parallel run" && return [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return local file_count @@ -13476,6 +14374,8 @@ test_300c() { run_test 300c "chown && check ls under striped directory" test_300d() { + [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.7.0) ] && + skip "skipped for lustre < 2.7.0" && return [ $PARALLEL == "yes" ] && skip "skip parallel run" && return [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return local stripe_count @@ -13507,6 +14407,8 @@ test_300d() { run_test 300d "check default stripe under striped directory" test_300e() { + [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.7.55) ] && + skip "Need MDS version at least 2.7.55" && return [ $PARALLEL == "yes" ] && skip "skip parallel run" && return [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return local stripe_count @@ -13551,6 +14453,8 @@ test_300e() { run_test 300e "check rename under striped directory" test_300f() { + [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.7.55) ] && + skip "Need MDS version at least 2.7.55" && return [ $PARALLEL == "yes" ] && skip "skip parallel run" && return [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return local stripe_count @@ -13638,6 +14542,8 @@ test_300_check_default_striped_dir() } test_300g() { + [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.7.55) ] && + skip "Need MDS version at least 2.7.55" && return [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return local dir local stripe_count @@ -13669,6 +14575,8 @@ test_300g() { run_test 300g "check default striped directory for normal directory" test_300h() { + [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.7.55) ] && + skip "Need MDS version at least 2.7.55" && return [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return local dir local stripe_count @@ -13697,6 +14605,8 @@ test_300h() { run_test 300h "check default striped directory for striped directory" test_300i() { + [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.7.55) ] && + skip "Need MDS version at least 2.7.55" && return [ $PARALLEL == "yes" ] && skip "skip parallel run" && return [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return local stripe_count @@ -13763,9 +14673,9 @@ test_300j() { run_test 300j "test large update record" test_300k() { - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.7.55) ] && skip "Need MDS version at least 2.7.55" && return + [ $PARALLEL == "yes" ] && skip "skip parallel run" && return [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return local stripe_count local file @@ -13788,6 +14698,8 @@ test_300k() { run_test 300k "test large striped directory" test_300l() { + [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.7.55) ] && + skip "Need MDS version at least 2.7.55" && return [ $PARALLEL == "yes" ] && skip "skip parallel run" && return [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return local stripe_index @@ -13809,6 +14721,8 @@ test_300l() { run_test 300l "non-root user to create dir under striped dir with stale layout" test_300m() { + [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.7.55) ] && + skip "Need MDS version at least 2.7.55" && return [ $PARALLEL == "yes" ] && skip "skip parallel run" && return [ $MDSCOUNT -ge 2 ] && skip "Only for single MDT" && return @@ -13849,6 +14763,8 @@ cleanup_300n() { } test_300n() { + [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.7.55) ] && + skip "Need MDS version at least 