X-Git-Url: https://git.whamcloud.com/?a=blobdiff_plain;f=lustre%2Ftests%2Fsanity.sh;h=199b055333aa707d86707d1460169e8f5cfe9d25;hb=979dae4f485b3663d55c129e86b4d8101dbe4bed;hp=a90ba467b5c2d3d44d93709c80392ec9833a107e;hpb=2f5556627a19d5d9b62e7c223a134f1b8b9f2a74;p=fs%2Flustre-release.git diff --git a/lustre/tests/sanity.sh b/lustre/tests/sanity.sh index a90ba46..199b055 100644 --- a/lustre/tests/sanity.sh +++ b/lustre/tests/sanity.sh @@ -8,7 +8,7 @@ set -e ONLY=${ONLY:-"$*"} # bug number for skipped test: 13297 2108 9789 3637 9789 3561 12622 12653 12653 5188 10764 16260 -ALWAYS_EXCEPT=" 27u 42a 42b 42c 42d 45 51d 65a 65e 68 75 119d $SANITY_EXCEPT" +ALWAYS_EXCEPT=" 27u 42a 42b 42c 42d 45 51d 65a 65e 68b 75 119d $SANITY_EXCEPT" # bug number for skipped test: 2108 9789 3637 9789 3561 5188/5749 1443 #ALWAYS_EXCEPT=${ALWAYS_EXCEPT:-"27m 42a 42b 42c 42d 45 68 76"} # UPDATE THE COMMENT ABOVE WITH BUG NUMBERS WHEN CHANGING ALWAYS_EXCEPT! @@ -40,13 +40,11 @@ GETSTRIPE=${GETSTRIPE:-"$LFS getstripe"} LSTRIPE=${LSTRIPE:-"$LFS setstripe"} LFIND=${LFIND:-"$LFS find"} LVERIFY=${LVERIFY:-ll_dirstripe_verify} -LSTRIPEINFO=${LSTRIPEINFO:-ll_getstripe_info} LCTL=${LCTL:-lctl} MCREATE=${MCREATE:-mcreate} OPENFILE=${OPENFILE:-openfile} OPENUNLINK=${OPENUNLINK:-openunlink} READS=${READS:-"reads"} -TOEXCL=${TOEXCL:-toexcl} TRUNCATE=${TRUNCATE:-truncate} MUNLINK=${MUNLINK:-munlink} SOCKETSERVER=${SOCKETSERVER:-socketserver} @@ -73,7 +71,7 @@ LUSTRE=${LUSTRE:-$(cd $(dirname $0)/..; echo $PWD)} init_test_env $@ . ${CONFIG:=$LUSTRE/tests/cfg/${NAME}.sh} -[ "$SLOW" = "no" ] && EXCEPT_SLOW="24o 27m 36f 36g 51b 51c 60c 63 64b 68 71 73 77f 78 101 103 115 120g 124b" +[ "$SLOW" = "no" ] && EXCEPT_SLOW="24o 24v 27m 36f 36g 51b 51c 60c 63 64b 68 71 73 77f 78 101 103 115 120g 124b" SANITYLOG=${TESTSUITELOG:-$TMP/$(basename $0 .sh).log} FAIL_ON_ERROR=false @@ -91,11 +89,12 @@ setup() { } check_kernel_version() { - VERSION_FILE=version WANT_VER=$1 - GOT_VER=$(lctl get_param -n $VERSION_FILE | awk '/kernel:/ {print $2}') - [ $GOT_VER == "patchless" ] && return 0 - [ $GOT_VER -ge $WANT_VER ] && return 0 + GOT_VER=$(lctl get_param -n version | awk '/kernel:/ {print $2}') + case $GOT_VER in + patchless|patchless_client) return 0;; + *) [ $GOT_VER -ge $WANT_VER ] && return 0 ;; + esac log "test needs at least kernel version $WANT_VER, running $GOT_VER" return 1 } @@ -112,11 +111,13 @@ check_and_setup_lustre DIR=${DIR:-$MOUNT} assert_DIR -LOVNAME=`lctl get_param -n llite.*.lov.common_name | tail -n 1` -OSTCOUNT=`lctl get_param -n lov.$LOVNAME.numobd` -STRIPECOUNT=`lctl get_param -n lov.$LOVNAME.stripecount` -STRIPESIZE=`lctl get_param -n lov.$LOVNAME.stripesize` -ORIGFREE=`lctl get_param -n lov.$LOVNAME.kbytesavail` +MDT0=$($LCTL get_param -n mdc.*.mds_server_uuid | \ + awk '{gsub(/_UUID/,""); print $1}' | head -1) +LOVNAME=$($LCTL get_param -n llite.