X-Git-Url: https://git.whamcloud.com/?a=blobdiff_plain;f=lustre%2Ftests%2Fsanity.sh;h=146525ceb22b1771a3c95aaaf0411724c3b4e1a4;hb=89c31f6814a49a2697259ffe57e480e6653d31ee;hp=50b3bfa795a20c6b4175d6fa3b384e22b59490a2;hpb=764218793b300af6ae16706314392518aee55024;p=fs%2Flustre-release.git diff --git a/lustre/tests/sanity.sh b/lustre/tests/sanity.sh index 50b3bfa..146525c 100755 --- a/lustre/tests/sanity.sh +++ b/lustre/tests/sanity.sh @@ -13,8 +13,8 @@ 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 LU-8139 -ALWAYS_EXCEPT=" 76 101g $ALWAYS_EXCEPT" +# bug number for skipped tests: LU-2036 LU-8411 +ALWAYS_EXCEPT=" 76 407 $ALWAYS_EXCEPT" is_sles11() # LU-4341 { @@ -79,10 +79,10 @@ init_logging [ "$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" - # 4 13 (min)" - [ "$SLOW" = "no" ] && EXCEPT_SLOW="$EXCEPT_SLOW 51b 51ba" + # bug number for skipped test: LU-4536 LU-1957 + ALWAYS_EXCEPT="$ALWAYS_EXCEPT 65ic 180" + # 13 (min)" + [ "$SLOW" = "no" ] && EXCEPT_SLOW="$EXCEPT_SLOW 51b" fi FAIL_ON_ERROR=false @@ -264,6 +264,21 @@ test_6g() { test_mkdir $DIR/$tdir/d/subdir || error "mkdir $tdir/d/subdir failed" $CHECKSTAT -g \#$RUNAS_GID $DIR/$tdir/d/subdir || error "$tdir/d/subdir should be GID $RUNAS_GID" + if [[ $MDSCOUNT -gt 1 ]]; then + # check remote dir sgid inherite + $LFS mkdir -i 0 $DIR/$tdir.local || + error "mkdir $tdir.local failed" + chmod g+s $DIR/$tdir.local || + error "chmod $tdir.local failed" + chgrp $RUNAS_GID $DIR/$tdir.local || + error "chgrp $tdir.local failed" + $LFS mkdir -i 1 $DIR/$tdir.local/$tdir.remote || + error "mkdir $tdir.remote failed" + $CHECKSTAT -g \#$RUNAS_GID $DIR/$tdir.local/$tdir.remote || + error "$tdir.remote should be owned by $UID.$RUNAS_ID" + $CHECKSTAT -p 02755 $DIR/$tdir.local/$tdir.remote || + error "$tdir.remote should be mode 02755" + fi } run_test 6g "Is new dir in sgid dir inheriting group?" @@ -1650,7 +1665,7 @@ test_27u() { # bug 4900 createmany -o $DIR/$tdir/t- 1000 do_nodes $list $LCTL set_param fail_loc=0 - TLOG=$DIR/$tfile.getstripe + TLOG=$TMP/$tfile.getstripe $GETSTRIPE $DIR/$tdir > $TLOG OBJS=$(awk -vobj=0 '($1 == 0) { obj += 1 } END { print obj; }' $TLOG) unlinkmany $DIR/$tdir/t- 1000 @@ -1894,6 +1909,9 @@ check_seq_oid() local ff_pstripe if echo $ff_parent | grep -q 'stripe='; then ff_pstripe=$(echo $ff_parent | sed -e 's/.*stripe=//') + if echo $ff_pstripe | grep -q 'stripe_size='; then + ff_pstripe=$(echo $ff_pstripe | cut -d' ' -f1) + fi else # # $LL_DECODE_FILTER_FID does not print "stripe="; look @@ -1904,6 +1922,15 @@ check_seq_oid() sed -e 's/\]//') fi + if echo $ff_parent | grep -q 'stripe_count='; then + local ff_scnt=$(echo $ff_parent | + sed -e 's/.*stripe_count=//' | + cut -d' ' -f1) + + [ $lmm_count -eq $ff_scnt ] || + error "FF stripe count $lmm_count != $ff_scnt" + fi + # compare lmm_seq and filter_fid->ff_parent.f_seq [ $ff_pseq = $lmm_seq ] || error "FF parent SEQ $ff_pseq != $lmm_seq" @@ -2038,11 +2065,11 @@ test_27D() { local skip27D [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.7.64) ] && - skip27D = "-s 29" + skip27D = "-s 29,30,31" llapi_layout_test -d$DIR/$tdir -p$POOL -o$OSTCOUNT $skip27D || error "llapi_layout_test failed" - cleanup_pools || error "cleanup_pools failed" + destroy_test_pools || error "destroy test pools failed" } run_test 27D "validate llapi_layout API" @@ -2411,10 +2438,10 @@ run_test 31p "remove of open striped directory" cleanup_test32_mount() { local rc=0 trap 0 - local loopdev=$(losetup -a | grep $EXT2_DEV | sed -ne 's/:.*$/p') + 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 + rm -rf $DIR/$tdir return $rc } @@ -2598,59 +2625,49 @@ test_32o() { run_test 32o "stat d32o/symlink->tmp/symlink->lustre-root/$tfile" test_32p() { - log 32p_1 + log 32p_1 rm -fr $DIR/d32p - log 32p_2 + log 32p_2 rm -f $DIR/$tfile - log 32p_3 + log 32p_3 touch $DIR/$tfile - log 32p_4 + log 32p_4 test_mkdir -p $DIR/d32p/tmp - log 32p_5 + log 32p_5 TMP_DIR=$DIR/d32p/tmp - log 32p_6 + log 32p_6 ln -s $DIR/$tfile $TMP_DIR/symlink12 - log 32p_7 + log 32p_7 ln -s $TMP_DIR/symlink12 $TMP_DIR/../symlink02 - log 32p_8 + log 32p_8 cat $DIR/d32p/tmp/symlink12 || error - log 32p_9 + log 32p_9 cat $DIR/d32p/symlink02 || error - log 32p_10 + log 32p_10 } run_test 32p "open d32p/symlink->tmp/symlink->lustre-root/$tfile" -cleanup_testdir_mount() { - local rc=0 - trap 0 - 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() { [ $PARALLEL == "yes" ] && skip "skip parallel run" && return [ -e $DIR/$tdir ] && rm -fr $DIR/$tdir - trap cleanup_testdir_mount EXIT - test_mkdir -p $DIR/$tdir - touch $DIR/$tdir/under_the_mount - mount -t ext2 -o loop $EXT2_DEV $DIR/$tdir - ls $DIR/$tdir | grep "\" && error - cleanup_testdir_mount + trap cleanup_test32_mount EXIT + test_mkdir -p $DIR/$tdir/ext2-mountpoint + touch $DIR/$tdir/ext2-mountpoint/under_the_mount + mount -t ext2 -o loop $EXT2_DEV $DIR/$tdir/ext2-mountpoint + ls $DIR/$tdir/ext2-mountpoint | grep "\" && error + cleanup_test32_mount } run_test 32q "stat follows mountpoints in Lustre (should return error)" test_32r() { [ $PARALLEL == "yes" ] && skip "skip parallel run" && return [ -e $DIR/$tdir ] && rm -fr $DIR/$tdir - trap cleanup_testdir_mount EXIT - test_mkdir -p $DIR/$tdir - touch $DIR/$tdir/under_the_mount - mount -t ext2 -o loop $EXT2_DEV $DIR/$tdir - ls $DIR/$tdir | grep -q under_the_mount && error || true - cleanup_testdir_mount + trap cleanup_test32_mount EXIT + test_mkdir -p $DIR/$tdir/ext2-mountpoint + touch $DIR/$tdir/ext2-mountpoint/under_the_mount + mount -t ext2 -o loop $EXT2_DEV $DIR/$tdir/ext2-mountpoint + ls $DIR/$tdir/ext2-mountpoint | grep -q under_the_mount && error || true + cleanup_test32_mount } run_test 32r "opendir follows mountpoints in Lustre (should return error)" @@ -2850,6 +2867,15 @@ test_33f() { } run_test 33f "nonroot user can create, access, and remove a striped directory" +test_33g() { + mkdir -p $DIR/$tdir/dir2 + + local err=$($RUNAS mkdir $DIR/$tdir/dir2 2>&1) + echo $err + [[ $err =~ "exists" ]] || error "Not exists error" +} +run_test 33g "nonroot user create already existing root created file" + TEST_34_SIZE=${TEST_34_SIZE:-2000000000000} test_34a() { rm -f $DIR/f34 @@ -3103,7 +3129,7 @@ test_38() { } run_test 38 "open a regular file with O_DIRECTORY should return -ENOTDIR ===" -test_39() { +test_39a() { # was test_39 touch $DIR/$tfile touch $DIR/${tfile}2 # ls -l $DIR/$tfile $DIR/${tfile}2 @@ -3121,7 +3147,7 @@ test_39() { error "O_TRUNC didn't change timestamps" fi } -run_test 39 "mtime changed on create ===========================" +run_test 39a "mtime changed on create ===========================" test_39b() { test_mkdir -p -c1 $DIR/$tdir @@ -3630,7 +3656,7 @@ test_42a() { error "$BEFOREWRITES < $AFTERWRITES" start_writeback } -run_test 42a "ensure that we don't flush on close ==============" +run_test 42a "ensure that we don't flush on close" test_42b() { [ $PARALLEL == "yes" ] && skip "skip parallel run" && return @@ -3783,7 +3809,7 @@ test_42e() { # bug22074 } run_test 42e "verify sub-RPC writes are not done synchronously" -test_43() { +test_43A() { # was test_43 test_mkdir -p $DIR/$tdir cp -p /bin/ls $DIR/$tdir/$tfile $MULTIOP $DIR/$tdir/$tfile Ow_c & @@ -3791,56 +3817,53 @@ test_43() { # give multiop a chance to open sleep 1 - $DIR/$tdir/$tfile && error || true + $DIR/$tdir/$tfile && error "execute $DIR/$tdir/$tfile succeeded" || true kill -USR1 $pid } -run_test 43 "execution of file opened for write should return -ETXTBSY" +run_test 43A "execution of file opened for write should return -ETXTBSY" test_43a() { - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return - test_mkdir -p $DIR/$tdir - cp -p `which $MULTIOP` $DIR/$tdir/multiop || - cp -p multiop $DIR/$tdir/multiop + test_mkdir $DIR/$tdir + cp -p $(which $MULTIOP) $DIR/$tdir/multiop || + cp -p multiop $DIR/$tdir/multiop MULTIOP_PROG=$DIR/$tdir/multiop multiop_bg_pause $TMP/$tfile.junk O_c || - return 1 - MULTIOP_PID=$! - $MULTIOP $DIR/$tdir/multiop Oc && error "expected error, got success" - kill -USR1 $MULTIOP_PID || return 2 - wait $MULTIOP_PID || return 3 - rm $TMP/$tfile.junk $DIR/$tdir/multiop + error "multiop open $TMP/$tfile.junk failed" + MULTIOP_PID=$! + $MULTIOP $DIR/$tdir/multiop Oc && error "expected error, got success" + kill -USR1 $MULTIOP_PID || error "kill -USR1 PID $MULTIOP_PID failed" + wait $MULTIOP_PID || error "wait PID $MULTIOP_PID failed" } run_test 43a "open(RDWR) of file being executed should return -ETXTBSY" test_43b() { [ $PARALLEL == "yes" ] && skip "skip parallel run" && return - test_mkdir -p $DIR/$tdir - cp -p `which $MULTIOP` $DIR/$tdir/multiop || - cp -p multiop $DIR/$tdir/multiop + test_mkdir $DIR/$tdir + cp -p $(which $MULTIOP) $DIR/$tdir/multiop || + cp -p multiop $DIR/$tdir/multiop MULTIOP_PROG=$DIR/$tdir/multiop multiop_bg_pause $TMP/$tfile.junk O_c || - return 1 - MULTIOP_PID=$! - $TRUNCATE $DIR/$tdir/multiop 0 && error "expected error, got success" - kill -USR1 $MULTIOP_PID || return 2 - wait $MULTIOP_PID || return 3 - rm $TMP/$tfile.junk $DIR/$tdir/multiop + error "multiop open $TMP/$tfile.junk failed" + MULTIOP_PID=$! + $TRUNCATE $DIR/$tdir/multiop 0 && error "expected error, got success" + kill -USR1 $MULTIOP_PID || error "kill -USR1 PID $MULTIOP_PID failed" + wait $MULTIOP_PID || error "wait PID $MULTIOP_PID failed" } run_test 43b "truncate of file being executed should return -ETXTBSY" test_43c() { local testdir="$DIR/$tdir" - test_mkdir -p $DIR/$tdir + test_mkdir $testdir cp $SHELL $testdir/ - ( cd $(dirname $SHELL) && md5sum $(basename $SHELL) ) | \ - ( cd $testdir && md5sum -c) + ( cd $(dirname $SHELL) && md5sum $(basename $SHELL) ) | + ( cd $testdir && md5sum -c ) } -run_test 43c "md5sum of copy into lustre========================" +run_test 43c "md5sum of copy into lustre" -test_44() { +test_44A() { # was test_44 [[ $OSTCOUNT -lt 2 ]] && skip_env "skipping 2-stripe test" && return dd if=/dev/zero of=$DIR/f1 bs=4k count=1 seek=1023 dd if=$DIR/f1 bs=4k count=1 > /dev/null } -run_test 44 "zero length read from a sparse stripe =============" +run_test 44A "zero length read from a sparse stripe" test_44a() { local nstripe=$($LCTL lov_getconfig $DIR | grep default_stripe_count: | @@ -4097,6 +4120,12 @@ test_51a() { # was test_51 } run_test 51a "special situations: split htree with empty entry ==" +cleanup_print_lfs_df () { + trap 0 + $LFS df + $LFS df -i +} + test_51b() { [ $PARALLEL == "yes" ] && skip "skip parallel run" && return local dir=$DIR/$tdir @@ -4118,10 +4147,12 @@ test_51b() { # need to check free space for the directories as well local blkfree=$(lctl get_param -n mdc.