X-Git-Url: https://git.whamcloud.com/?p=fs%2Flustre-release.git;a=blobdiff_plain;f=lustre%2Ftests%2Fsanity.sh;h=16546255fe167635e7d7482735fe49626eb5848e;hp=a5e8bdf0d938c301d8a921d7a43e358093005902;hb=e240fb5099af8e62c532d314317095800ebb6864;hpb=a2ba6c0c6e7b22ae31280253cd916fa3e29bd749 diff --git a/lustre/tests/sanity.sh b/lustre/tests/sanity.sh index a5e8bdf..1654625 100755 --- a/lustre/tests/sanity.sh +++ b/lustre/tests/sanity.sh @@ -81,8 +81,8 @@ init_logging if [ $(facet_fstype $SINGLEMDS) = "zfs" ]; then # bug number for skipped test: LU-4536 LU-1957 ALWAYS_EXCEPT="$ALWAYS_EXCEPT 65ic 180" - # 4 13 (min)" - [ "$SLOW" = "no" ] && EXCEPT_SLOW="$EXCEPT_SLOW 51b 51ba" + # 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 @@ -2850,6 +2865,15 @@ 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 +3127,7 @@ 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 +3145,7 @@ 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 @@ -3783,7 +3807,7 @@ 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 & @@ -3794,7 +3818,7 @@ test_43() { $DIR/$tdir/$tfile && error || 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 @@ -3835,12 +3859,12 @@ test_43c() { } 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 +4121,12 @@ 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/$tdir @@ -4118,10 +4148,12 @@ test_51b() { # need to check free space for the directories as well local blkfree=$(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:$dir" @@ -4133,6 +4165,9 @@ test_51b() { local left=100 local delete=$((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: $nrdirs" @@ -4151,6 +4186,8 @@ test_51b() { local after=$(stat -c %h $dir) [[ $after -gt 2 ]] && error "nlink after: $after > 2" || echo "nlink after: $after" + + cleanup_print_lfs_df } run_test 51b "exceed 64k subdirectory nlink limit on create, verify unlink" @@ -4423,30 +4460,47 @@ test_56a() { # was test_56 error "$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 $FILENUM ]] || @@ -4792,7 +4846,7 @@ run_test 56s "check lfs find -stripe-count works" test_56t() { # LU-611 TDIR=$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} @@ -5692,10 +5746,6 @@ cleanup_68() { rm -f $LLOOP unset LLOOP fi - if [ ! -z "$LLITELOOPLOAD" ]; then - rmmod llite_lloop - unset LLITELOOPLOAD - fi rm -f $DIR/f68* } @@ -5707,74 +5757,6 @@ 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() { @@ -6155,6 +6137,11 @@ 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; } @@ -6184,6 +6171,8 @@ 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)) @@ -6196,7 +6185,7 @@ 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 ============" @@ -6678,6 +6667,10 @@ 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 @@ -6688,13 +6681,16 @@ 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" @@ -7127,29 +7123,31 @@ 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) @@ -7286,7 +7284,8 @@ 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" @@ -7296,6 +7295,15 @@ 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" @@ -8331,6 +8339,7 @@ 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" | @@ -8373,6 +8382,7 @@ 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" | @@ -8667,11 +8677,18 @@ 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 @@ -10128,20 +10145,43 @@ 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 $(version_code 2.2.51) ]] || @@ -10416,67 +10456,114 @@ test_155_big_load() { true } +save_writethrough() { + local facets=$(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 $p } run_test 155h "Verify big file correctness: read cache:off write_cache:off" @@ -10487,12 +10574,14 @@ test_156() { local BEFORE local AFTER local file="$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" @@ -10621,10 +10710,18 @@ 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 $* @@ -10665,10 +10762,12 @@ test_160a() { local CL_USERS="mdd.$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 @@ -10712,9 +10811,9 @@ 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 $USER_REC2" @@ -10729,7 +10828,7 @@ test_160a() { # LU-3446 changelog index reset on MDT restart local MDT_DEV=$(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' ') @@ -10738,19 +10837,19 @@ 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 "$CL_USER other changelog users; can't verify off" fi } run_test 160a "changelog sanity" @@ -10763,10 +10862,12 @@ test_160b() { # LU-3587 local CL_USERS="mdd.