*.lov.common_name | tail -n 1) +OSTCOUNT=$($LCTL get_param -n lov.$LOVNAME.numobd) +STRIPECOUNT=$($LCTL get_param -n lov.$LOVNAME.stripecount) +STRIPESIZE=$($LCTL get_param -n lov.$LOVNAME.stripesize) +ORIGFREE=$($LCTL get_param -n lov.$LOVNAME.kbytesavail) MAXFREE=${MAXFREE:-$((200000 * $OSTCOUNT))} [ -f $DIR/d52a/foo ] && chattr -a $DIR/d52a/foo @@ -126,7 +127,7 @@ rm -rf $DIR/[Rdfs][0-9]* # $RUNAS_ID may get set incorrectly somewhere else [ $UID -eq 0 -a $RUNAS_ID -eq 0 ] && error "\$RUNAS_ID set to 0, but \$UID is also 0!" -check_runas_id $RUNAS_ID $RUNAS +check_runas_id $RUNAS_ID $RUNAS_ID $RUNAS build_test_filter @@ -464,6 +465,23 @@ test_17f() { } run_test 17f "symlinks: long and very long symlink name ========================" +test_17g() { + mkdir -p $DIR/$tdir + LONGSYMLINK="$(dd if=/dev/zero bs=4095 count=1 | tr '\0' 'x')" + ln -s $LONGSYMLINK $DIR/$tdir/$tfile + ls -l $DIR/$tdir +} +run_test 17g "symlinks: really long symlink name ===============================" + +test_17h() { #bug 17378 + mkdir -p $DIR/$tdir + $SETSTRIPE $DIR/$tdir -c -1 +#define OBD_FAIL_MDS_LOV_PREP_CREATE 0x141 + do_facet mds lctl set_param fail_loc=0x80000141 + touch $DIR/$tdir/$tfile || true +} +run_test 17h "create objects: lov_free_memmd() doesn't lbug" + test_18() { touch $DIR/f ls $DIR || error @@ -532,12 +550,28 @@ test_22() { } run_test 22 "unpack tar archive as non-root user ===============" -test_23() { - mkdir $DIR/d23 - $TOEXCL $DIR/d23/f23 - $TOEXCL -e $DIR/d23/f23 || error +# was test_23 +test_23a() { + mkdir -p $DIR/$tdir + local file=$DIR/$tdir/$tfile + + openfile -f O_CREAT:O_EXCL $file || error "$file create failed" + openfile -f O_CREAT:O_EXCL $file && + error "$file recreate succeeded" || true } -run_test 23 "O_CREAT|O_EXCL in subdir ==========================" +run_test 23a "O_CREAT|O_EXCL in subdir ==========================" + +test_23b() { # bug 18988 + mkdir -p $DIR/$tdir + local file=$DIR/$tdir/$tfile + + rm -f $file + echo foo > $file || error "write filed" + echo bar >> $file || error "append filed" + $CHECKSTAT -s 8 $file || error "wrong size" + rm $file +} +run_test 23b "O_APPEND check ==========================" test_24a() { echo '== rename sanity ==============================================' @@ -621,7 +655,7 @@ test_24i() { mrename $DIR/R9/f $DIR/R9/a $CHECKSTAT -t file $DIR/R9/f || error $CHECKSTAT -t dir $DIR/R9/a || error - $CHECKSTAT -a file $DIR/R9/a/f || error + $CHECKSTAT -a $DIR/R9/a/f || error } run_test 24i "rename file to dir error: touch f ; mkdir a ; rename f a" @@ -734,6 +768,21 @@ test_24u() { # bug12192 } run_test 24u "create stripe file" +test_24v() { + local NRFILES=100000 + local FREE_INODES=`lfs df -i|grep "filesystem summary" | awk '{print $5}'` + [ $FREE_INODES -lt $NRFILES ] && \ + skip "not enough free inodes $FREE_INODES required $NRFILES" && \ + return + + mkdir -p $DIR/d24v + createmany -m $DIR/d24v/$tfile $NRFILES + ls $DIR/d24v >/dev/null || error "error in listing large dir" + + rm $DIR/d24v -rf +} +run_test 24v "list directory with large files (handle hash collision, bug: 17560)" + test_25a() { echo '== symlink sanity =============================================' @@ -807,6 +856,7 @@ run_test 26f "rm -r of a directory which has recursive symlink =" test_27a() { echo '== stripe sanity ==============================================' mkdir -p $DIR/d27 || error "mkdir failed" + $GETSTRIPE $DIR/d27 $SETSTRIPE $DIR/d27/f0 -c 1 || error "lstripe failed" $CHECKSTAT -t file $DIR/d27/f0 || error "checkstat failed" pass @@ -841,7 +891,7 @@ test_27e() { $SETSTRIPE $DIR/d27/f12 -c 2 && error "lstripe succeeded twice" $CHECKSTAT -t file $DIR/d27/f12 || error "checkstat failed" } -run_test 27e "lstripe existing file (should return error) ======" +run_test 27e "setstripe existing file (should return error) ======" test_27f() { mkdir -p $DIR/d27 @@ -849,7 +899,7 @@ test_27f() { dd if=/dev/zero of=$DIR/d27/f12 bs=4k count=4 || error "dd failed" $GETSTRIPE $DIR/d27/fbad || error "lfs getstripe failed" } -run_test 27f "lstripe with bad stripe size (should return error)" +run_test 27f "setstripe with bad stripe size (should return error)" test_27g() { mkdir -p $DIR/d27 @@ -868,7 +918,7 @@ test_27j() { mkdir -p $DIR/d27 $SETSTRIPE $DIR/d27/f27j -i $OSTCOUNT && error "lstripe failed"||true } -run_test 27j "lstripe with bad stripe offset (should return error)" +run_test 27j "setstripe with bad stripe offset (should return error)" test_27k() { # bug 2844 mkdir -p $DIR/d27 @@ -922,29 +972,42 @@ run_test 27m "create file while OST0 was full ==================" # osc's keep a NOSPC stick flag that gets unset with rmdir reset_enospc() { - [ "$1" ] && FAIL_LOC=$1 || FAIL_LOC=0 + local FAIL_LOC=${1:-0} + local OSTIDX=${2:-""} + mkdir -p $DIR/d27/nospc rmdir $DIR/d27/nospc - do_nodes $(comma_list $(osts_nodes)) lctl set_param fail_loc=$FAIL_LOC + local list=$(comma_list $(osts_nodes)) + [ "$OSTIDX" ] && list=$(facet_host ost$((OSTIDX + 1))) + + do_nodes $list lctl set_param fail_loc=$FAIL_LOC } exhaust_precreations() { - OSTIDX=$1 + local OSTIDX=$1 + local MDSIDX=$(get_mds_dir "$DIR/d27") + echo OSTIDX=$OSTIDX MDSIDX=$MDSIDX + + local OST=$(lfs osts | grep ${OSTIDX}": " | \ + awk '{print $2}' | sed -e 's/_UUID$//') + local MDT_INDEX=$(lfs df | grep "\[MDT:$((MDSIDX - 1))\]" | awk '{print $1}' | \ + sed -e 's/_UUID$//;s/^.*-//') - OST=$(lfs osts | grep ${OSTIDX}": " | \ - awk '{print $2}' | sed -e 's/_UUID$//') # on the mdt's osc - last_id=$(do_facet $SINGLEMDS lctl get_param -n osc.*${OST}-osc-MDT0000.prealloc_last_id) - next_id=$(do_facet $SINGLEMDS lctl get_param -n osc.*${OST}-osc-MDT0000.prealloc_next_id) + local last_id=$(do_facet mds${MDSIDX} lctl get_param -n osc.