2.7.55" && return [ $PARALLEL == "yes" ] && skip "skip parallel run" && return [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return local stripe_index @@ -13887,6 +14803,83 @@ test_300n() { } run_test 300n "non-root user to create dir under striped dir with default EA" +test_300o() { + [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.7.55) ] && + skip "Need MDS version at least 2.7.55" && return + [ $PARALLEL == "yes" ] && skip "skip parallel run" && return + [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return + local numfree1 + local numfree2 + + mkdir -p $DIR/$tdir + + numfree1=$(lctl get_param -n mdc.*MDT0000*.filesfree) + numfree2=$(lctl get_param -n mdc.*MDT0001*.filesfree) + if [ $numfree1 -lt 66000 -o $numfree2 -lt 66000 ]; then + skip "not enough free inodes $numfree1 $numfree2" + return + fi + + numfree1=$(lctl get_param -n mdc.*MDT0000-mdc-*.kbytesfree) + numfree2=$(lctl get_param -n mdc.*MDT0001-mdc-*.kbytesfree) + if [ $numfree1 -lt 300000 -o $numfree2 -lt 300000 ]; then + skip "not enough free space $numfree1 $numfree2" + return + fi + + $LFS setdirstripe -c2 $DIR/$tdir/striped_dir || + error "setdirstripe fails" + + createmany -d $DIR/$tdir/striped_dir/d 131000 || + error "create dirs fails" + + $LCTL set_param ldlm.namespaces.*mdc-*.lru_size=0 + ls $DIR/$tdir/striped_dir > /dev/null || + error "ls striped dir fails" + unlinkmany -d $DIR/$tdir/striped_dir/d 131000 || + error "unlink big striped dir fails" +} +run_test 300o "unlink big sub stripe(> 65000 subdirs)" + +test_300p() { + [ $PARALLEL == "yes" ] && skip "skip parallel run" && return + [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return + remote_mds_nodsh && skip "remote MDS with nodsh" && return + + mkdir -p $DIR/$tdir + + #define OBD_FAIL_OUT_ENOSPC 0x1704 + do_facet mds2 lctl set_param fail_loc=0x80001704 + $LFS setdirstripe -c2 $DIR/$tdir/bad_striped_dir > /dev/null 2>&1 && + error "create striped directory should fail" + + [ -e $DIR/$tdir/bad_striped_dir ] && error "striped dir exists" + + $LFS setdirstripe -c2 $DIR/$tdir/bad_striped_dir + true +} +run_test 300p "create striped directory without space" + +test_300q() { + [ $PARALLEL == "yes" ] && skip "skip parallel run" && return + [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return + + local fd=$(free_fd) + local cmd="exec $fd<$tdir" + cd $DIR + $LFS mkdir -c $MDSCOUNT $tdir || error "create $tdir fails" + eval $cmd + cmd="exec $fd<&-" + trap "eval $cmd" EXIT + cd $tdir || error "cd $tdir fails" + rmdir ../$tdir || error "rmdir $tdir fails" + mkdir local_dir && error "create dir succeeds" + $LFS setdirstripe -i1 remote_dir && error "create remote dir succeeds" + eval $cmd + return 0 +} +run_test 300q "create remote directory under orphan directory" + prepare_remote_file() { mkdir $DIR/$tdir/src_dir || error "create remote source failed" @@ -13961,6 +14954,154 @@ test_310c() { } run_test 310c "open-unlink remote file with multiple links" +#LU-4825 +test_311() { + [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.8.54) ] && + skip "lustre < 2.8.54 does not contain LU-4825 fix" && return + [ $PARALLEL == "yes" ] && skip "skip parallel run" && return + + local old_iused=$($LFS df -i | grep OST0000 | awk '{ print $3 }') + + mkdir -p $DIR/$tdir + $SETSTRIPE -i 0 -c 1 $DIR/$tdir + createmany -o $DIR/$tdir/$tfile. 1000 + + # statfs data is not real time, let's just calculate it + old_iused=$((old_iused + 1000)) + + local count=$(do_facet $SINGLEMDS "lctl get_param -n \ + osp.