$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) @@ -10778,11 +10879,8 @@ 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" @@ -10799,8 +10897,9 @@ 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 @@ -10812,11 +10911,10 @@ 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" @@ -10832,21 +10930,22 @@ 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 $USER + cleanup_changelog } run_test 160d "verify that changelog log catch the migrate event" @@ -10953,8 +11052,10 @@ test_161c() { # define CLF_RENAME_LAST 0x0001 # rename overwrite a target having nlink = 1 (changelog flag 0x1) - local USER=$(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 @@ -10963,10 +11064,8 @@ 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," \ @@ -10979,10 +11078,8 @@ 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," \ @@ -10993,10 +11090,8 @@ 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," \ @@ -11007,10 +11102,8 @@ 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," \ @@ -11021,15 +11114,13 @@ test_161c() { rm -f $DIR/$tdir/foobar_161c $LFS changelog $MDT0 | grep UNLNK flags=$($LFS changelog $MDT0 | grep UNLNK | tail -1 | cut -f5 -d' ') - $LFS changelog_clear $MDT0 $USER 0 + $LFS changelog_clear $MDT0 $CL_USER 0 if [ x$flags != "x0x0" ]; then - do_facet $SINGLEMDS $LCTL --device $MDT0 changelog_deregister \ - $USER error "flag $flags is not 0x0" fi echo "unlink a file having nlink > 1," \ "changelog record has flags of $flags" - do_facet $SINGLEMDS $LCTL --device $MDT0 changelog_deregister $USER + cleanup_changelog } run_test 161c "check CL_RENME[UNLINK] changelog record flags" @@ -11951,10 +12042,11 @@ jobstats_set() { } cleanup_205() { + trap 0 do_facet $SINGLEMDS \ $LCTL set_param mdt.*.job_cleanup_interval=$OLD_INTERVAL [ $OLD_JOBENV != $JOBENV ] && jobstats_set $OLD_JOBENV - do_facet $SINGLEMDS lctl --device $MDT0 changelog_deregister $CL_USER + cleanup_changelog } test_205() { # Job stats @@ -12022,8 +12114,10 @@ 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 | @@ -12572,6 +12666,11 @@ 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) @@ -12589,10 +12688,11 @@ 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" @@ -12600,6 +12700,8 @@ test_224c() { # LU-6441 $timeout || error "conf_param timeout=$timeout failed" $LCTL set_param -n $pages_per_rpc + restore_lustre_params < $p + rm -f $p } run_test 224c "Don't hang if one of md lost during large bulk RPC" @@ -13617,7 +13719,7 @@ test_241_dio() { done } -test_241() { +test_241a() { # was test_241 dd if=/dev/zero of=$DIR/$tfile count=1 bs=40960 ls -la $DIR/$tfile cancel_lru_locks osc @@ -13626,7 +13728,7 @@ 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/$tfile count=1 bs=40960 @@ -13793,18 +13895,9 @@ test_247e() { run_test 247e "mount .. as fileset" test_248() { - local my_error=error - local fast_read_sav=$($LCTL get_param -n llite.*.fast_read 2>/dev/null) [ -z "$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 @@ -13826,7 +13919,8 @@ 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 @@ -13843,7 +13937,7 @@ 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 @@ -13947,6 +14041,129 @@ 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 | grep ${ostidx}": " | \ + awk '{print $2}' | sed -e 's/_UUID$//') + # 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" + + do_facet mgs $LCTL pool_new $FSNAME.$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 @@ -14023,6 +14240,7 @@ 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 @@ -14072,8 +14290,8 @@ ladvise_willread_performance() local lowest_speedup=$((average_cache / 2)) [ $average_ladvise -gt $lowest_speedup ] || - error "Speedup with willread is less than $lowest_speedup%,"\ - "got $average_ladvise%" + error_not_in_vm "Speedup with willread is less than " \ + "$lowest_speedup%, got $average_ladvise%" echo "Speedup with willread ladvise: $average_ladvise%" echo "Speedup with cache: $average_cache%" } @@ -14157,6 +14375,71 @@ test_255a() { } 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 @@ -14221,6 +14504,29 @@ 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 @@ -14986,16 +15292,156 @@ 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 + [ $((old_iused - new_iused)) -gt 800 ] && 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 800 ] || error "objs not destroyed