*${OST}-osc-${MDT_INDEX}.prealloc_last_id) + local next_id=$(do_facet mds${MDSIDX} lctl get_param -n osc.*${OST}-osc-${MDT_INDEX}.prealloc_next_id) - mkdir -p $DIR/d27/${OST} - $SETSTRIPE $DIR/d27/${OST} -i $OSTIDX -c 1 + echo ${OST}-osc-${MDT_INDEX}.prealloc_last_id=$last_id + echo ${OST}-osc-${MDT_INDEX}.prealloc_next_id=$next_id + + mkdir -p $DIR/d27 + $SETSTRIPE $DIR/d27 -i $OSTIDX -c 1 #define OBD_FAIL_OST_ENOSPC 0x215 do_facet ost$((OSTIDX + 1)) lctl set_param fail_loc=0x215 echo "Creating to objid $last_id on ost $OST..." - createmany -o $DIR/d27/${OST}/f $next_id $((last_id - next_id + 2)) - do_facet $SINGLEMDS lctl get_param -n osc.*${OST}-osc-MDT0000.prealloc* | grep '[0-9]' - reset_enospc $2 + createmany -o $DIR/d27/${OST}-f $next_id $((last_id - next_id + 2)) + do_facet mds${MDSIDX} lctl get_param osc.*${OST}-osc-${MDT_INDEX}.prealloc* | grep '[0-9]' + reset_enospc $2 $OSTIDX } exhaust_all_precreations() { @@ -958,14 +1021,14 @@ exhaust_all_precreations() { test_27n() { [ "$OSTCOUNT" -lt "2" ] && skip "too few OSTs" && return remote_mds_nodsh && skip "remote MDS with nodsh" && return - remote_ost_nodsh && skip "remote OST with nodsh" && return + remote_ost_nodsh && skip "remote OST with nodsh" && return reset_enospc rm -f $DIR/d27/f27n exhaust_precreations 0 0x80000215 - + $SETSTRIPE -c -1 $DIR/d27 touch $DIR/d27/f27n || error - + $GETSTRIPE $DIR/d27/f27n reset_enospc } run_test 27n "create file with some full OSTs ==================" @@ -973,7 +1036,7 @@ run_test 27n "create file with some full OSTs ==================" test_27o() { [ "$OSTCOUNT" -lt "2" ] && skip "too few OSTs" && return remote_mds_nodsh && skip "remote MDS with nodsh" && return - remote_ost_nodsh && skip "remote OST with nodsh" && return + remote_ost_nodsh && skip "remote OST with nodsh" && return reset_enospc rm -f $DIR/d27/f27o @@ -990,18 +1053,18 @@ run_test 27o "create file with all full OSTs (should error) ====" test_27p() { [ "$OSTCOUNT" -lt "2" ] && skip "too few OSTs" && return remote_mds_nodsh && skip "remote MDS with nodsh" && return - remote_ost_nodsh && skip "remote OST with nodsh" && return + remote_ost_nodsh && skip "remote OST with nodsh" && return reset_enospc rm -f $DIR/d27/f27p - $MCREATE $DIR/d27/f27p || error - $TRUNCATE $DIR/d27/f27p 80000000 || error - $CHECKSTAT -s 80000000 $DIR/d27/f27p || error + $MCREATE $DIR/d27/f27p || error "mcreate failed" + $TRUNCATE $DIR/d27/f27p 80000000 || error "truncate failed" + $CHECKSTAT -s 80000000 $DIR/d27/f27p || error "checkstat failed" exhaust_precreations 0 0x80000215 - echo foo >> $DIR/d27/f27p || error - $CHECKSTAT -s 80000004 $DIR/d27/f27p || error + echo foo >> $DIR/d27/f27p || error "append failed" + $CHECKSTAT -s 80000004 $DIR/d27/f27p || error "checkstat failed" reset_enospc } @@ -1010,7 +1073,7 @@ run_test 27p "append to a truncated file