*OST0000*MDT0000.create_count") + local max_count=$(do_facet $SINGLEMDS "lctl get_param -n \ + osp.*OST0000*MDT0000.max_create_count") + for idx in $(seq $MDSCOUNT); do + do_facet mds$idx "lctl set_param -n \ + osp.*OST0000*MDT000?.max_create_count=0" + done + + $SETSTRIPE -i 0 $DIR/$tdir/$tfile || error "setstripe failed" + local index=$($GETSTRIPE -i $DIR/$tdir/$tfile) + [ $index -ne 0 ] || error "$tfile stripe index is 0" + + unlinkmany $DIR/$tdir/$tfile. 1000 + + for idx in $(seq $MDSCOUNT); do + do_facet mds$idx "lctl set_param -n \ + osp.*OST0000*MDT000?.max_create_count=$max_count" + do_facet mds$idx "lctl set_param -n \ + osp.*OST0000*MDT000?.create_count=$count" + done + + local new_iused + for i in $(seq 120); do + new_iused=$($LFS df -i | grep OST0000 | awk '{ print $3 }') + [ $((old_iused - new_iused)) -gt 800 ] && break + sleep 1 + done + + echo "waited $i sec, old Iused $old_iused, new Iused $new_iused" + [ $((old_iused - new_iused)) -gt 800 ] || + error "objs not destroyed after unlink" +} +run_test 311 "disable OSP precreate, and unlink should destroy objs" + +zfs_oid_to_objid() +{ + local ost=$1 + local objid=$2 + + local vdevdir=$(dirname $(facet_vdevice $ost)) + local cmd="$ZDB -e -p $vdevdir -dddd $(facet_device $ost)" + local zfs_zapid=$(do_facet $ost $cmd | + grep -w "/O/0/d$((objid%32))" -C 5 | + awk '/Object/{getline; print $1}') + local zfs_objid=$(do_facet $ost $cmd $zfs_zapid | + awk "/$objid = /"'{printf $3}') + + echo $zfs_objid +} + +zfs_object_blksz() { + local ost=$1 + local objid=$2 + + local vdevdir=$(dirname $(facet_vdevice $ost)) + local cmd="$ZDB -e -p $vdevdir -dddd $(facet_device $ost)" + local blksz=$(do_facet $ost $cmd $objid | + awk '/dblk/{getline; printf $4}') + + case "${blksz: -1}" in + k|K) blksz=$((${blksz:0:$((${#blksz} - 1))}*1024)) ;; + m|M) blksz=$((${blksz:0:$((${#blksz} - 1))}*1024*1024)) ;; + *) ;; + esac + + echo $blksz +} + +test_312() { # LU-4856 + [ $(facet_fstype ost1) = "zfs" ] || + { skip "the test only applies to zfs" && return; } + + local max_blksz=$(do_facet ost1 \ + $ZFS get -p recordsize $(facet_device ost1) | + awk '!/VALUE/{print $3}') + + # to make life a little bit easier + $LFS mkdir -c 1 -i 0 $DIR/$tdir + $LFS setstripe -c 1 -i 0 $DIR/$tdir + + local tf=$DIR/$tdir/$tfile + touch $tf + local oid=$($LFS getstripe $tf | awk '/obdidx/{getline; print $2}') + + # Get ZFS object id + local zfs_objid=$(zfs_oid_to_objid ost1 $oid) + + # block size change by sequential over write + local blksz + for ((bs=4096; bs <= max_blksz; bs <<= 2)); do + dd if=/dev/zero of=$tf bs=$bs count=1 oflag=sync conv=notrunc + + blksz=$(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=4K count=1 oflag=sync conv=notrunc + oid=$($LFS getstripe $tf | awk '/obdidx/{getline; print $2}') + zfs_objid=$(zfs_oid_to_objid ost1 $oid) + + for ((count = 1; count < $((max_blksz / 4096)); count *= 2)); do + dd if=/dev/zero of=$tf bs=4K count=$count seek=$count \ + oflag=sync conv=notrunc + + blksz=$(zfs_object_blksz ost1 $zfs_objid) + blksz=$((blksz / 8192)) # in 2*4K unit + [ $blksz -eq $count ] || + error "blksz error(in 8k): $blksz, expected: $count" + done + rm -f $tf + + # random write + touch $tf + oid=$($LFS getstripe $tf | awk '/obdidx/{getline; print $2}') + zfs_objid=$(zfs_oid_to_objid ost1 $oid) + + dd if=/dev/zero of=$tf bs=8K count=1 oflag=sync conv=notrunc + blksz=$(zfs_object_blksz ost1 $zfs_objid) + [ $blksz -eq 8192 ] || error "blksz error: $blksz, expected: 8k" + + dd if=/dev/zero of=$tf bs=64K count=1 oflag=sync conv=notrunc seek=128 + blksz=$(zfs_object_blksz ost1 $zfs_objid) + [ $blksz -eq 65536 ] || error "blksz error: $blksz, expected: 64k" + + dd if=/dev/zero of=$tf bs=1M count=1 oflag=sync conv=notrunc + blksz=$(zfs_object_blksz ost1 $zfs_objid) + [ $blksz -eq 65536 ] || error "rewrite error: $blksz, expected: 64k" +} +run_test 312 "make sure ZFS adjusts its block size by write pattern" + test_400a() { # LU-1606, was conf-sanity test_74 local extra_flags='' local out=$TMP/$tfile @@ -13983,6 +15124,7 @@ test_400a() { # LU-1606, was conf-sanity test_74 $CC -Wall -Werror $extra_flags -llustreapi -o $out $prog || error "client api broken" done + rm -f $out } run_test 400a "Lustre client api program can compile and link" @@ -14015,9 +15157,281 @@ test_400b() { # LU-1606, LU-5011 $CC -Wall -Werror -include $header -c -x c /dev/null -o $out || error "cannot compile '$header'" done + rm -f $out } run_test 400b "packaged headers can be compiled" +test_401a() { #LU-7437 + local printf_arg=$(find -printf 2>&1 | grep "unrecognized:") + [ -n "$printf_arg" ] && skip_env "find does not support -printf" && + return + #count the number of parameters by "list_param -R" + local params=$($LCTL list_param -R '*' 2>/dev/null | wc -l) + #count the number of parameters by listing proc files + local procs=$(find -L $proc_dirs -mindepth 1 -printf '%P\n' 2>/dev/null| + sort -u | wc -l) + + [ $params -eq $procs ] || + error "found $params parameters vs. $procs proc files" + + # test the list_param -D option only returns directories + params=$($LCTL list_param -R -D '*' 2>/dev/null | wc -l) + #count the number of parameters by listing proc directories + procs=$(find -L $proc_dirs -mindepth 1 -type d -printf '%P\n' 2>/dev/null | + sort -u | wc -l) + + [ $params -eq $procs ] || + error "found $params parameters vs. $procs proc files" +} +run_test 401a "Verify if 'lctl list_param -R' can list parameters recursively" + +test_401b() { + local save=$($LCTL get_param -n jobid_var) + local tmp=testing + + $LCTL set_param foo=bar jobid_var=$tmp bar=baz && + error "no error returned when setting bad parameters" + + local jobid_new=$($LCTL get_param -n foe jobid_var baz) + [[ "$jobid_new" == "$tmp" ]] || error "jobid tmp $jobid_new != $tmp" + + $LCTL set_param -n fog=bam jobid_var=$save bat=fog + local jobid_old=$($LCTL get_param -n foe jobid_var bag) + [[ "$jobid_old" == "$save" ]] || error "jobid new $jobid_old != $save" +} +run_test 401b "Verify 'lctl {get,set}_param' continue after error" + +test_401c() { + local jobid_var_old=$($LCTL get_param -n jobid_var) + local jobid_var_new + + $LCTL set_param jobid_var= && + error "no error returned for 'set_param a='" + + jobid_var_new=$($LCTL get_param -n jobid_var) + [[ "$jobid_var_old" == "$jobid_var_new" ]] || + error "jobid_var was changed by setting without value" + + $LCTL set_param jobid_var && + error "no error returned for 'set_param a'" + + jobid_var_new=$($LCTL get_param -n jobid_var) + [[ "$jobid_var_old" == "$jobid_var_new" ]] || + error "jobid_var was changed by setting without value" +} +run_test 401c "Verify 'lctl set_param' without value fails in either format." + +test_401d() { + local jobid_var_old=$($LCTL get_param -n jobid_var) + local jobid_var_new + local new_value="foo=bar" + + $LCTL set_param jobid_var=$new_value || + error "'set_param a=b' did not accept a value containing '='" + + jobid_var_new=$($LCTL get_param -n jobid_var) + [[ "$jobid_var_new" == "$new_value" ]] || + error "'set_param a=b' failed on a value containing '='" + + # Reset the jobid_var to test the other format + $LCTL set_param jobid_var=$jobid_var_old + jobid_var_new=$($LCTL get_param -n jobid_var) + [[ "$jobid_var_new" == "$jobid_var_old" ]] || + error "failed to reset jobid_var" + + $LCTL set_param jobid_var $new_value || + error "'set_param a b' did not accept a value containing '='" + + jobid_var_new=$($LCTL get_param -n jobid_var) + [[ "$jobid_var_new" == "$new_value" ]] || + error "'set_param a b' failed on a value containing '='" + + $LCTL set_param jobid_var $jobid_var_old + jobid_var_new=$($LCTL get_param -n jobid_var) + [[ "$jobid_var_new" == "$jobid_var_old" ]] || + error "failed to reset jobid_var" +} +run_test 401d "Verify 'lctl set_param' accepts values containing '='" + +test_402() { + remote_mds_nodsh && skip "remote MDS with nodsh" && return + $LFS setdirstripe -i 0 $DIR/$tdir || error "setdirstripe -i 0 failed" +#define OBD_FAIL_MDS_FLD_LOOKUP 0x15c + do_facet mds1 "lctl set_param fail_loc=0x8000015c" + touch $DIR/$tdir/$tfile && error "touch should fail with ENOENT" || + echo "Touch failed - OK" +} +run_test 402 "Return ENOENT to lod_generate_and_set_lovea" + +test_403() { + local file1=$DIR/$tfile.1 + local file2=$DIR/$tfile.2 + local tfile=$TMP/$tfile + + rm -f $file1 $file2 $tfile + + touch $file1 + ln $file1 $file2 + + # 30 sec OBD_TIMEOUT in ll_getattr() + # right before populating st_nlink + $LCTL set_param fail_loc=0x80001409 + stat -c %h $file1 > $tfile & + + # create an alias, drop all locks and reclaim the dentry + < $file2 + cancel_lru_locks mdc + cancel_lru_locks osc + sysctl -w vm.drop_caches=2 + + wait + + [ `cat $tfile` -gt 0 ] || error "wrong nlink count: `cat $tfile`" + + rm -f $tfile $file1 $file2 +} +run_test 403 "i_nlink should not drop to zero due to aliasing" + +test_404() { # LU-6601 + remote_mds_nodsh && skip "remote MDS with nodsh" && return + local mosps=$(do_facet $SINGLEMDS $LCTL dl | + awk '/osp .*-osc-MDT/ { print $4}') + + local osp + for osp in $mosps; do + echo "Deactivate: " $osp + do_facet $SINGLEMDS $LCTL --device %$osp deactivate + local stat=$(do_facet $SINGLEMDS $LCTL dl | + awk -vp=$osp '$4 == p { print $2 }') + [ $stat = IN ] || { + do_facet $SINGLEMDS $LCTL dl | grep -w $osp + error "deactivate error" + } + echo "Activate: " $osp + do_facet $SINGLEMDS $LCTL --device %$osp activate + local stat=$(do_facet $SINGLEMDS $LCTL dl | + awk -vp=$osp '$4 == p { print $2 }') + [ $stat = UP ] || { + do_facet $SINGLEMDS $LCTL dl | grep -w $osp + error "activate error" + } + done +} +run_test 404 "validate manual {de}activated works properly for OSPs" + +test_405() { + [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.6.92) ] && + skip "Layout swap lock is not supported" && return + + check_swap_layouts_support && return 0 + + test_mkdir -p $DIR/$tdir + swap_lock_test -d $DIR/$tdir || + error "One layout swap locked test failed" +} +run_test 405 "Various layout swap lock tests" + +test_406() { + [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return + [ $OSTCOUNT -lt 2 ] && skip "needs >= 2 OSTs" && return + [ -n "$FILESET" ] && skip "SKIP due to FILESET set" && return + [ $PARALLEL == "yes" ] && skip "skip parallel run" && return + [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.8.50) ] && + skip "Need MDS version at least 2.8.50" && return + + local