$FSNAME-MDT$mdtidx*.kbytesavail) - numfree=$((blkfree / 4)) + numfree=$(( blkfree / $(fs_inode_ksize) )) [[ $numfree -lt $nrdirs ]] && skip "not enough blocks ($numfree)" && return + trap cleanup_print_lfsdf EXIT + # create files createmany -d $dir/d $nrdirs || error "failed to create $nrdirs subdirs in MDT$mdtidx:$dir" @@ -4133,6 +4164,9 @@ test_51b() { local left=100 local delete=$((nrdirs - left)) + $LFS df + $LFS df -i + # 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 $dir), created before: $nrdirs" @@ -4151,6 +4185,8 @@ test_51b() { local after=$(stat -c %h $dir) [[ $after -gt 2 ]] && error "nlink after: $after > 2" || echo "nlink after: $after" + + cleanup_print_lfs_df } run_test 51b "exceed 64k subdirectory nlink limit on create, verify unlink" @@ -4209,6 +4245,35 @@ test_51e() { } run_test 51e "check file nlink limit" +test_51f() { + test_mkdir $DIR/$tdir + + local max=100000 + local ulimit_old=$(ulimit -n) + local spare=20 # number of spare fd's for scripts/libraries, etc. + local mdt=$(lfs getstripe -M $DIR/$tdir) + local numfree=$(lfs df -i $DIR/$tdir | awk '/MDT:'$mdt'/ { print $4 }') + + echo "MDT$mdt numfree=$numfree, max=$max" + [[ $numfree -gt $max ]] && numfree=$max || numfree=$((numfree * 7 / 8)) + if [ $((numfree + spare)) -gt $ulimit_old ]; then + while ! ulimit -n $((numfree + spare)); do + numfree=$((numfree * 3 / 4)) + done + echo "changed ulimit from $ulimit_old to $((numfree + spare))" + else + echo "left ulimit at $ulimit_old" + fi + + createmany -o -k -t 120 $DIR/$tdir/f $numfree || + error "create+open $numfree files in $DIR/$tdir failed" + ulimit -n $ulimit_old + + # if createmany exits at 120s there will be fewer than $numfree files + unlinkmany $DIR/$tdir/f $numfree || true +} +run_test 51f "check many open files limit" + test_52a() { [ -f $DIR/$tdir/foo ] && chattr -a $DIR/$tdir/foo test_mkdir -p $DIR/$tdir @@ -4423,30 +4488,47 @@ test_56a() { # was test_56 error "$GETSTRIPE $DIR/$tdir: found $FILENUM, not $NUMFILES" echo "$GETSTRIPE --recursive passed." - # test lfs getstripe with file instead of dir + # test lfs getstripe with file instead of dir FILENUM=$($GETSTRIPE $DIR/$tdir/file1 | grep -c obdidx) [[ $FILENUM -eq 1 ]] || error "$GETSTRIPE $DIR/$tdir/file1: found $FILENUM, not 1" echo "$GETSTRIPE file1 passed." - #test lfs getstripe with --verbose + #test lfs getstripe with --verbose [[ $($GETSTRIPE --verbose $DIR/$tdir | grep -c lmm_magic) -eq $NUMFILES ]] || error "$GETSTRIPE --verbose $DIR/$tdir: want $NUMFILES" [[ $($GETSTRIPE $DIR/$tdir | grep -c lmm_magic) -eq 0 ]] || - rror "$GETSTRIPE $DIR/$tdir: showed lmm_magic" + error "$GETSTRIPE $DIR/$tdir: showed lmm_magic" + + #test lfs getstripe with -v prints lmm_fid + [[ $($GETSTRIPE -v $DIR/$tdir | grep -c lmm_fid) -eq $NUMFILES ]] || + error "$GETSTRIPE -v $DIR/$tdir: want $NUMFILES lmm_fid: lines" + [[ $($GETSTRIPE $DIR/$tdir | grep -c lmm_fid) -eq 0 ]] || + error "$GETSTRIPE $DIR/$tdir: showed lmm_fid" echo "$GETSTRIPE --verbose passed." - #test lfs getstripe with --obd - $GETSTRIPE --obd wrong_uuid $DIR/$tdir 2>&1 | + #check for FID information + local fid1=$($GETSTRIPE --fid $DIR/$tdir/file1) + local fid2=$($GETSTRIPE --verbose $DIR/$tdir/file1 | + awk '/lmm_fid: / { print $2 }') + local fid3=$($LFS path2fid $DIR/$tdir/file1) + [ "$fid1" != "$fid2" ] && + error "getstripe --fid $fid1 != getstripe --verbose $fid2" + [ "$fid1" != "$fid3" ] && + error "getstripe --fid $fid1 != lfs path2fid $fid3" + echo "$GETSTRIPE --fid passed." + + #test lfs getstripe with --obd + $GETSTRIPE --obd wrong_uuid $DIR/$tdir 2>&1 | grep -q "unknown obduuid" || error "$GETSTRIPE --obd wrong_uuid should return error message" [[ $OSTCOUNT -lt 2 ]] && - skip_env "skipping other $GETSTRIPE --obd test" && return + skip_env "skipping other $GETSTRIPE --obd test" && return - OSTIDX=1 - OBDUUID=$(ostuuid_from_index $OSTIDX) + OSTIDX=1 + OBDUUID=$(ostuuid_from_index $OSTIDX) FILENUM=$($GETSTRIPE -ir $DIR/$tdir | grep "^$OSTIDX\$" | wc -l) FOUND=$($GETSTRIPE -r --obd $OBDUUID $DIR/$tdir | grep obdidx | wc -l) [[ $FOUND -eq $FILENUM ]] || @@ -4459,6 +4541,50 @@ 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 $LFS getdirstripe" + +test_56c() { + local ost_idx=0 + local ost_name=$(ostname_from_index $ost_idx) + + local old_status=$(ost_dev_status $ost_idx) + [[ -z "$old_status" ]] || + { skip_env "OST $ost_name is in $old_status status"; return 0; } + + do_facet ost1 $LCTL set_param -n obdfilter.$ost_name.degraded=1 + sleep_maxage + + local new_status=$(ost_dev_status $ost_idx) + [[ "$new_status" = "D" ]] || + error "OST $ost_name is in status of '$new_status', not 'D'" + + do_facet ost1 $LCTL set_param -n obdfilter.$ost_name.degraded=0 + sleep_maxage + + new_status=$(ost_dev_status $ost_idx) + [[ -z "$new_status" ]] || + error "OST $ost_name is in status of '$new_status', not ''" +} +run_test 56c "check 'lfs df' showing device status" + NUMFILES=3 NUMDIRS=3 setup_56() { @@ -4772,7 +4898,7 @@ run_test 56s "check lfs find -stripe-count works" test_56t() { # LU-611 TDIR=$DIR/${tdir}t - setup_56 $NUMFILES $NUMDIRS "-s 512k" + setup_56 $NUMFILES $NUMDIRS "--stripe-size 512k" $SETSTRIPE -S 256k $TDIR/$tfile.{0,1,2,3} @@ -5288,7 +5414,23 @@ test_60a() { start mgs $(mgsdevname) $MGS_MOUNT_OPTS || error "start mgs failed" $pass || error "test failed, see FAILED test_60a messages for specifics" } -run_test 60a "llog_test run from kernel module and test llog_reader ==========" +run_test 60a "llog_test run from kernel module and test llog_reader" + +test_60aa() { + # test old logid format + if [ $(lustre_version_code mgs) -le $(version_code 3.1.53) ]; then + do_facet mgs $LCTL dl | grep MGS + do_facet mgs "$LCTL --device %MGS llog_print \\\\\\\$$FSNAME-client" || + error "old llog_print failed" + fi + + # test new logid format + if [ $(lustre_version_code mgs) -ge $(version_code 2.9.53) ]; then + do_facet mgs "$LCTL --device MGS llog_print $FSNAME-client" || + error "new llog_print failed" + fi +} +run_test 60aa "llog_print works with FIDs and simple names" test_60b() { # bug 6411 [ $PARALLEL == "yes" ] && skip "skip parallel run" && return @@ -5640,6 +5782,12 @@ 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 @@ -5666,10 +5814,6 @@ cleanup_68() { rm -f $LLOOP unset LLOOP fi - if [ ! -z "$LLITELOOPLOAD" ]; then - rmmod llite_lloop - unset LLITELOOPLOAD - fi rm -f $DIR/f68* } @@ -5681,74 +5825,6 @@ swap_used() { swapon -s | awk '($1 == "'$1'") { print $4 }' } -# test case for lloop driver, basic function -test_68a() { - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return - [ "$UID" != 0 ] && skip_env "must run as root" && return - llite_lloop_enabled || \ - { skip_env "llite_lloop module disabled" && return; } - - trap cleanup_68 EXIT - - if ! module_loaded llite_lloop; then - if load_module llite/llite_lloop; then - LLITELOOPLOAD=yes - else - skip_env "can't find module llite_lloop" - return - fi - fi - - LLOOP=$TMP/lloop.`date +%s`.`date +%N` - dd if=/dev/zero of=$DIR/f68a bs=4k count=1024 - $LCTL blockdev_attach $DIR/f68a $LLOOP || error "attach failed" - - directio rdwr $LLOOP 0 1024 4096 || error "direct write failed" - directio rdwr $LLOOP 0 1025 4096 && error "direct write should fail" - - cleanup_68 -} -run_test 68a "lloop driver - basic test ========================" - -# excercise swapping to lustre by adding a high priority swapfile entry -# and then consuming memory until it is used. -test_68b() { # was test_68 - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return - [ "$UID" != 0 ] && skip_env "must run as root" && return - lctl get_param -n devices | grep -q obdfilter && \ - skip "local OST" && return - - grep -q llite_lloop /proc/modules - [ $? -ne 0 ] && skip "can't find module llite_lloop" && return - - [ -z "`$LCTL list_nids | grep -v tcp`" ] && \ - skip "can't reliably test swap with TCP" && return - - MEMTOTAL=`meminfo MemTotal` - NR_BLOCKS=$((MEMTOTAL>>8)) - [[ $NR_BLOCKS -le 2048 ]] && NR_BLOCKS=2048 - - LLOOP=$TMP/lloop.`date +%s`.