with some full OSTs ===" test_27q() { [ "$OSTCOUNT" -lt "2" ] && skip "too few OSTs" && return remote_mds_nodsh && skip "remote MDS with nodsh" && return - remote_ost_nodsh && skip "remote OST with nodsh" && return + remote_ost_nodsh && skip "remote OST with nodsh" && return reset_enospc rm -f $DIR/d27/f27q @@ -1031,7 +1094,7 @@ run_test 27q "append to truncated file with all OSTs full (should error) ===" test_27r() { [ "$OSTCOUNT" -lt "2" ] && skip "too few OSTs" && return remote_mds_nodsh && skip "remote MDS with nodsh" && return - remote_ost_nodsh && skip "remote OST with nodsh" && return + remote_ost_nodsh && skip "remote OST with nodsh" && return reset_enospc rm -f $DIR/d27/f27r @@ -1044,9 +1107,13 @@ test_27r() { run_test 27r "stripe file with some full OSTs (shouldn't LBUG) =" test_27s() { # bug 10725 - mkdir -p $DIR/$tdir - $LSTRIPE $DIR/$tdir $((2048 * 1024 * 1024)) -1 2 && \ - error "stripe width >= 2^32 succeeded" || true + mkdir -p $DIR/$tdir + local stripe_size=$((4096 * 1024 * 1024)) # 2^32 + local stripe_count=0 + [ $OSTCOUNT -eq 1 ] || stripe_count=2 + $SETSTRIPE $DIR/$tdir -s $stripe_size -c $stripe_count && \ + error "stripe width >= 2^32 succeeded" || true + } run_test 27s "lsm_xfersize overflow (should error) (bug 10725)" @@ -1064,9 +1131,8 @@ test_27u() { # bug 4900 [ "$OSTCOUNT" -lt "2" ] && skip "too few OSTs" && return remote_mds_nodsh && skip "remote MDS with nodsh" && return - #define OBD_FAIL_MDS_OSC_PRECREATE 0x13d - - do_facet $SINGLEMDS lctl set_param fail_loc=0x13d +#define OBD_FAIL_MDS_OSC_PRECREATE 0x139 + do_facet $SINGLEMDS lctl set_param fail_loc=0x139 mkdir -p $DIR/d27u createmany -o $DIR/d27u/t- 1000 do_facet $SINGLEMDS lctl set_param fail_loc=0 @@ -1083,7 +1149,7 @@ run_test 27u "skip object creation on OSC w/o objects ==========" test_27v() { # bug 4900 [ "$OSTCOUNT" -lt "2" ] && skip "too few OSTs" && return remote_mds_nodsh && skip "remote MDS with nodsh" && return - remote_ost_nodsh && skip "remote OST with nodsh" && return + remote_ost_nodsh && skip "remote OST with nodsh" && return exhaust_all_precreations @@ -1092,15 +1158,17 @@ test_27v() { # bug 4900 touch $DIR/$tdir/$tfile #define OBD_FAIL_TGT_DELAY_PRECREATE 0x705 - lctl set_param fail_loc=0x705 - START=`date +%s` - for F in `seq 1 32`; do - touch $DIR/$tdir/$tfile.$F + # all except ost1 + for (( i=0; i < OSTCOUNT; i++ )) ; do + do_facet ost$i lctl set_param fail_loc=0x705 done - lctl set_param fail_loc=0 + local START=`date +%s` + createmany -o $DIR/$tdir/$tfile 32 - FINISH=`date +%s` - TIMEOUT=`lctl get_param -n timeout` + reset_enospc + + local FINISH=`date +%s` + local TIMEOUT=`lctl get_param -n timeout` [ $((FINISH - START)) -ge $((TIMEOUT / 2)) ] && \ error "$FINISH - $START >= $TIMEOUT / 2" @@ -1111,22 +1179,24 @@ run_test 27v "skip object creation on slow OST =================" test_27w() { # bug 