X-Git-Url: https://git.whamcloud.com/?p=fs%2Flustre-release.git;a=blobdiff_plain;f=lustre%2Ftests%2Fsanity.sh;h=5ec81b23d288f59cf63da31f6b2b7f9e70f177ba;hp=507812df6c524d601d11b9a57c7ea199d0b46865;hb=bc49f2b84d4403583b485ab9736cf6a9464977d4;hpb=6cde14a5df781ae29da88f98a2559eb4342fe1f3 diff --git a/lustre/tests/sanity.sh b/lustre/tests/sanity.sh index 507812d..5ec81b2 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 { @@ -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 @@ -2042,7 +2057,7 @@ test_27D() { llapi_layout_test -d$DIR/$tdir -p$POOL -o$OSTCOUNT $skip27D || error "llapi_layout_test failed" - cleanup_pools || error "cleanup_pools failed" + destroy_test_pools || error "destroy test pools failed" } run_test 27D "validate llapi_layout API" @@ -2411,10 +2426,10 @@ run_test 31p "remove of open striped directory" cleanup_test32_mount() { local rc=0 trap 0 - local loopdev=$(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 $rc } @@ -2598,59 +2613,49 @@ test_32o() { run_test 32o "stat d32o/symlink->tmp/symlink->lustre-root/$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 +2855,15 @@ 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 +3117,7 @@ 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 +3135,7 @@ 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 +3797,7 @@ 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 +3808,7 @@ 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 +3849,12 @@ 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 +4111,12 @@ 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/$tdir @@ -4118,10 +4138,12 @@ test_51b() { # need to check free space for the directories as well local blkfree=$(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:$dir" @@ -4133,6 +4155,9 @@ test_51b() { local left=100 local delete=$((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: $nrdirs" @@ -4151,6 +4176,8 @@ test_51b() { local after=$(stat -c %h $dir) [[ $after -gt 2 ]] && error "nlink after: $after > 2" || echo "nlink after: $after" + + cleanup_print_lfs_df } run_test 51b "exceed 64k subdirectory nlink limit on create, verify unlink" @@ -4423,30 +4450,47 @@ test_56a() { # was test_56 error "$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 $FILENUM ]] || @@ -4792,7 +4836,7 @@ run_test 56s "check lfs find -stripe-count works" test_56t() { # LU-611 TDIR=$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 +5736,6 @@ cleanup_68() { rm -f $LLOOP unset LLOOP fi - if [ ! -z "$LLITELOOPLOAD" ]; then - rmmod llite_lloop - unset LLITELOOPLOAD - fi rm -f $DIR/f68* } @@ -5707,74 +5747,6 @@ 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() { @@ -6118,22 +6090,6 @@ 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 @@ -6141,10 +6097,15 @@ 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 $MOUNT @@ -6685,6 +6646,10 @@ run_test 101e "check read-ahead for small read(1k) for small files(500k)" test_101f() { which iozone || { skip "no iozone installed" && return; } + local old_debug=$($LCTL get_param debug) + old_debug=${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 @@ -6695,71 +6660,87 @@ 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 } @@ -7134,29 +7115,31 @@ 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) @@ -7293,7 +7276,8 @@ 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" @@ -7572,20 +7556,24 @@ 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() @@ -7608,9 +7596,9 @@ 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 $SINGLEMDS \ lctl get_param -n *.*MDT0000-mdtlov.qos_threshold_rr | head -n1) @@ -7624,15 +7612,15 @@ test_116a() { # was previously test_116() else # generate uneven OSTs. Write 2% over the QOS threshold value echo "no" - DIFF=$(($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-$i \ bs=2M count=1 2>/dev/null echo -n . @@ -7643,10 +7631,10 @@ test_116a() { # was previously test_116() free_min_max fi - DIFF=$(($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 $threshold ]; then echo "ok" else echo "failed - QOS mode won't be used" @@ -7655,12 +7643,14 @@ test_116a() { # was previously test_116() return fi - MINI1=$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 [ $FILL -gt 600 ]; then FILL=600 fi @@ -7678,31 +7668,31 @@ test_116a() { # was previously test_116() echo "Note: free space may not be updated, so measurements might be off" free_min_max - DIFF2=$(($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 ]] || @@ -8347,6 +8337,7 @@ 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" | @@ -8389,6 +8380,7 @@ 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" | @@ -8683,11 +8675,18 @@ 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 @@ -9345,22 +9344,27 @@ 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 $want" fi } @@ -9460,17 +9464,17 @@ test_133c() { dd if=/dev/zero of=${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} } @@ -10144,20 +10148,43 @@ 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 $(version_code 2.2.51) ]] || @@ -10925,6 +10952,36 @@ test_160d() { } run_test 160d "verify that changelog log catch the migrate event" +test_160e() { + # Create a user + CL_USER=$(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 @@ -11826,7 +11883,7 @@ test_200() { break done - cleanup_pools + destroy_test_pools return $rc } run_test 200 "OST pools" @@ -12026,6 +12083,9 @@ cleanup_205() { } 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 @@ -12090,8 +12150,10 @@ 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 $SINGLEMDS lctl get_param *.*.job_stats | @@ -12521,7 +12583,7 @@ test_220() { #LU-325 do_facet ost$((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 @@ -13693,7 +13755,7 @@ 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 @@ -13702,7 +13764,7 @@ 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 @@ -14015,6 +14077,128 @@ 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 @@ -14148,6 +14332,9 @@ ladvise_willread_performance() } 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 && @@ -14156,9 +14343,6 @@ test_255a() { ladvise_no_ioctl $DIR/$tfile && skip "ladvise ioctl is not supported" && return - [ $(lustre_version_code ost1) -lt $(version_code 2.8.54) ] && - skip "lustre < 2.8.54 does not support ladvise " && return - local size_mb=100 local size=$((size_mb * 1048576)) dd if=/dev/zero of=$DIR/$tfile bs=1048576 count=$size_mb || @@ -14226,6 +14410,71 @@ 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 @@ -14636,6 +14885,24 @@ 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 @@ -15078,12 +15345,14 @@ test_311() { local new_iused for i in $(seq 120); do new_iused=$($LFS df -i | grep OST0000 | awk '{ print $3 }') - [ $((old_iused - new_iused)) -gt 800 ] && break + # 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" - [ $((old_iused - new_iused)) -gt 800 ] || + [ $((old_iused - new_iused)) -gt 400 ] || error "objs not destroyed after unlink" } run_test 311 "disable OSP precreate, and unlink should destroy objs" @@ -15186,7 +15455,30 @@ test_312() { # LU-4856 } run_test 312 "make sure ZFS adjusts its block size by write pattern" -test_399() { # LU-7655 for OST fake write +test_313() { + local file=$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) $LCTL set_param debug=0 @@ -15198,35 +15490,56 @@ test_399() { # LU-7655 for OST fake write local blocks=$((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=$(date +%s.%N) - dd if=/dev/zero of=$DIR/$tfile bs=1M count=$blocks oflag=sync || - error "real dd writing error" + $dd_cmd bs=1M count=$blocks oflag=sync || + error "real dd $read_write error" local duration=$(bc <<< "$(date +%s.%N) - $start_time") - rm -f $DIR/$tfile - # define OBD_FAIL_OST_FAKE_WRITE 0x238 + 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 if=/dev/zero of=$DIR/$tfile bs=1M count=$blocks oflag=sync || - error "fake dd writing error" + $dd_cmd bs=1M count=$blocks oflag=sync || + error "fake dd $read_write error" local duration_fake=$(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 $blocks MB" - + if [ "$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 write $duration_fake vs. normal write $duration in seconds" + 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 } -run_test 399 "fake write should not be slower than normal write" +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 [ "$(facet_fstype ost1)" != "ldiskfs" ]; then + skip "only for ldiskfs" && return 0 + fi + test_fake_rw read +} +run_test 399b "fake read should not be slower than normal read" test_400a() { # LU-1606, was conf-sanity test_74 local extra_flags='' @@ -15380,6 +15693,14 @@ test_401d() { run_test 401d "Verify 'lctl set_param' accepts values containing '='" test_402() { + local server_version=$(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 @@ -15558,6 +15879,58 @@ test_407() { } run_test 407 "transaction fail should cause operation fail" +test_408() { + dd if=/dev/zero of=$DIR/$tfile bs=4096 count=1 oflag=direct + + #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" + # # tests that do cleanup/setup should be run at the end #