def_stripenr=$($GETSTRIPE -c $MOUNT) + local def_stripe_size=$($GETSTRIPE -S $MOUNT) + local def_stripe_offset=$($GETSTRIPE -i $MOUNT) + local def_pool=$($GETSTRIPE -p $MOUNT) + + local test_pool=$TESTNAME + pool_add $test_pool || error "pool_add failed" + pool_add_targets $test_pool 0 $(($OSTCOUNT - 1)) 1 || + error "pool_add_targets failed" + + # parent set default stripe count only, child will stripe from both + # parent and fs default + $SETSTRIPE -c 1 -i 1 -S $((def_stripe_size * 2)) -p $test_pool $MOUNT || + error "setstripe $MOUNT failed" + $LFS mkdir -c $MDSCOUNT $DIR/$tdir || error "mkdir $tdir failed" + $SETSTRIPE -c $OSTCOUNT $DIR/$tdir || error "setstripe $tdir failed" + for i in $(seq 10); do + local f=$DIR/$tdir/$tfile.$i + touch $f || error "touch failed" + local count=$($GETSTRIPE -c $f) + [ $count -eq $OSTCOUNT ] || + error "$f stripe count $count != $OSTCOUNT" + local offset=$($GETSTRIPE -i $f) + [ $offset -eq 1 ] || error "$f stripe offset $offset != 1" + local size=$($GETSTRIPE -S $f) + [ $size -eq $((def_stripe_size * 2)) ] || + error "$f stripe size $size != $((def_stripe_size * 2))" + local pool=$($GETSTRIPE -p $f) + [ $pool == $test_pool ] || error "$f pool $pool != $test_pool" + done + + # change fs default striping, delete parent default striping, now child + # will stripe from new fs default striping only + $SETSTRIPE -c 1 -S $def_stripe_size -i 0 $MOUNT || + error "change $MOUNT default stripe failed" + $SETSTRIPE -c 0 $DIR/$tdir || error "delete $tdir default stripe failed" + for i in $(seq 11 20); do + local f=$DIR/$tdir/$tfile.$i + touch $f || error "touch $f failed" + local count=$($GETSTRIPE -c $f) + [ $count -eq 1 ] || error "$f stripe count $count != 1" + local offset=$($GETSTRIPE -i $f) + [ $offset -eq 0 ] || error "$f stripe offset $offset != 0" + local size=$($GETSTRIPE -S $f) + [ $size -eq $def_stripe_size ] || + error "$f stripe size $size != $def_stripe_size" + local pool=$($GETSTRIPE -p $f) + [ "#$pool" == "#" ] || error "$f pool $pool is set" + + done + + unlinkmany $DIR/$tdir/$tfile. 1 20 + + # restore FS default striping + if [ -z $def_pool ]; then + $SETSTRIPE -c $def_stripenr -S $def_stripe_size \ + -i $def_stripe_offset $MOUNT || + error "restore default striping failed" + else + $SETSTRIPE -c $def_stripenr -S $def_stripe_size -p $def_pool \ + -i $def_stripe_offset $MOUNT || + error "restore default striping with $def_pool failed" + fi + + local f=$DIR/$tdir/$tfile + pool_remove_all_targets $test_pool $f + pool_remove $test_pool $f +} +run_test 406 "DNE support fs default striping" + +test_407() { + [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return + + [[ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.8.55) ]] && + skip "Need MDS version at least 2.8.55" && return + + $LFS mkdir -i 0 -c 1 $DIR/$tdir.0 || + error "$LFS mkdir -i 0 -c 1 $tdir.0 failed" + $LFS mkdir -i 1 -c 1 $DIR/$tdir.1 || + error "$LFS mkdir -i 1 -c 1 $tdir.1 failed" + touch $DIR/$tdir.0/$tfile.0 || error "touch $tdir.0/$tfile.0 failed" + + #define OBD_FAIL_DT_TXN_STOP 0x2019 + for idx in $(seq $MDSCOUNT); do + do_facet mds$idx "lctl set_param fail_loc=0x2019" + done + $LFS mkdir -c 2 $DIR/$tdir && error "$LFS mkdir -c 2 $tdir should fail" + mv $DIR/$tdir.0/$tfile.0 $DIR/$tdir.1/$tfile.1 && + error "mv $tdir.0/$tfile.0 $tdir.1/$tfile.1 should fail" + true +} +run_test 407 "transaction fail should cause operation fail" + # # tests that do cleanup/setup should be run at the end #