`date +%N` - dd if=/dev/zero of=$DIR/f68b bs=64k seek=$NR_BLOCKS count=1 - mkswap $DIR/f68b - - $LCTL blockdev_attach $DIR/f68b $LLOOP || error "attach failed" - - trap cleanup_68 EXIT - - swapon -p 32767 $LLOOP || error "swapon $LLOOP failed" - - echo "before: `swapon -s | grep $LLOOP`" - $MEMHOG $MEMTOTAL || error "error allocating $MEMTOTAL kB" - echo "after: `swapon -s | grep $LLOOP`" - SWAPUSED=`swap_used $LLOOP` - - cleanup_68 - - [ $SWAPUSED -eq 0 ] && echo "no swap used???" || true -} -run_test 68b "support swapping to Lustre ========================" - # bug5265, obdfilter oa2dentry return -ENOENT # #define OBD_FAIL_SRV_ENOENT 0x217 test_69() { @@ -6092,22 +6168,6 @@ test_77g() { # bug 10889 } run_test 77g "checksum error on OST write, read" -test_77i() { # bug 13805 - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return - $GSS && skip "could not run with gss" && return - #define OBD_FAIL_OSC_CONNECT_CKSUM 0x40b - lctl set_param fail_loc=0x40b - remount_client $MOUNT - lctl set_param fail_loc=0 - for VALUE in `lctl get_param osc.*osc-[^mM]*.checksum_type`; do - PARAM=`echo ${VALUE[0]} | cut -d "=" -f1` - algo=`lctl get_param -n $PARAM | sed 's/.*\[\(.*\)\].*/\1/g'` - [ "$algo" = "adler" ] || error "algo set to $algo instead of adler" - done - remount_client $MOUNT -} -run_test 77i "client not supporting OSD_CONNECT_CKSUM" - test_77j() { # bug 13805 [ $PARALLEL == "yes" ] && skip "skip parallel run" && return $GSS && skip "could not run with gss" && return @@ -6115,10 +6175,15 @@ test_77j() { # bug 13805 lctl set_param fail_loc=0x40c remount_client $MOUNT lctl set_param fail_loc=0 - sleep 2 # wait async osc connect to finish - for VALUE in `lctl get_param osc.*osc-[^mM]*.checksum_type`; do - PARAM=`echo ${VALUE[0]} | cut -d "=" -f1` - algo=`lctl get_param -n $PARAM | sed 's/.*\[\(.*\)\].*/\1/g'` + # wait async osc connect to finish and reflect updated state value + local i + for (( i=0; i < OSTCOUNT; i++ )) ; do + wait_osc_import_state client ost$((i+1)) FULL + done + + for VALUE in $(lctl get_param osc.*osc-[^mM]*.checksum_type); do + PARAM=$(echo ${VALUE[0]} | cut -d "=" -f1) + algo=$(lctl get_param -n $PARAM | sed 's/.*\[\(.*\)\].*/\1/g') [ "$algo" = "adler" ] || error "algo set to $algo instead of adler" done remount_client $MOUNT @@ -6129,6 +6194,11 @@ 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; } @@ -6158,6 +6228,8 @@ 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)) @@ -6170,7 +6242,7 @@ 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 ============" @@ -6652,6 +6724,10 @@ run_test 101e "check read-ahead for small read(1k) for small files(500k)" test_101f() { which iozone || { skip "no iozone installed" && return; } + local old_debug=$($LCTL get_param debug) + old_debug=${old_debug#*=} + $LCTL set_param debug="reada mmap" + # create a test file iozone -i 0 -+n -r 1m -s 128m -w -f $DIR/$tfile > /dev/null 2>&1 @@ -6662,71 +6738,87 @@ test_101f() { $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 + iozone -i 1 -u 1 -l 1 -+n -r 32k -s 128m -B -f $DIR/$tfile \ + > /dev/null 2>&1 echo checking missing pages + $LCTL get_param llite.*.read_ahead_stats local miss=$($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!" + $LCTL set_param debug="$old_debug" + [ $miss -lt 3 ] || error "misses too much pages ('$miss')!" rm -f $DIR/$tfile } run_test 101f "check mmap read performance" +test_101g_brw_size_test() { + local mb=$1 + local pages=$((mb * 1048576 / $(page_size))) + + $LCTL set_param osc.*.max_pages_per_rpc=${mb}M || + { error "unable to set max_pages_per_rpc=${mb}M"; return 1; } + for mp in $($LCTL get_param -n osc.*.max_pages_per_rpc); do + [ $mp -ne $pages ] && error "max_pages_per_rpc $mp != $pages" && + return 2 + done + + $LCTL set_param -n osc.*.rpc_stats=0 + + # 10 RPCs should be enough for the test + local count=10 + dd if=/dev/zero of=$DIR/$tfile bs=${mb}M count=$count || + { error "dd write ${mb} MB blocks failed"; return 3; } + cancel_lru_locks osc + dd of=/dev/null if=$DIR/$tfile bs=${mb}M count=$count || + { error "dd write ${mb} MB blocks failed"; return 4; } + + # calculate number of full-sized read and write RPCs + rpcs=($($LCTL get_param -n 'osc.*.rpc_stats' | + sed -n '/pages per rpc/,/^$/p' | + awk '/'$pages':/ { reads += $2; writes += $5 };' \ + 'END { print reads,writes }')) + [ ${rpcs[0]} -ne $count ] && error "${rpcs[0]} != $count read RPCs" && + return 5 + [ ${rpcs[1]} -ne $count ] && error "${rpcs[1]} != $count write RPCs" && + return 6 + + return 0 +} + test_101g() { local rpcs local osts=$(get_facets OST) local list=$(comma_list $(osts_nodes)) local p="$TMP/$TESTSUITE-$TESTNAME.parameters" + local brw_size="obdfilter.*.brw_size" - save_lustre_params $osts "obdfilter.*.brw_size" > $p - - $LFS setstripe -c 1 $DIR/$tfile + $LFS setstripe -i 0 -c 1 $DIR/$tfile + local orig_mb=$(do_facet ost1 $LCTL get_param -n $brw_size | head -n 1) 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" - - 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 + [ $(lustre_version_code ost1) -ge $(version_code 2.9.52) ] && + suffix="M" + if [[ $orig_mb < 16 ]]; then + save_lustre_params $osts "$brw_size" > $p + do_nodes $list $LCTL set_param -n $brw_size=16$suffix || + error "set 16MB RPC size failed" + + echo "remount client to enable new RPC size" + remount_client $MOUNT || error "remount_client failed" + fi - # calculate 16 MiB RPCs - rpcs=$($LCTL get_param 'osc.*.rpc_stats' | - sed -n '/pages per rpc/,/^$/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" + test_101g_brw_size_test 16 || error "16MB RPC test failed" + # should be able to set brw_size=12, but no rpc_stats for that + test_101g_brw_size_test 8 || error "8MB RPC test failed" fi - echo "set RPC size to 4MB" + test_101g_brw_size_test 4 || error "4MB RPC test failed" - $LCTL set_param -n osc.*.max_pages_per_rpc=4M osc.*.rpc_stats=0 - dd if=/dev/zero of=$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" - - restore_lustre_params < $p - remount_client $MOUNT || error "remount_client failed" + if [[ $orig_mb < 16 ]]; then + restore_lustre_params < $p + remount_client $MOUNT || error "remount_client restore failed" + fi rm -f $p $DIR/$tfile } @@ -7101,29 +7193,31 @@ getxattr() { # getxattr path name } test_102n() { # LU-4101 mdt: protect internal xattrs - local file0=$DIR/$tfile.0 - local file1=$DIR/$tfile.1 - local xattr0=$TMP/$tfile.0 - local xattr1=$TMP/$tfile.1 - local name - local value - [ -z "$(which setfattr 2>/dev/null)" ] && skip "could not find setfattr" && return - if [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.5.50) ] then skip "MDT < 2.5.50 allows setxattr on internal trusted xattrs" return fi + local file0=$DIR/$tfile.0 + local file1=$DIR/$tfile.1 + local xattr0=$TMP/$tfile.0 + local xattr1=$TMP/$tfile.1 + local namelist="lov lma lmv link fid version som hsm" + local name + local value + rm -rf $file0 $file1 $xattr0 $xattr1 touch $file0 $file1 # Get 'before' xattrs of $file1. getfattr --absolute-names --dump --match=- $file1 > $xattr0 - for name in lov lma lmv link fid version som hsm; do + [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.8.53) ] && + namelist+=" lfsck_namespace" + for name in $namelist; do # Try to copy xattr from $file0 to $file1. value=$(getxattr $file0 trusted.$name 2> /dev/null) @@ -7260,7 +7354,8 @@ test_103a() { SAVE_UMASK=$(umask) umask 0022 - cd $DIR + mkdir -p $DIR/$tdir + cd $DIR/$tdir echo "performing cp ..." run_acl_subtest cp || error "run_acl_subtest cp failed" @@ -7270,6 +7365,15 @@ test_103a() { run_acl_subtest misc || error "misc test failed" echo "performing permissions..." run_acl_subtest permissions || error "permissions failed" + # LU-1482 mdd: Setting xattr are properly checked with and without ACLs + if [ $(lustre_version_code $SINGLEMDS) -gt $(version_code 2.8.55) -o \ + \( $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.6) -a \ + $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.5.29) \) ] + then + echo "performing permissions xattr..." + run_acl_subtest permissions_xattr || + error "permissions_xattr failed" + fi echo "performing setfacl..." run_acl_subtest setfacl || error "setfacl test failed" @@ -7530,20 +7634,24 @@ run_test 115 "verify dynamic thread creation====================" free_min_max () { wait_delete_completed AVAIL=($(lctl get_param -n osc.*[oO][sS][cC]-[^M]*.kbytesavail)) - echo OST kbytes available: ${AVAIL[@]} - MAXI=0; MAXV=${AVAIL[0]} - MINI=0; MINV=${AVAIL[0]} + echo "OST kbytes available: ${AVAIL[@]}" + MAXV=${AVAIL[0]} + MAXI=0 + MINV=${AVAIL[0]} + MINI=0 for ((i = 0; i < ${#AVAIL[@]}; i++)); do #echo OST $i: ${AVAIL[i]}kb if [[ ${AVAIL[i]} -gt $MAXV ]]; then - MAXV=${AVAIL[i]}; MAXI=$i + MAXV=${AVAIL[i]} + MAXI=$i fi if [[ ${AVAIL[i]} -lt $MINV ]]; then - MINV=${AVAIL[i]}; MINI=$i + MINV=${AVAIL[i]} + MINI=$i fi done - echo Min free space: OST $MINI: $MINV - echo Max free space: OST $MAXI: $MAXV + echo "Min free space: OST $MINI: $MINV" + echo "Max free space: OST $MAXI: $MAXV" } test_116a() { # was previously test_116() @@ -7566,9 +7674,9 @@ test_116a() { # was previously test_116() # Check if we need to generate uneven OSTs test_mkdir -p $DIR/$tdir/OST${MINI} - local FILL=$(($MINV / 4)) - local DIFF=$(($MAXV - $MINV)) - local DIFF2=$(($DIFF * 100 / $MINV)) + local FILL=$((MINV / 4)) + local DIFF=$((MAXV - MINV)) + local DIFF2=$((DIFF * 100 / MINV)) local threshold=$(do_facet $SINGLEMDS \ lctl get_param -n *.