10997 mkdir -p $DIR/d27w || error "mkdir failed" $LSTRIPE $DIR/d27w/f0 -s 65536 || error "lstripe failed" - size=`$LSTRIPEINFO $DIR/d27w/f0 | awk {'print $1'}` + size=`$GETSTRIPE $DIR/d27w/f0 -qs` [ $size -ne 65536 ] && error "stripe size $size != 65536" || true [ "$OSTCOUNT" -lt "2" ] && skip "skipping multiple stripe count/offset test" && return for i in `seq 1 $OSTCOUNT`; do offset=$(($i-1)) $LSTRIPE $DIR/d27w/f$i -c $i -i $offset || error "lstripe -c $i -i $offset failed" - count=`$LSTRIPEINFO $DIR/d27w/f$i | awk {'print $2'}` - index=`$LSTRIPEINFO $DIR/d27w/f$i | awk {'print $3'}` + count=`$GETSTRIPE -qc $DIR/d27w/f$i` + index=`$GETSTRIPE -qo $DIR/d27w/f$i` [ $count -ne $i ] && error "stripe count $count != $i" || true [ $index -ne $offset ] && error "stripe offset $index != $offset" || true done } run_test 27w "check lfs setstripe -c -s -i options =============" -test_28() { +# createtest also checks that device nodes are created and +# then visible correctly (#2091) +test_28() { # bug 2091 mkdir $DIR/d28 $CREATETEST $DIR/d28/ct || error } @@ -1725,9 +1795,15 @@ test_37() { run_test 37 "ls a mounted file system to check old content =====" test_38() { - o_directory $DIR/$tfile + local file=$DIR/$tfile + touch $file + openfile -f O_DIRECTORY $file + local RC=$? + local ENOTDIR=20 + [ $RC -eq 0 ] && error "opened file $file with O_DIRECTORY" || true + [ $RC -eq $ENOTDIR ] || error "error $RC should be ENOTDIR ($ENOTDIR)" } -run_test 38 "open a regular file with O_DIRECTORY ==============" +run_test 38 "open a regular file with O_DIRECTORY should return -ENOTDIR ===" test_39() { touch $DIR/$tfile @@ -2054,11 +2130,7 @@ test_46() { } run_test 46 "dirtying a previously written page ================" -# Check that device nodes are created and then visible correctly (#2091) -test_47() { - cmknod $DIR/test_47_node || error -} -run_test 47 "Device nodes check ================================" +# test_47 is removed "Device nodes check" is moved to test_28 test_48a() { # bug 2399 check_kernel_version 34 || return 0 @@ -2213,7 +2285,7 @@ test_51b() { run_test 51b "mkdir .../t-0 --- .../t-$NUMTEST ====================" test_51bb() { - [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return + [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return local ndirs=${TEST51BB_NDIRS:-10} local nfiles=${TEST51BB_NFILES:-100} @@ -2235,7 +2307,7 @@ test_51bb() { declare -a dirs for ((i=0; i < $ndirs; i++)); do dirs[i]=$dir/$RANDOM - echo Creating directory ${dirs[i]} + echo Creating directory ${dirs[i]} mkdir -p ${dirs[i]} ls $dir echo Creating $nfiles in dir ${dirs[i]} ... @@ -2254,7 +2326,7 @@ test_51bb() { echo "mds $i: inodes count OLD ${OLDUSED[$i]} NEW ${NEWUSED[$i]}" [ ${OLDUSED[$i]} -lt ${NEWUSED[$i]} ] || rc=$((rc + 1)) done - + lctl set_param -n lmv.