after unlink" } run_test 311 "disable OSP precreate, and unlink should destroy objs" +zfs_oid_to_objid() +{ + local ost=$1 + local objid=$2 + + local vdevdir=$(dirname $(facet_vdevice $ost)) + local cmd="$ZDB -e -p $vdevdir -dddd $(facet_device $ost)" + local zfs_zapid=$(do_facet $ost $cmd | + grep -w "/O/0/d$((objid%32))" -C 5 | + awk '/Object/{getline; print $1}') + local zfs_objid=$(do_facet $ost $cmd $zfs_zapid | + awk "/$objid = /"'{printf $3}') + + echo $zfs_objid +} + +zfs_object_blksz() { + local ost=$1 + local objid=$2 + + local vdevdir=$(dirname $(facet_vdevice $ost)) + local cmd="$ZDB -e -p $vdevdir -dddd $(facet_device $ost)" + local blksz=$(do_facet $ost $cmd $objid | + awk '/dblk/{getline; printf $4}') + + case "${blksz: -1}" in + k|K) blksz=$((${blksz:0:$((${#blksz} - 1))}*1024)) ;; + m|M) blksz=$((${blksz:0:$((${#blksz} - 1))}*1024*1024)) ;; + *) ;; + esac + + echo $blksz +} + +test_312() { # LU-4856 + [ $(facet_fstype ost1) = "zfs" ] || + { skip "the test only applies to zfs" && return; } + + local max_blksz=$(do_facet ost1 \ + $ZFS get -p recordsize $(facet_device ost1) | + awk '!/VALUE/{print $3}') + + # to make life a little bit easier + $LFS mkdir -c 1 -i 0 $DIR/$tdir + $LFS setstripe -c 1 -i 0 $DIR/$tdir + + local tf=$DIR/$tdir/$tfile + touch $tf + local oid=$($LFS getstripe $tf | awk '/obdidx/{getline; print $2}') + + # Get ZFS object id + local zfs_objid=$(zfs_oid_to_objid ost1 $oid) + + # block size change by sequential over write + local blksz + for ((bs=4096; bs <= max_blksz; bs <<= 2)); do + dd if=/dev/zero of=$tf bs=$bs count=1 oflag=sync conv=notrunc + + blksz=$(zfs_object_blksz ost1 $zfs_objid) + [ $blksz -eq $bs ] || error "blksz error: $blksz, expected: $bs" + done + rm -f $tf + + # block size change by sequential append write + dd if=/dev/zero of=$tf bs=4K count=1 oflag=sync conv=notrunc + oid=$($LFS getstripe $tf | awk '/obdidx/{getline; print $2}') + zfs_objid=$(zfs_oid_to_objid ost1 $oid) + + for ((count = 1; count < $((max_blksz / 4096)); count *= 2)); do + dd if=/dev/zero of=$tf bs=4K count=$count seek=$count \ + oflag=sync conv=notrunc + + blksz=$(zfs_object_blksz ost1 $zfs_objid) + blksz=$((blksz / 8192)) # in 2*4K unit + [ $blksz -eq $count ] || + error "blksz error(in 8k): $blksz, expected: $count" + done + rm -f $tf + + # random write + touch $tf + oid=$($LFS getstripe $tf | awk '/obdidx/{getline; print $2}') + zfs_objid=$(zfs_oid_to_objid ost1 $oid) + + dd if=/dev/zero of=$tf bs=8K count=1 oflag=sync conv=notrunc + blksz=$(zfs_object_blksz ost1 $zfs_objid) + [ $blksz -eq 8192 ] || error "blksz error: $blksz, expected: 8k" + + dd if=/dev/zero of=$tf bs=64K count=1 oflag=sync conv=notrunc seek=128 + blksz=$(zfs_object_blksz ost1 $zfs_objid) + [ $blksz -eq 65536 ] || error "blksz error: $blksz, expected: 64k" + + dd if=/dev/zero of=$tf bs=1M count=1 oflag=sync conv=notrunc + blksz=$(zfs_object_blksz ost1 $zfs_objid) + [ $blksz -eq 65536 ] || error "rewrite error: $blksz, expected: 64k" +} +run_test 312 "make sure ZFS adjusts its block size by write pattern" + +test_399() { # LU-7655 for OST fake write + # turn off debug for performance testing + local saved_debug=$($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 + [ $blocks -gt 1000 ] && blocks=1000 # 1G in maximum + + local start_time=$(date +%s.%N) + dd if=/dev/zero of=$DIR/$tfile bs=1M count=$blocks oflag=sync || + error "real dd writing error" + local duration=$(bc <<< "$(date +%s.%N) - $start_time") + rm -f $DIR/$tfile + + # define OBD_FAIL_OST_FAKE_WRITE 0x238 + do_facet ost1 $LCTL set_param fail_loc=0x238 + + local start_time=$(date +%s.%N) + dd if=/dev/zero of=$DIR/$tfile bs=1M count=$blocks oflag=sync || + error "fake dd writing error" + local duration_fake=$(bc <<< "$(date +%s.%N) - $start_time") + + # verify file size + cancel_lru_locks osc + $CHECKSTAT -t file -s $((blocks * 1024 * 1024)) $DIR/$tfile || + error "$tfile size not $blocks MB" + + do_facet ost1 $LCTL set_param fail_loc=0 + + echo "fake write $duration_fake vs. normal 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 +} +run_test 399 "fake write should not be slower than normal write" + test_400a() { # LU-1606, was conf-sanity test_74 local extra_flags='' local out=$TMP/$tfile @@ -15303,6 +15749,49 @@ 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" + # # tests that do cleanup/setup should be run at the end #