*MDT0000-mdtlov.qos_threshold_rr | head -n1) @@ -7582,15 +7690,15 @@ test_116a() { # was previously test_116() else # generate uneven OSTs. Write 2% over the QOS threshold value echo "no" - DIFF=$(($threshold - $DIFF2 + 2)) - DIFF2=$(( ($MINV * $DIFF)/100 )) - echo "Fill ${DIFF}% remaining space in OST${MINI} with ${DIFF2}KB" + DIFF=$((threshold - DIFF2 + 2)) + DIFF2=$((MINV * DIFF / 100)) + echo "Fill $DIFF% remaining space in OST$MINI with ${DIFF2}KB" $SETSTRIPE -i $MINI -c 1 $DIR/$tdir/OST${MINI} || error "setstripe failed" - DIFF=$(($DIFF2 / 2048)) + DIFF=$((DIFF2 / 2048)) i=0 while [ $i -lt $DIFF ]; do - i=$(($i + 1)) + i=$((i + 1)) dd if=/dev/zero of=$DIR/$tdir/OST${MINI}/$tfile-$i \ bs=2M count=1 2>/dev/null echo -n . @@ -7601,10 +7709,10 @@ test_116a() { # was previously test_116() free_min_max fi - DIFF=$(($MAXV - $MINV)) - DIFF2=$(($DIFF * 100 / $MINV)) - echo -n "diff=${DIFF}=${DIFF2}% must be > ${threshold}% for QOS mode..." - if [[ $DIFF2 -gt $threshold ]]; then + DIFF=$((MAXV - MINV)) + DIFF2=$((DIFF * 100 / MINV)) + echo -n "diff=$DIFF=$DIFF2% must be > $threshold% for QOS mode..." + if [ $DIFF2 -gt $threshold ]; then echo "ok" else echo "failed - QOS mode won't be used" @@ -7613,12 +7721,14 @@ test_116a() { # was previously test_116() return fi - MINI1=$MINI; MINV1=$MINV - MAXI1=$MAXI; MAXV1=$MAXV + MINI1=$MINI + MINV1=$MINV + MAXI1=$MAXI + MAXV1=$MAXV # now fill using QOS $SETSTRIPE -c 1 $DIR/$tdir - FILL=$(($FILL / 200)) + FILL=$((FILL / 200)) if [ $FILL -gt 600 ]; then FILL=600 fi @@ -7636,31 +7746,31 @@ test_116a() { # was previously test_116() echo "Note: free space may not be updated, so measurements might be off" free_min_max - DIFF2=$(($MAXV - $MINV)) + DIFF2=$((MAXV - MINV)) echo "free space delta: orig $DIFF final $DIFF2" [ $DIFF2 -gt $DIFF ] && echo "delta got worse!" - DIFF=$(($MINV1 - ${AVAIL[$MINI1]})) + DIFF=$((MINV1 - ${AVAIL[$MINI1]})) echo "Wrote ${DIFF}KB to smaller OST $MINI1" - DIFF2=$(($MAXV1 - ${AVAIL[$MAXI1]})) + DIFF2=$((MAXV1 - ${AVAIL[$MAXI1]})) echo "Wrote ${DIFF2}KB to larger OST $MAXI1" if [[ $DIFF -gt 0 ]]; then - FILL=$(($DIFF2 * 100 / $DIFF - 100)) + FILL=$((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.${FSNAME}-clilov-*.target_obd | - awk '/'$MINI1': / {print $2; exit}') + awk '/'$MINI1': / {print $2; exit}') echo $UUID MINC=$($GETSTRIPE --ost $UUID $DIR/$tdir | grep $DIR | wc -l) echo "$MINC files created on smaller OST $MINI1" UUID=$(lctl get_param -n lov.${FSNAME}-clilov-*.target_obd | - awk '/'$MAXI1': / {print $2; exit}') + awk '/'$MAXI1': / {print $2; exit}') echo $UUID MAXC=$($GETSTRIPE --ost $UUID $DIR/$tdir | grep $DIR | wc -l) echo "$MAXC files created on larger OST $MAXI1" if [[ $MINC -gt 0 ]]; then - FILL=$(($MAXC * 100 / $MINC - 100)) + FILL=$((MAXC * 100 / MINC - 100)) echo "Wrote ${FILL}% more files to larger OST $MAXI1" fi [[ $MAXC -gt $MINC ]] || @@ -8305,6 +8415,7 @@ test_120e() { 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. + sleep 2 cancel_lru_locks osc can1=$(do_facet $SINGLEMDS \ "$LCTL get_param -n ldlm.services.ldlm_canceld.stats" | @@ -8347,6 +8458,7 @@ test_120f() { stat $DIR/$tdir/d1 $DIR/$tdir/d2 $DIR/$tdir/d1/f1 $DIR/$tdir/d2/f2 > /dev/null # XXX client can not do early lock cancel of OST lock # during rename (LU-4206), so cancel osc lock now. + sleep 2 cancel_lru_locks osc can1=$(do_facet $SINGLEMDS \ "$LCTL get_param -n ldlm.services.ldlm_canceld.stats" | @@ -8641,11 +8753,18 @@ test_124b() { NR=$(($(default_lru_size)*20)) if [[ $NR -gt $LIMIT ]]; then - log "Limit lock number by $LIMIT locks" - NR=$LIMIT - fi - lru_resize_disable mdc - test_mkdir -p $DIR/$tdir/disable_lru_resize || + log "Limit lock number by $LIMIT locks" + NR=$LIMIT + fi + + IFree=$(mdsrate_inodes_available) + if [ $IFree -lt $NR ]; then + log "Limit lock number by $IFree inodes" + NR=$IFree + fi + + lru_resize_disable mdc + test_mkdir -p $DIR/$tdir/disable_lru_resize || error "failed to create $DIR/$tdir/disable_lru_resize" createmany -o $DIR/$tdir/disable_lru_resize/f $NR @@ -8882,6 +9001,9 @@ check_mds_dmesg() { } test_129() { + [[ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.5.56) ]] || + { skip "Need MDS version with at least 2.5.56"; return 0; } + [ $PARALLEL == "yes" ] && skip "skip parallel run" && return if [ "$(facet_fstype $SINGLEMDS)" != ldiskfs ]; then skip "Only applicable to ldiskfs-based MDTs" @@ -9303,22 +9425,27 @@ test_131e() { run_test 131e "test read hitting hole" check_stats() { + local facet=$1 + local op=$2 + local want=${3:-0} local res - local count - case $1 in - $SINGLEMDS) res=`do_facet $SINGLEMDS $LCTL get_param mdt.$FSNAME-MDT0000.md_stats | grep "$2"` + + case $facet in + mds*) res=$(do_facet $facet \ + $LCTL get_param mdt.$FSNAME-MDT0000.md_stats | grep "$op") ;; - ost) res=`do_facet ost1 $LCTL get_param obdfilter.$FSNAME-OST0000.stats | grep "$2"` + ost*) res=$(do_facet $facet \ + $LCTL get_param obdfilter.$FSNAME-OST0000.stats | grep "$op") ;; - *) error "Wrong argument $1" ;; + *) error "Wrong facet '$facet'" ;; esac echo $res - [ -z "$res" ] && error "The counter for $2 on $1 was not incremented" + [ "$res" ] || error "The counter for $op on $facet was not incremented" # if the argument $3 is zero, it means any stat increment is ok. - if [[ $3 -gt 0 ]]; then - count=$(echo $res | awk '{ print $2 }') - [[ $count -ne $3 ]] && - error "The $2 counter on $1 is wrong - expected $3" + if [[ $want -gt 0 ]]; then + local count=$(echo $res | awk '{ print $2 }') + [[ $count -ne $want ]] && + error "The $op counter on $facet is $count, not $want" fi } @@ -9418,17 +9545,17 @@ test_133c() { dd if=/dev/zero of=${testdir}/${tfile} conv=notrunc bs=512k count=1 || error "dd failed" sync cancel_lru_locks osc - check_stats ost "write" 1 + check_stats ost1 "write" 1 dd if=${testdir}/${tfile} of=/dev/null bs=1k count=1 || error "dd failed" - check_stats ost "read" 1 + check_stats ost1 "read" 1 > ${testdir}/${tfile} || error "truncate failed" - check_stats ost "punch" 1 + check_stats ost1 "punch" 1 rm -f ${testdir}/${tfile} || error "file remove failed" wait_delete_completed - check_stats ost "destroy" 1 + check_stats ost1 "destroy" 1 rm -rf $DIR/${tdir} } @@ -9605,17 +9732,15 @@ test_133e() { } run_test 133e "Verifying OST {read,write}_bytes nid stats =================" -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+=" $dir" -done +proc_regexp="/{proc,sys}/{fs,sys,kernel/debug}/{lustre,lnet}/" test_133f() { remote_mds_nodsh && skip "remote MDS with nodsh" && return remote_ost_nodsh && skip "remote OST with nodsh" && return # First without trusting modes. + local proc_dirs=$(eval \ls -d $proc_regexp 2>/dev/null) + echo "proc_dirs='$proc_dirs'" + [ -n "$proc_dirs" ] || error "no proc_dirs on $HOSTNAME" find $proc_dirs -exec cat '{}' \; &> /dev/null # Second verifying readability. @@ -9624,12 +9749,16 @@ test_133f() { # 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 $proc_dirs \ + for facet in mds1 ost1; do + local facet_proc_dirs=$(do_facet $facet \ + \\\ls -d $proc_regexp 2>/dev/null) + echo "${facet}_proc_dirs='$facet_proc_dirs'" + [ -z "$facet_proc_dirs" ] && error "no proc_dirs on $facet" + do_facet $facet find $facet_proc_dirs \ ! -name req_history \ -exec cat '{}' \\\; &> /dev/null - do_facet $facet find $proc_dirs \ + do_facet $facet find $facet_proc_dirs \ ! -name req_history \ -type f \ -exec cat '{}' \\\; &> /dev/null || @@ -9642,6 +9771,9 @@ test_133g() { remote_mds_nodsh && skip "remote MDS with nodsh" && return remote_ost_nodsh && skip "remote OST with nodsh" && return # Second verifying writability. + local proc_dirs=$(eval \ls -d $proc_regexp 2>/dev/null) + echo "proc_dirs='$proc_dirs'" + [ -n "$proc_dirs" ] || error "no proc_dirs on $HOSTNAME" find $proc_dirs \ -type f \ -not -name force_lbug \ @@ -9649,21 +9781,20 @@ test_133g() { -exec badarea_io '{}' \; &> /dev/null || error "find $proc_dirs failed" - [ $(lustre_version_code $SINGLEMDS) -le $(version_code 2.5.54) ] && - skip "Too old lustre on MDS" && return - - [ $(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 \ + for facet in mds1 ost1; do + [ $(lustre_version_code $facet) -le $(version_code 2.5.54) ] && + skip "Too old lustre on $facet" && continue + local facet_proc_dirs=$(do_facet $facet \ + \\\ls -d $proc_regexp 2> /dev/null) + echo "${facet}_proc_dirs='$facet_proc_dirs'" + [ -z "$facet_proc_dirs" ] && error "no proc_dirs on $facet" + do_facet $facet find $facet_proc_dirs \ -type f \ -not -name force_lbug \ -not -name changelog_mask \ -exec badarea_io '{}' \\\; &> /dev/null || - error "$facet find $proc_dirs failed" - + error "$facet find $facet_proc_dirs failed" done # remount the FS in case writes/reads /proc break the FS @@ -9673,6 +9804,34 @@ test_133g() { } run_test 133g "Check for Oopses on bad io area writes/reads in /proc" +test_133h() { + remote_mds_nodsh && skip "remote MDS with nodsh" && return + remote_ost_nodsh && skip "remote OST with nodsh" && return + [[ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.9.54) ]] && + skip "Need MDS version at least 2.9.54" && return + + local facet + for facet in client mds1 ost1; do + local facet_proc_dirs=$(do_facet $facet \ + \\\ls -d $proc_regexp 2> /dev/null) + [ -z "$facet_proc_dirs" ] && error "no proc_dirs on $facet" + echo "${facet}_proc_dirs='$facet_proc_dirs'" + # Get the list of files that are missing the terminating newline + local missing=($(do_facet $facet \ + find ${facet_proc_dirs} -type f \| \ + while read F\; do \ + awk -v FS='\v' -v RS='\v\v' \ + "'END { if(NR>0 && \ + \\\$NF !~ /.