*.placement=$savePOLICY [ $rc -ne $MDSCOUNT ] || \ @@ -2350,8 +2422,9 @@ test_53() { # only test MDT0000 for value in $(do_facet $SINGLEMDS lctl get_param osc.*-osc-MDT0000.prealloc_last_id) ; do param=`echo ${value[0]} | cut -d "=" -f1` - mds_last=$(do_facet $SINGLEMDS lctl get_param -n $param) ostname=`echo $param | cut -d "." -f2 | cut -d - -f 1-2` + mds_last=$(do_facet $SINGLEMDS lctl get_param -n $param) + ostnum=$(echo $ostname | sed "s/${FSNAME}-OST//g" | awk '{print ($1+1)}' ) ost_last=$(do_facet ost$ostnum lctl get_param -n obdfilter.$ostname.last_id | head -n 1) echo "$ostname.last_id=$ost_last ; MDS.last_id=$mds_last" if [ $ost_last != $mds_last ]; then @@ -2641,11 +2714,11 @@ test_56o() { setup_56 $NUMFILES $NUMDIRS TDIR=$DIR/${tdir}g - utime $TDIR/file1 > /dev/null || error - utime $TDIR/file2 > /dev/null || error - utime $TDIR/dir1 > /dev/null || error - utime $TDIR/dir2 > /dev/null || error - utime $TDIR/dir1/file1 > /dev/null || error + utime $TDIR/file1 > /dev/null || error "utime (1)" + utime $TDIR/file2 > /dev/null || error "utime (2)" + utime $TDIR/dir1 > /dev/null || error "utime (3)" + utime $TDIR/dir2 > /dev/null || error "utime (4)" + utime $TDIR/dir1/file1 > /dev/null || error "utime (5)" EXPECTED=5 NUMS=`$LFIND -mtime +1 $TDIR | wc -l` @@ -2708,7 +2781,7 @@ test_57a() { DEV=$(do_facet $SINGLEMDS lctl get_param -n $MNTDEV) [ -z "$DEV" ] && error "can't access $MNTDEV" for DEV in $(do_facet $SINGLEMDS lctl get_param -n $MNTDEV); do - do_facet $SINGLEMDS dumpe2fs -h $DEV > $TMP/t57a.dump || error "can't access $DEV" + do_facet $SINGLEMDS $DUMPE2FS -h $DEV > $TMP/t57a.dump || error "can't access $DEV" DEVISIZE=`awk '/Inode size:/ { print $3 }' $TMP/t57a.dump` [ "$DEVISIZE" -gt 128 ] || error "inode size $DEVISIZE" rm $TMP/t57a.dump @@ -2717,42 +2790,53 @@ test_57a() { run_test 57a "verify MDS filesystem created with large inodes ==" test_57b() { - FILECOUNT=100 - FILE1=$DIR/d57b/f1 - FILEN=$DIR/d57b/f$FILECOUNT - rm -rf $DIR/d57b || error "removing $DIR/d57b" - mkdir -p $DIR/d57b || error "creating $DIR/d57b" + local dir=$DIR/d57b + + local FILECOUNT=100 + local FILE1=$dir/f1 + local FILEN=$dir/f$FILECOUNT + + rm -rf $dir || error "removing $dir" + mkdir -p $dir || error "creating $dir" + local num=$(get_mds_dir $dir) + local mymds=mds$num + echo "mcreating $FILECOUNT files" - createmany -m $DIR/d57b/f 1 $FILECOUNT || \ - error "creating files in $DIR/d57b" + createmany -m $dir/f 1 $FILECOUNT || \ + error "creating files in $dir" # verify that files do not have EAs yet $GETSTRIPE $FILE1 2>&1 | grep -q "no stripe" || error "$FILE1 has an EA" $GETSTRIPE $FILEN 2>&1 | grep -q "no stripe" || error "$FILEN has an EA" - MDSFREE="`lctl get_param -n osd.*MDT0000.kbytesfree 2> /dev/null`" - MDCFREE="`lctl get_param -n mdc.*.kbytesfree | head -n 1`" + sync + sleep 1 + df $dir #make sure we get new statfs data + local MDSFREE=$(do_facet $mymds lctl get_param -n osd.