*\\\n\$/) \ + print FILENAME}'" \ + '\$F'\; \ + done 2>/dev/null)) + [ ${#missing[*]} -eq 0 ] || + error "files do not end with newline: ${missing[*]}" + done +} +run_test 133h "Proc files should end with newlines" + test_134a() { remote_mds_nodsh && skip "remote MDS with nodsh" && return [[ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.7.54) ]] && @@ -10102,20 +10261,43 @@ test_154A() { [[ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.4.1) ]] && skip "Need MDS version at least 2.4.1" && return - touch $DIR/$tfile - local FID=$($LFS path2fid $DIR/$tfile) - [ -z "$FID" ] && error "path2fid unable to get $DIR/$tfile FID" + local tf=$DIR/$tfile + touch $tf - # check that we get the same pathname back - local FOUND=$($LFS fid2path $MOUNT "$FID") - [ -z "$FOUND" ] && error "fid2path unable to get $FID path" - [ "$FOUND" != "$DIR/$tfile" ] && - error "fid2path(path2fid($DIR/$tfile)) = $FOUND != $DIR/$tfile" + local fid=$($LFS path2fid $tf) + [ -z "$fid" ] && error "path2fid unable to get $tf FID" - rm -rf $DIR/$tfile + # check that we get the same pathname back + local found=$($LFS fid2path $MOUNT "$fid") + [ -z "$found" ] && error "fid2path unable to get '$fid' path" + [ "$found" == "$tf" ] || + error "fid2path($fid=path2fid($tf)) = $found != $tf" } run_test 154A "lfs path2fid and fid2path basic checks" +test_154B() { + [[ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.4.1) ]] && + skip "Need MDS version at least 2.4.1" && return + + mkdir -p $DIR/$tdir || error "mkdir $tdir failed" + touch $DIR/$tdir/$tfile || error "touch $DIR/$tdir/$tfile failed" + local linkea=$($LL_DECODE_LINKEA $DIR/$tdir/$tfile | grep 'pfid') + [ -z "$linkea" ] && error "decode linkea $DIR/$tdir/$tfile failed" + + local name=$(echo $linkea | awk '/pfid/ {print $5}' | sed -e "s/'//g") + local PFID=$(echo $linkea | awk '/pfid/ {print $3}' | sed -e "s/,//g") + + # check that we get the same pathname + echo "PFID: $PFID, name: $name" + local FOUND=$($LFS fid2path $MOUNT "$PFID") + [ -z "$FOUND" ] && error "fid2path unable to get $PFID path" + [ "$FOUND/$name" != "$DIR/$tdir/$tfile" ] && + error "ll_decode_linkea has $FOUND/$name != $DIR/$tdir/$tfile" + + rm -rf $DIR/$tdir || error "Can not delete directory $DIR/$tdir" +} +run_test 154B "verify the ll_decode_linkea tool" + test_154a() { [ $PARALLEL == "yes" ] && skip "skip parallel run" && return [[ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.2.51) ]] || @@ -10390,67 +10572,114 @@ test_155_big_load() { true } +save_writethrough() { + local facets=$(get_facets OST) + + save_lustre_params $facets "obdfilter.*.writethrough_cache_enable" > $1 + save_lustre_params $facets "osd-*.*.writethrough_cache_enable" >> $1 +} + test_155a() { [ $PARALLEL == "yes" ] && skip "skip parallel run" && return + local p="$TMP/$TESTSUITE-$TESTNAME.parameters" + save_writethrough $p + set_cache read on set_cache writethrough on test_155_small_load + restore_lustre_params < $p + rm -f $p } run_test 155a "Verify small file correctness: read cache:on write_cache:on" test_155b() { [ $PARALLEL == "yes" ] && skip "skip parallel run" && return + local p="$TMP/$TESTSUITE-$TESTNAME.parameters" + save_writethrough $p + set_cache read on set_cache writethrough off test_155_small_load + restore_lustre_params < $p + rm -f $p } run_test 155b "Verify small file correctness: read cache:on write_cache:off" test_155c() { [ $PARALLEL == "yes" ] && skip "skip parallel run" && return + local p="$TMP/$TESTSUITE-$TESTNAME.parameters" + save_writethrough $p + set_cache read off set_cache writethrough on test_155_small_load + restore_lustre_params < $p + rm -f $p } run_test 155c "Verify small file correctness: read cache:off write_cache:on" test_155d() { [ $PARALLEL == "yes" ] && skip "skip parallel run" && return + local p="$TMP/$TESTSUITE-$TESTNAME.parameters" + save_writethrough $p + set_cache read off set_cache writethrough off test_155_small_load + restore_lustre_params < $p + rm -f $p } run_test 155d "Verify small file correctness: read cache:off write_cache:off" test_155e() { [ $PARALLEL == "yes" ] && skip "skip parallel run" && return + local p="$TMP/$TESTSUITE-$TESTNAME.parameters" + save_writethrough $p + set_cache read on set_cache writethrough on test_155_big_load + restore_lustre_params < $p + rm -f $p } run_test 155e "Verify big file correctness: read cache:on write_cache:on" test_155f() { [ $PARALLEL == "yes" ] && skip "skip parallel run" && return + local p="$TMP/$TESTSUITE-$TESTNAME.parameters" + save_writethrough $p + set_cache read on set_cache writethrough off test_155_big_load + restore_lustre_params < $p + rm -f $p } run_test 155f "Verify big file correctness: read cache:on write_cache:off" test_155g() { [ $PARALLEL == "yes" ] && skip "skip parallel run" && return + local p="$TMP/$TESTSUITE-$TESTNAME.parameters" + save_writethrough $p + set_cache read off set_cache writethrough on test_155_big_load + restore_lustre_params < $p + rm -f $p } run_test 155g "Verify big file correctness: read cache:off write_cache:on" test_155h() { [ $PARALLEL == "yes" ] && skip "skip parallel run" && return + local p="$TMP/$TESTSUITE-$TESTNAME.parameters" + save_writethrough $p + set_cache read off set_cache writethrough off test_155_big_load + restore_lustre_params < $p + rm -f $p } run_test 155h "Verify big file correctness: read cache:off write_cache:off" @@ -10461,12 +10690,14 @@ test_156() { local BEFORE local AFTER local file="$DIR/$tfile" + local p="$TMP/$TESTSUITE-$TESTNAME.parameters" [ "$(facet_fstype ost1)" = "zfs" -a \ $(lustre_version_code ost1 -lt $(version_code 2.6.93)) ] && skip "LU-1956/LU-2261: stats not implemented on OSD ZFS" && return + save_writethrough $p roc_hit_init log "Turn on read and write cache" @@ -10595,10 +10826,18 @@ test_156() { fi rm -f $file + restore_lustre_params < $p + rm -f $p } 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 $* @@ -10639,10 +10878,12 @@ test_160a() { local CL_USERS="mdd.$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 @@ -10686,9 +10927,9 @@ 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 $USER_REC2" @@ -10703,7 +10944,7 @@ test_160a() { # LU-3446 changelog index reset on MDT restart local MDT_DEV=$(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' ') @@ -10712,19 +10953,19 @@ 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 "$CL_USER other changelog users; can't verify off" fi } run_test 160a "changelog sanity" @@ -10737,10 +10978,12 @@ test_160b() { # LU-3587 local CL_USERS="mdd.$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) @@ -10752,11 +10995,8 @@ 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" @@ -10773,8 +11013,9 @@ 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 @@ -10786,11 +11027,10 @@ 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" @@ -10806,25 +11046,56 @@ test_160d() { [ $PARALLEL == "yes" ] && skip "skip parallel run" && return # Registration step - local USER=$(do_facet mds1 $LCTL --device $MDT0 \ + CL_USER=$(do_facet mds1 $LCTL --device $MDT0 \ changelog_register -n) + trap cleanup_changelog EXIT mkdir -p $DIR/$tdir/migrate_dir - $LFS changelog_clear $MDT0 $USER 0 + $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 $USER 0 + $LFS changelog_clear $MDT0 $CL_USER 0 [ $MIGRATES -eq 1 ] || error "MIGRATE changelog mask count $MIGRATES != 1" # Deregistration step - do_facet mds1 $LCTL --device $MDT0 changelog_deregister $USER + cleanup_changelog } run_test 160d "verify that changelog log catch the migrate event" -test_161a() { +test_160e() { + # Create a user + CL_USER=$(do_facet $SINGLEMDS $LCTL --device $MDT0 \ + changelog_register -n) + echo "Registered as changelog user $CL_USER" + trap cleanup_changelog EXIT + + # Delete a future user (expect fail) + do_facet $SINGLEMDS $LCTL --device $MDT0 changelog_deregister cl77 + local rc=$? + + if [ $rc -eq 0 ]; then + error "Deleted non-existant user cl77" + elif [ $rc -ne 2 ]; then + error "changelog_deregister failed with $rc, " \ + "expected 2 (ENOENT)" + fi + + # Clear to a bad index (1 billion should be safe) + $LFS changelog_clear $MDT0 $CL_USER 1000000000 + rc=$? + + if [ $rc -eq 0 ]; then + error "Successfully cleared to invalid CL index" + elif [ $rc -ne 22 ]; then + error "changelog_clear failed with $rc, expected 22 (EINVAL)" + fi +} +run_test 160e "changelog negative testing" + +test_161a() { [ $PARALLEL == "yes" ] && skip "skip parallel run" && return test_mkdir -p -c1 $DIR/$tdir cp /etc/hosts $DIR/$tdir/$tfile @@ -10927,8 +11198,10 @@ test_161c() { # define CLF_RENAME_LAST 0x0001 # rename overwrite a target having nlink = 1 (changelog flag 0x1) - local USER=$(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 @@ -10937,10 +11210,8 @@ 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," \ @@ -10953,10 +11224,8 @@ 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," \ @@ -10967,10 +11236,8 @@ 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," \ @@ -10981,10 +11248,8 @@ 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," \ @@ -10995,18 +11260,65 @@ 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" +test_161d() { + local user + local pid + local fid + + # cleanup previous run + rm -rf $DIR/$tdir/$tfile + + user=$(do_facet $SINGLEMDS $LCTL --device $MDT0 \ + changelog_register -n) + [[ $? -eq 0 ]] || error "changelog_register failed" + + # work in a standalone dir to avoid locking on $DIR/$MOUNT to + # interfer with $MOUNT/.lustre/fid/ access + mkdir $DIR/$tdir + [[ $? -eq 0 ]] || error "mkdir failed" + + #define OBD_FAIL_LLITE_CREATE_NODE_PAUSE 0x140c | OBD_FAIL_ONCE + $LCTL set_param fail_loc=0x8000140c + # 5s pause + $LCTL set_param fail_val=5 + + # create file + echo foofoo > $DIR/$tdir/$tfile & + pid=$! + + # wait for create to be delayed + sleep 2 + + ps -p $pid + [[ $? -eq 0 ]] || error "create should be blocked" + + local tempfile=$(mktemp) + fid=$(changelog_extract_field $MDT0 "CREAT" "$tfile" "t=") + cat $MOUNT/.lustre/fid/$fid 2>/dev/null >$tempfile || error "cat failed" + # some delay may occur during ChangeLog publishing and file read just + # above, that could allow file write to happen finally + [[ -s $tempfile ]] && echo "file should be empty" + + $LCTL set_param fail_loc=0 + + wait $pid + [[ $? -eq 0 ]] || error "create failed" + + $LFS changelog_clear $MDT0 $user 0 + do_facet $SINGLEMDS $LCTL --device $MDT0 changelog_deregister $user +} +run_test 161d "create with concurrent .lustre/fid access" + check_path() { local expected=$1 shift @@ -11733,7 +12045,7 @@ test_200() { break done - cleanup_pools + destroy_test_pools return $rc } run_test 200 "OST pools" @@ -11925,13 +12237,17 @@ 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 + [[ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.7.1) ]] || + { skip "Need MDS version with at least 2.7.1"; return 0; } + [ $PARALLEL == "yes" ] && skip "skip parallel run" && return remote_mgs_nodsh && skip "remote MGS with nodsh" && return remote_mds_nodsh && skip "remote MDS with nodsh" && return @@ -11996,8 +12312,10 @@ 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 + 2 - (SECONDS - start))) - [ $left -ge 0 ] && echo "sleep $left for expiry" && sleep $((left + 1)) + local left=$((interval_new + 5 - (SECONDS - start))) + [ $left -ge 0 ] && wait_update_facet $SINGLEMDS \ + "lctl get_param *.*.job_stats | grep -c 'job_id.*mkdir'" \ + "0" $left cmd="mkdir $DIR/$tdir.expire" verify_jobstats "$cmd" "$SINGLEMDS" [ $(do_facet $SINGLEMDS lctl get_param *.*.job_stats | @@ -12427,7 +12745,7 @@ test_220() { #LU-325 do_facet ost$((OSTIDX + 1)) lctl set_param fail_val=-1 #define OBD_FAIL_OST_ENOINO 0x229 do_facet ost$((OSTIDX + 1)) lctl set_param fail_loc=0x229 - do_facet mgs $LCTL pool_new $FSNAME.$TESTNAME || return 1 + create_pool $FSNAME.$TESTNAME || return 1 do_facet mgs $LCTL pool_add $FSNAME.$TESTNAME $OST || return 2 $SETSTRIPE $DIR/$tdir -i $OSTIDX -c 1 -p $FSNAME.$TESTNAME @@ -12546,6 +12864,11 @@ run_test 224b "Don't panic on bulk IO failure" test_224c() { # LU-6441 [ $PARALLEL == "yes" ] && skip "skip parallel run" && return remote_mds_nodsh && skip "remote MDS with nodsh" && return + + local p="$TMP/$TESTSUITE-$TESTNAME.parameters" + save_writethrough $p + set_cache writethrough on + local pages_per_rpc=$($LCTL get_param \ osc.*.max_pages_per_rpc) local at_max=$($LCTL get_param -n at_max) @@ -12563,10 +12886,11 @@ test_224c() { # LU-6441 error "conf_param timeout=5 failed" #define OBD_FAIL_PTLRPC_CLIENT_BULK_CB3 0x520 - $LCTL set_param fail_loc=0x520 + do_facet ost1 $LCTL set_param fail_loc=0x520 + $LFS setstripe -c 1 -i 0 $DIR/$tfile dd if=/dev/zero of=$DIR/$tfile bs=8MB count=1 sync - $LCTL set_param fail_loc=0 + do_facet ost1 $LCTL set_param fail_loc=0 set_conf_param_and_check client "$test_at" "$param_at" $at_max || error "conf_param at_max=$at_max failed" @@ -12574,6 +12898,8 @@ test_224c() { # LU-6441 $timeout || error "conf_param timeout=$timeout failed" $LCTL set_param -n $pages_per_rpc + restore_lustre_params < $p + rm -f $p } run_test 224c "Don't hang if one of md lost during large bulk RPC" @@ -13591,7 +13917,7 @@ test_241_dio() { done } -test_241() { +test_241a() { # was test_241 dd if=/dev/zero of=$DIR/$tfile count=1 bs=40960 ls -la $DIR/$tfile cancel_lru_locks osc @@ -13600,7 +13926,7 @@ test_241() { test_241_dio 1000 wait $PID } -run_test 241 "bio vs dio" +run_test 241a "bio vs dio" test_241b() { dd if=/dev/zero of=$DIR/$tfile count=1 bs=40960 @@ -13767,18 +14093,9 @@ test_247e() { run_test 247e "mount .. as fileset" test_248() { - local my_error=error - local fast_read_sav=$($LCTL get_param -n llite.*.fast_read 2>/dev/null) [ -z "$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 @@ -13800,7 +14117,8 @@ test_248() { # 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 $t_slow" + error_not_in_vm "fast read was not 4 times faster: " \ + "$t_fast vs $t_slow" echo "Test 2: verify the performance between big and small read" $LCTL set_param -n llite.*.fast_read=1 @@ -13817,7 +14135,7 @@ test_248() { # 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" + error_not_in_vm "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 @@ -13921,6 +14239,128 @@ test_252() { } run_test 252 "check lr_reader tool" +test_253_fill_ost() { + local size_mb #how many MB should we write to pass watermark + local lwm=$3 #low watermark + local free_10mb #10% of free space + + free_kb=$($LFS df $MOUNT | grep $1 | awk '{ print $4 }') + size_mb=$((free_kb / 1024 - lwm)) + free_10mb=$((free_kb / 10240)) + #If 10% of free space cross low watermark use it + if (( free_10mb > size_mb )); then + size_mb=$free_10mb + else + #At least we need to store 1.1 of difference between + #free space and low watermark + size_mb=$((size_mb + size_mb / 10)) + fi + if (( lwm <= $((free_kb / 1024)) )) || [ ! -f $DIR/$tdir/1 ]; then + dd if=/dev/zero of=$DIR/$tdir/1 bs=1M count=$size_mb \ + oflag=append conv=notrunc + fi + + sleep_maxage + + free_kb=$($LFS df $MOUNT | grep $1 | awk '{ print $4 }') + echo "OST still has $((free_kb / 1024)) mbytes free" +} + +test_253() { + local ostidx=0 + local rc=0 + + [ $PARALLEL == "yes" ] && skip "skip parallel run" && return + remote_mds_nodsh && skip "remote MDS with nodsh" && return + remote_mgs_nodsh && skip "remote MGS with nodsh" && return + + local ost_name=$($LFS osts | + sed -n 's/^'$ostidx': \(.*\)_UUID .*/\1/p') + # on the mdt's osc + local mdtosc_proc1=$(get_mdtosc_proc_path $SINGLEMDS $ost_name) + do_facet $SINGLEMDS $LCTL get_param -n \ + osp.$mdtosc_proc1.reserved_mb_high || + { skip "remote MDS does not support reserved_mb_high" && + return; } + + rm -rf $DIR/$tdir + wait_mds_ost_sync + wait_delete_completed + mkdir $DIR/$tdir + + local last_wm_h=$(do_facet $SINGLEMDS $LCTL get_param -n \ + osp.$mdtosc_proc1.reserved_mb_high) + local last_wm_l=$(do_facet $SINGLEMDS $LCTL get_param -n \ + osp.$mdtosc_proc1.reserved_mb_low) + echo "prev high watermark $last_wm_h, prev low watermark $last_wm_l" + + create_pool $FSNAME.$TESTNAME || error "Pool creation failed" + do_facet mgs $LCTL pool_add $FSNAME.$TESTNAME $ost_name || + error "Adding $ost_name to pool failed" + + # Wait for client to see a OST at pool + wait_update $HOSTNAME "$LCTL get_param -n + lov.$FSNAME-*.pools.$TESTNAME | sort -u | + grep $ost_name" "$ost_name""_UUID" $((TIMEOUT/2)) || + error "Client can not see the pool" + $SETSTRIPE $DIR/$tdir -i $ostidx -c 1 -p $FSNAME.$TESTNAME || + error "Setstripe failed" + + dd if=/dev/zero of=$DIR/$tdir/0 bs=1M count=10 + local blocks=$($LFS df $MOUNT | grep $ost_name | awk '{ print $4 }') + echo "OST still has $((blocks/1024)) mbytes free" + + local new_lwm=$((blocks/1024-10)) + do_facet $SINGLEMDS $LCTL set_param \ + osp.$mdtosc_proc1.reserved_mb_high=$((new_lwm+5)) + do_facet $SINGLEMDS $LCTL set_param \ + osp.$mdtosc_proc1.reserved_mb_low=$new_lwm + + test_253_fill_ost $ost_name $mdtosc_proc1 $new_lwm + + #First enospc could execute orphan deletion so repeat. + test_253_fill_ost $ost_name $mdtosc_proc1 $new_lwm + + local oa_status=$(do_facet $SINGLEMDS $LCTL get_param -n \ + osp.$mdtosc_proc1.prealloc_status) + echo "prealloc_status $oa_status" + + dd if=/dev/zero of=$DIR/$tdir/2 bs=1M count=1 && + error "File creation should fail" + #object allocation was stopped, but we still able to append files + dd if=/dev/zero of=$DIR/$tdir/1 bs=1M seek=6 count=5 oflag=append || + error "Append failed" + rm -f $DIR/$tdir/1 $DIR/$tdir/0 $DIR/$tdir/r* + + wait_delete_completed + + sleep_maxage + + for i in $(seq 10 12); do + dd if=/dev/zero of=$DIR/$tdir/$i bs=1M count=1 2>/dev/null || + error "File creation failed after rm"; + done + + oa_status=$(do_facet $SINGLEMDS $LCTL get_param -n \ + osp.$mdtosc_proc1.prealloc_status) + echo "prealloc_status $oa_status" + + if (( oa_status != 0 )); then + error "Object allocation still disable after rm" + fi + do_facet $SINGLEMDS $LCTL set_param \ + osp.$mdtosc_proc1.reserved_mb_high=$last_wm_h + do_facet $SINGLEMDS $LCTL set_param \ + osp.$mdtosc_proc1.reserved_mb_low=$last_wm_l + + + do_facet mgs $LCTL pool_remove $FSNAME.$TESTNAME $ost_name || + error "Remove $ost_name from pool failed" + do_facet mgs $LCTL pool_destroy $FSNAME.$TESTNAME || + error "Pool destroy fialed" +} +run_test 253 "Check object allocation limit" + test_254() { local cl_user @@ -13962,6 +14402,241 @@ test_254() { } 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 $lowest_speedup ] || + error_not_in_vm "Speedup with willread is less than " \ + "$lowest_speedup%, got $average_ladvise%" + echo "Speedup with willread ladvise: $average_ladvise%" + echo "Speedup with cache: $average_cache%" +} + +test_255a() { + [ $(lustre_version_code ost1) -lt $(version_code 2.8.54) ] && + skip "lustre < 2.8.54 does not support ladvise " && return + + 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 + + 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'" + +facet_meminfo() { + local facet=$1 + local info=$2 + + do_facet $facet "cat /proc/meminfo | grep ^${info}:" | awk '{print $2}' +} + +test_255b() { + ladvise_no_type dontneed $DIR/$tfile && + skip "dontneed ladvise is not supported" && return + + ladvise_no_ioctl $DIR/$tfile && + skip "ladvise ioctl is not supported" && return + + [ $(lustre_version_code ost1) -lt $(version_code 2.8.54) ] && + skip "lustre < 2.8.54 does not support ladvise" && return + + [ "$(facet_fstype ost1)" = "zfs" ] && + skip "zfs-osd does not support dontneed advice" && return + + lfs setstripe -c 1 -i 0 $DIR/$tfile + + local size_mb=100 + local size=$((size_mb * 1048576)) + # In order to prevent disturbance of other processes, only check 3/4 + # of the memory usage + local kibibytes=$((size_mb * 1024 * 3 / 4)) + + dd if=/dev/zero of=$DIR/$tfile bs=1048576 count=$size_mb || + error "dd to $DIR/$tfile failed" + + local total=$(facet_meminfo ost1 MemTotal) + echo "Total memory: $total KiB" + + do_facet ost1 "sync && echo 3 > /proc/sys/vm/drop_caches" + local before_read=$(facet_meminfo ost1 Cached) + echo "Cache used before read: $before_read KiB" + + lfs ladvise -a willread $DIR/$tfile || + error "Ladvise willread failed" + local after_read=$(facet_meminfo ost1 Cached) + echo "Cache used after read: $after_read KiB" + + lfs ladvise -a dontneed $DIR/$tfile || + error "Ladvise dontneed again failed" + local no_read=$(facet_meminfo ost1 Cached) + echo "Cache used after dontneed ladvise: $no_read KiB" + + if [ $total -lt $((before_read + kibibytes)) ]; then + echo "Memory is too small, abort checking" + return 0 + fi + + if [ $((before_read + kibibytes)) -gt $after_read ]; then + error "Ladvise willread should use more memory" \ + "than $kibibytes KiB" + fi + + if [ $((no_read + kibibytes)) -gt $after_read ]; then + error "Ladvise dontneed should release more memory" \ + "than $kibibytes KiB" + fi +} +run_test 255b "check 'lfs ladvise -a dontneed'" + test_256() { local cl_user local cat_sl @@ -14026,6 +14701,29 @@ test_256() { } run_test 256 "Check llog delete for empty and not full state" +test_257() { + remote_mds_nodsh && skip "remote MDS with nodsh" && return + [[ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.8.55) ]] && + skip "Need MDS version at least 2.8.55" && return + + test_mkdir -p $DIR/$tdir + + setfattr -n trusted.name1 -v value1 $DIR/$tdir || + error "setfattr -n trusted.name1=value1 $DIR/$tdir failed" + stat $DIR/$tdir + +#define OBD_FAIL_MDS_XATTR_REP 0x161 + local mdtidx=$($LFS getstripe -M $DIR/$tdir) + local facet=mds$((mdtidx + 1)) + set_nodes_failloc $(facet_active_host $facet) 0x80000161 + getfattr -n trusted.name1 $DIR/$tdir 2> /dev/null + + stop $facet || error "stop MDS failed" + start $facet $(mdsdevname $((mdtidx + 1))) $MDS_MOUNT_OPTS || + error "start MDS fail" +} +run_test 257 "xattr locks are not lost" + test_260() { #define OBD_FAIL_MDC_CLOSE 0x806 $LCTL set_param fail_loc=0x80000806 @@ -14059,7 +14757,11 @@ test_striped_dir() { error "getdirstripe failed" stripe_count=$($LFS getdirstripe -c $DIR/$tdir/striped_dir) if [ "$stripe_count" != "2" ]; then - error "stripe_count is $stripe_count, expect 2" + error "1:stripe_count is $stripe_count, expect 2" + fi + stripe_count=$($LFS getdirstripe -T $DIR/$tdir/striped_dir) + if [ "$stripe_count" != "2" ]; then + error "2:stripe_count is $stripe_count, expect 2" fi stripe_index=$($LFS getdirstripe -i $DIR/$tdir/striped_dir) @@ -14225,13 +14927,13 @@ test_300e() { mkdir $DIR/$tdir/striped_dir/dir_c $LFS setdirstripe -i 0 -c 2 -t all_char $DIR/$tdir/striped_dir/stp_a || - error "set striped dir under striped dir error" + error "set striped adir under striped dir error" - $LFS setdirstripe -i 0 -c 2 -t all_char $DIR/$tdir/striped_dir/stp_b || - error "set striped dir under striped dir error" + $LFS setdirstripe -i 0 -c 2 -H all_char $DIR/$tdir/striped_dir/stp_b || + error "set striped bdir under striped dir error" $LFS setdirstripe -i 0 -c 2 -t all_char $DIR/$tdir/striped_dir/stp_c || - error "set striped dir under striped dir error" + error "set striped cdir under striped dir error" mrename $DIR/$tdir/striped_dir/dir_a $DIR/$tdir/striped_dir/dir_b || error "rename dir under striped dir fails" @@ -14349,6 +15051,24 @@ test_300g() { mkdir $DIR/$tdir mkdir $DIR/$tdir/normal_dir + #Checking when client cache stripe index + $LFS setdirstripe -c$MDSCOUNT $DIR/$tdir/striped_dir + $LFS setdirstripe -D -i1 $DIR/$tdir/striped_dir || + error "create striped_dir failed" + + mkdir $DIR/$tdir/striped_dir/dir1 || + error "create dir1 fails" + stripe_index=$($LFS getdirstripe -i $DIR/$tdir/striped_dir/dir1) + [ $stripe_index -eq 1 ] || + error "dir1 expect 1 got $stripe_index" + + $LFS setdirstripe -i2 $DIR/$tdir/striped_dir/dir2 || + error "create dir2 fails" + stripe_index=$($LFS getdirstripe -i $DIR/$tdir/striped_dir/dir2) + [ $stripe_index -eq 2 ] || + error "dir2 expect 2 got $stripe_index" + + #check default stripe count/stripe index test_300_check_default_striped_dir normal_dir $MDSCOUNT 1 test_300_check_default_striped_dir normal_dir 1 0 test_300_check_default_striped_dir normal_dir 2 1 @@ -14417,11 +15137,23 @@ test_300i() { createmany -o $DIR/$tdir/striped_dir/f- 10 || error "create files under striped dir failed" + $LFS setdirstripe -i0 -c$MDSCOUNT -H all_char $DIR/$tdir/hashdir || + error "set striped hashdir error" + + $LFS setdirstripe -i0 -c$MDSCOUNT -H all_char $DIR/$tdir/hashdir/d0 || + error "create dir0 under hash dir failed" + $LFS setdirstripe -i0 -c$MDSCOUNT -H fnv_1a_64 $DIR/$tdir/hashdir/d1 || + error "create dir1 under hash dir failed" + # unfortunately, we need to umount to clear dir layout cache for now # once we fully implement dir layout, we can drop this umount_client $MOUNT || error "umount failed" mount_client $MOUNT || error "mount failed" + $LFS find -H fnv_1a_64 $DIR/$tdir/hashdir + local dircnt=$($LFS find -H fnv_1a_64 $DIR/$tdir/hashdir | wc -l) + [ $dircnt -eq 1 ] || error "lfs find striped dir got:$dircnt,except:1" + #set the stripe to be unknown hash type #define OBD_FAIL_UNKNOWN_LMV_STRIPE 0x1901 $LCTL set_param fail_loc=0x1901 @@ -14791,16 +15523,202 @@ test_311() { local new_iused for i in $(seq 120); do new_iused=$($LFS df -i | grep OST0000 | awk '{ print $3 }') - [ $new_iused -lt $((old_iused - 900)) ] && break + # system may be too busy to destroy all objs in time, use + # a somewhat small value to not fail autotest + [ $((old_iused - new_iused)) -gt 400 ] && break sleep 1 done echo "waited $i sec, old Iused $old_iused, new Iused $new_iused" - [ $new_iused -lt $((old_iused - 900)) ] || + [ $((old_iused - new_iused)) -gt 400 ] || 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: $blksz, expected: 64k" +} +run_test 312 "make sure ZFS adjusts its block size by write pattern" + +test_313() { + local file=$DIR/$tfile + rm -f $file + $SETSTRIPE -c 1 -i 0 $file || error "setstripe failed" + + # define OBD_FAIL_TGT_RCVD_EIO 0x720 + do_facet ost1 "$LCTL set_param fail_loc=0x720" + dd if=/dev/zero of=$file bs=4096 oflag=direct count=1 && + error "write should failed" + do_facet ost1 "$LCTL set_param fail_loc=0" + rm -f $file +} +run_test 313 "io should fail after last_rcvd update fail" + +test_fake_rw() { + local read_write=$1 + if [ "$read_write" = "write" ]; then + local dd_cmd="dd if=/dev/zero of=$DIR/$tfile" + elif [ "$read_write" = "read" ]; then + local dd_cmd="dd of=/dev/null if=$DIR/$tfile" + else + error "argument error" + fi + + # turn off debug for performance testing + local saved_debug=$($LCTL get_param -n debug) + $LCTL set_param debug=0 + + $SETSTRIPE -c 1 -i 0 $DIR/$tfile + + # get ost1 size - lustre-OST0000 + local ost1_avail_size=$($LFS df | awk /${ost1_svc}/'{ print $4 }') + local blocks=$((ost1_avail_size/2/1024)) # half avail space by megabytes + [ $blocks -gt 1000 ] && blocks=1000 # 1G in maximum + + if [ "$read_write" = "read" ]; then + truncate -s $(expr 1048576 \* $blocks) $DIR/$tfile + fi + + local start_time=$(date +%s.%N) + $dd_cmd bs=1M count=$blocks oflag=sync || + error "real dd $read_write error" + local duration=$(bc <<< "$(date +%s.%N) - $start_time") + + if [ "$read_write" = "write" ]; then + rm -f $DIR/$tfile + fi + + # define OBD_FAIL_OST_FAKE_RW 0x238 + do_facet ost1 $LCTL set_param fail_loc=0x238 + + local start_time=$(date +%s.%N) + $dd_cmd bs=1M count=$blocks oflag=sync || + error "fake dd $read_write error" + local duration_fake=$(bc <<< "$(date +%s.%N) - $start_time") + + if [ "$read_write" = "write" ]; then + # verify file size + cancel_lru_locks osc + $CHECKSTAT -t file -s $((blocks * 1024 * 1024)) $DIR/$tfile || + error "$tfile size not $blocks MB" + fi + do_facet ost1 $LCTL set_param fail_loc=0 + + echo "fake $read_write $duration_fake vs. normal $read_write" \ + "$duration in seconds" + [ $(bc <<< "$duration_fake < $duration") -eq 1 ] || + error_not_in_vm "fake write is slower" + + $LCTL set_param -n debug="$saved_debug" + rm -f $DIR/$tfile +} +test_399a() { # LU-7655 for OST fake write + test_fake_rw write +} +run_test 399a "fake write should not be slower than normal write" + + +test_399b() { # LU-8726 for OST fake read + if [ "$(facet_fstype ost1)" != "ldiskfs" ]; then + skip "only for ldiskfs" && return 0 + fi + test_fake_rw read +} +run_test 399b "fake read should not be slower than normal read" + test_400a() { # LU-1606, was conf-sanity test_74 local extra_flags='' local out=$TMP/$tfile @@ -14867,6 +15785,9 @@ test_401a() { #LU-7437 #count the number of parameters by "list_param -R" local params=$($LCTL list_param -R '*' 2>/dev/null | wc -l) #count the number of parameters by listing proc files + local proc_dirs=$(eval \ls -d $proc_regexp 2>/dev/null) + echo "proc_dirs='$proc_dirs'" + [ -n "$proc_dirs" ] || error "no proc_dirs on $HOSTNAME" local procs=$(find -L $proc_dirs -mindepth 1 -printf '%P\n' 2>/dev/null| sort -u | wc -l) @@ -14953,6 +15874,14 @@ test_401d() { run_test 401d "Verify 'lctl set_param' accepts values containing '='" test_402() { + local server_version=$(lustre_version_code $SINGLEMDS) + [[ $server_version -ge $(version_code 2.7.66) ]] || + [[ $server_version -ge $(version_code 2.7.18.4) && + $server_version -lt $(version_code 2.7.50) ]] || + [[ $server_version -ge $(version_code 2.7.2) && + $server_version -lt $(version_code 2.7.11) ]] || + { skip "Need MDS version 2.7.2+ or 2.7.18.4+ or 2.7.66+"; + return; } remote_mds_nodsh && skip "remote MDS with nodsh" && return $LFS setdirstripe -i 0 $DIR/$tdir || error "setdirstripe -i 0 failed" #define OBD_FAIL_MDS_FLD_LOOKUP 0x15c @@ -15108,6 +16037,331 @@ test_406() { } 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" + +test_408() { + dd if=/dev/zero of=$DIR/$tfile bs=4096 count=1 oflag=direct + + #define OBD_FAIL_OSC_BRW_PREP_REQ2 0x40a + lctl set_param fail_loc=0x8000040a + # let ll_prepare_partial_page() fail + dd if=/dev/zero of=$DIR/$tfile bs=2048 count=1 conv=notrunc || true + + rm -f $DIR/$tfile + + # create at least 100 unused inodes so that + # shrink_icache_memory(0) should not return 0 + touch $DIR/$tfile-{0..100} + rm -f $DIR/$tfile-{0..100} + sync + + echo 2 > /proc/sys/vm/drop_caches +} +run_test 408 "drop_caches should not hang due to page leaks" + +test_409() +{ + [ $MDSCOUNT -lt 2 ] && + skip "We need at least 2 MDTs for this test" && return + + check_mount_and_prep + + mkdir -p $DIR/$tdir || error "(0) Fail to mkdir" + $LFS mkdir -i 1 -c 2 $DIR/$tdir/foo || error "(1) Fail to mkdir" + touch $DIR/$tdir/guard || error "(2) Fail to create" + + local PREFIX=$(str_repeat 'A' 128) + echo "Create 1K hard links start at $(date)" + createmany -l $DIR/$tdir/guard $DIR/$tdir/foo/${PREFIX}_ 1000 || + error "(3) Fail to hard link" + + echo "Links count should be right although linkEA overflow" + stat $DIR/$tdir/guard || error "(4) Fail to stat" + local linkcount=$(stat --format=%h $DIR/$tdir/guard) + [ $linkcount -eq 1001 ] || + error "(5) Unexpected hard links count: $linkcount" + + echo "List all links start at $(date)" + ls -l $DIR/$tdir/foo > /dev/null || + error "(6) Fail to list $DIR/$tdir/foo" + + echo "Unlink hard links start at $(date)" + unlinkmany $DIR/$tdir/foo/${PREFIX}_ 1000 || + error "(7) Fail to unlink" +} +run_test 409 "Large amount of cross-MDTs hard links on the same file" + +prep_801() { + start_full_debug_logging + # cleanup unused barrier locks before test + do_facet mgs $LCTL barrier_rescan $FSNAME || + error "Fail to prep barrier test env" +} + +post_801() { + stop_full_debug_logging +} + +test_801a() { + prep_801 + + #define OBD_FAIL_BARRIER_DELAY 0x2202 + do_facet mgs $LCTL set_param fail_val=3 fail_loc=0x2202 + do_facet mgs $LCTL barrier_freeze $FSNAME 10 & + + sleep 1 + local b_status=$(do_facet mgs $LCTL barrier_stat $FSNAME | + awk '/The barrier for/ { print $7 }') + [ "$b_status" = "'freezing_p1'" ] || + error "(1) unexpected barrier status $b_status" + + do_facet mgs $LCTL set_param fail_val=0 fail_loc=0 + wait + b_status=$(do_facet mgs $LCTL barrier_stat $FSNAME | + awk '/The barrier for/ { print $7 }') + [ "$b_status" = "'frozen'" ] || + error "(2) unexpected barrier status $b_status" + + local expired=$(do_facet mgs $LCTL barrier_stat $FSNAME | + awk '/will be expired/ { print $7 }') + echo "sleep $((expired + 3)) seconds, then the barrier will be expired" + sleep $((expired + 3)) + + b_status=$(do_facet mgs $LCTL barrier_stat $FSNAME | + awk '/The barrier for/ { print $7 }') + [ "$b_status" = "'expired'" ] || + error "(3) unexpected barrier status $b_status" + + do_facet mgs $LCTL barrier_freeze $FSNAME 10 || + error "(4) fail to freeze barrier" + + b_status=$(do_facet mgs $LCTL barrier_stat $FSNAME | + awk '/The barrier for/ { print $7 }') + [ "$b_status" = "'frozen'" ] || + error "(5) unexpected barrier status $b_status" + + #define OBD_FAIL_BARRIER_DELAY 0x2202 + do_facet mgs $LCTL set_param fail_val=3 fail_loc=0x2202 + do_facet mgs $LCTL barrier_thaw $FSNAME & + + sleep 1 + b_status=$(do_facet mgs $LCTL barrier_stat $FSNAME | + awk '/The barrier for/ { print $7 }') + [ "$b_status" = "'thawing'" ] || + error "(6) unexpected barrier status $b_status" + + do_facet mgs $LCTL set_param fail_val=0 fail_loc=0 + wait + b_status=$(do_facet mgs $LCTL barrier_stat $FSNAME | + awk '/The barrier for/ { print $7 }') + [ "$b_status" = "'thawed'" ] || + error "(7) unexpected barrier status $b_status" + + #define OBD_FAIL_BARRIER_FAILURE 0x2203 + do_facet $SINGLEMDS $LCTL set_param fail_loc=0x2203 + do_facet mgs $LCTL barrier_freeze $FSNAME + + b_status=$(do_facet mgs $LCTL barrier_stat $FSNAME | + awk '/The barrier for/ { print $7 }') + [ "$b_status" = "'failed'" ] || + error "(8) unexpected barrier status $b_status" + + do_facet $SINGLEMDS $LCTL set_param fail_loc=0 + do_facet mgs $LCTL barrier_thaw $FSNAME + + post_801 +} +run_test 801a "write barrier user interfaces and stat machine" + +test_801b() { + prep_801 + + mkdir $DIR/$tdir || error "(1) fail to mkdir" + createmany -d $DIR/$tdir/d 6 || "(2) fail to mkdir" + touch $DIR/$tdir/d2/f10 || error "(3) fail to touch" + touch $DIR/$tdir/d3/f11 || error "(4) fail to touch" + touch $DIR/$tdir/d4/f12 || error "(5) fail to touch" + + cancel_lru_locks mdc + + # 180 seconds should be long enough + do_facet mgs $LCTL barrier_freeze $FSNAME 180 + + local b_status=$(do_facet mgs $LCTL barrier_stat $FSNAME | + awk '/The barrier for/ { print $7 }') + [ "$b_status" = "'frozen'" ] || + error "(6) unexpected barrier status $b_status" + + mkdir $DIR/$tdir/d0/d10 & + mkdir_pid=$! + + touch $DIR/$tdir/d1/f13 & + touch_pid=$! + + ln $DIR/$tdir/d2/f10 $DIR/$tdir/d2/f14 & + ln_pid=$! + + mv $DIR/$tdir/d3/f11 $DIR/$tdir/d3/f15 & + mv_pid=$! + + rm -f $DIR/$tdir/d4/f12 & + rm_pid=$! + + stat $DIR/$tdir/d5 || error "(7) stat should succeed" + + # To guarantee taht the 'stat' is not blocked + b_status=$(do_facet mgs $LCTL barrier_stat $FSNAME | + awk '/The barrier for/ { print $7 }') + [ "$b_status" = "'frozen'" ] || + error "(8) unexpected barrier status $b_status" + + # let above commands to run at background + sleep 5 + + ps -p $mkdir_pid || error "(9) mkdir should be blocked" + ps -p $touch_pid || error "(10) touch should be blocked" + ps -p $ln_pid || error "(11) link should be blocked" + ps -p $mv_pid || error "(12) rename should be blocked" + ps -p $rm_pid || error "(13) unlink should be blocked" + + b_status=$(do_facet mgs $LCTL barrier_stat $FSNAME | + awk '/The barrier for/ { print $7 }') + [ "$b_status" = "'frozen'" ] || + error "(14) unexpected barrier status $b_status" + + do_facet mgs $LCTL barrier_thaw $FSNAME + b_status=$(do_facet mgs $LCTL barrier_stat $FSNAME | + awk '/The barrier for/ { print $7 }') + [ "$b_status" = "'thawed'" ] || + error "(15) unexpected barrier status $b_status" + + wait $mkdir_pid || error "(16) mkdir should succeed" + wait $touch_pid || error "(17) touch should succeed" + wait $ln_pid || error "(18) link should succeed" + wait $mv_pid || error "(19) rename should succeed" + wait $rm_pid || error "(20) unlink should succeed" + + post_801 +} +run_test 801b "modification will be blocked by write barrier" + +test_801c() { + [[ $MDSCOUNT -lt 2 ]] && skip "needs >= 2 MDTs" && return + + prep_801 + + stop mds2 || error "(1) Fail to stop mds2" + + do_facet mgs $LCTL barrier_freeze $FSNAME 30 + + local b_status=$(do_facet mgs $LCTL barrier_stat $FSNAME | + awk '/The barrier for/ { print $7 }') + [ "$b_status" = "'expired'" -o "$b_status" = "'failed'" ] || { + do_facet mgs $LCTL barrier_thaw $FSNAME + error "(2) unexpected barrier status $b_status" + } + + do_facet mgs $LCTL barrier_rescan $FSNAME || + error "(3) Fail to rescan barrier bitmap" + + do_facet mgs $LCTL barrier_freeze $FSNAME 10 + + b_status=$(do_facet mgs $LCTL barrier_stat $FSNAME | + awk '/The barrier for/ { print $7 }') + [ "$b_status" = "'frozen'" ] || + error "(4) unexpected barrier status $b_status" + + do_facet mgs $LCTL barrier_thaw $FSNAME + b_status=$(do_facet mgs $LCTL barrier_stat $FSNAME | + awk '/The barrier for/ { print $7 }') + [ "$b_status" = "'thawed'" ] || + error "(5) unexpected barrier status $b_status" + + local devname=$(mdsdevname 2) + + start mds2 $devname $MDS_MOUNT_OPTS || error "(6) Fail to start mds2" + + do_facet mgs $LCTL barrier_rescan $FSNAME || + error "(7) Fail to rescan barrier bitmap" + + post_801 +} +run_test 801c "rescan barrier bitmap" + +saved_MGS_MOUNT_OPTS=$MGS_MOUNT_OPTS +saved_MDS_MOUNT_OPTS=$MDS_MOUNT_OPTS +saved_OST_MOUNT_OPTS=$OST_MOUNT_OPTS + +cleanup_802() { + trap 0 + + stopall + MGS_MOUNT_OPTS=$saved_MGS_MOUNT_OPTS + MDS_MOUNT_OPTS=$saved_MDS_MOUNT_OPTS + OST_MOUNT_OPTS=$saved_OST_MOUNT_OPTS + setupall +} + +test_802() { + mkdir $DIR/$tdir || error "(1) fail to mkdir" + + cp $LUSTRE/tests/test-framework.sh $DIR/$tdir/ || + error "(2) Fail to copy" + + trap cleanup_802 EXIT + + # sync by force before remount as readonly + sync; sync_all_data; sleep 3; sync_all_data + + stopall + + MGS_MOUNT_OPTS=$(csa_add "$MGS_MOUNT_OPTS" -o rdonly_dev) + MDS_MOUNT_OPTS=$(csa_add "$MDS_MOUNT_OPTS" -o rdonly_dev) + OST_MOUNT_OPTS=$(csa_add "$OST_MOUNT_OPTS" -o rdonly_dev) + + echo "Mount the server as read only" + setupall server_only || error "(3) Fail to start servers" + + echo "Mount client without ro should fail" + mount_client $MOUNT && + error "(4) Mount client without 'ro' should fail" + + echo "Mount client with ro should succeed" + mount_client $MOUNT ro || + error "(5) Mount client with 'ro' should succeed" + + echo "Modify should be refused" + touch $DIR/$tdir/guard && error "(6) Touch should fail under ro mode" + + echo "Read should be allowed" + diff $LUSTRE/tests/test-framework.sh $DIR/$tdir/test-framework.sh || + error "(7) Read should succeed under ro mode" + + cleanup_802 +} +run_test 802 "simulate readonly device" + # # tests that do cleanup/setup should be run at the end #