*MDT000$((num -1)).kbytesfree) + local MDCFREE=$(lctl get_param -n mdc.*MDT000$((num -1))-mdc-*.kbytesfree) echo "opening files to create objects/EAs" - for FILE in `seq -f $DIR/d57b/f%g 1 $FILECOUNT`; do - $OPENFILE -f O_RDWR $FILE > /dev/null || error "opening $FILE" + local FILE + for FILE in `seq -f $dir/f%g 1 $FILECOUNT`; do + $OPENFILE -f O_RDWR $FILE > /dev/null 2>&1 || error "opening $FILE" done # verify that files have EAs now $GETSTRIPE $FILE1 | grep -q "obdidx" || error "$FILE1 missing EA" $GETSTRIPE $FILEN | grep -q "obdidx" || error "$FILEN missing EA" - sleep 1 # make sure we get new statfs data -# MDSFREE2="`lctl get_param -n mds.*.kbytesfree`" -# MDCFREE2="`lctl get_param -n mdc.*.kbytesfree`" -# if [ "$MDCFREE2" -lt "$((MDCFREE - 8))" ]; then -# if [ "$MDSFREE" != "$MDSFREE2" ]; then -# error "MDC before $MDCFREE != after $MDCFREE2" -# else -# echo "MDC before $MDCFREE != after $MDCFREE2" -# echo "unable to confirm if MDS has large inodes" -# fi -# fi - rm -rf $DIR/d57b + sleep 1 #make sure we get new statfs data + df $dir + local MDSFREE2=$(do_facet $mymds lctl get_param -n osd.*MDT000$((num -1)).kbytesfree) + local MDCFREE2=$(lctl get_param -n mdc.*MDT000$((num -1))-mdc-*.kbytesfree) + if [ "$MDCFREE2" -lt "$((MDCFREE - 8))" ]; then + if [ "$MDSFREE" != "$MDSFREE2" ]; then + error "MDC before $MDCFREE != after $MDCFREE2" + else + echo "MDC before $MDCFREE != after $MDCFREE2" + echo "unable to confirm if MDS has large inodes" + fi + fi + rm -rf $dir } run_test 57b "default LOV EAs are stored inside large inodes ===" @@ -2802,8 +2886,8 @@ run_test 60b "limit repeated messages from CERROR/CWARN ========" test_60c() { echo "create 5000 files" createmany -o $DIR/f60c- 5000 -#define OBD_FAIL_MDS_LLOG_CREATE_FAILED 0x13c - lctl set_param fail_loc=0x8000013c +#define OBD_FAIL_MDS_LLOG_CREATE_FAILED 0x137 + lctl set_param fail_loc=0x80000137 unlinkmany $DIR/f60c- 5000 lctl set_param fail_loc=0 } @@ -2981,6 +3065,21 @@ test_65i() { # bug6367 } run_test 65i "set non-default striping on root directory (bug 6367)=" +test_65ia() { # bug12836 + $LFS getstripe $MOUNT || error "getstripe $MOUNT failed" +} +run_test 65ia "getstripe on -1 default directory striping" + +test_65ib() { # bug12836 + $LFS getstripe -v $MOUNT || error "getstripe -v $MOUNT failed" +} +run_test 65ib "getstripe -v on -1 default directory striping" + +test_65ic() { # bug12836 + $LFS find -mtime -1 $MOUNT || error "find $MOUNT failed" +} +run_test 65ic "new find on -1 default directory striping" + test_65j() { # bug6367 sync; sleep 1 # if we aren't already remounting for each test, do so for this test @@ -3045,12 +3144,15 @@ LLOOP= cleanup_68() { trap 0 if [ ! -z "$LLOOP" ]; then - swapoff $LLOOP || error "swapoff failed" + if swapon -s | grep -q $LLOOP; then + swapoff $LLOOP || error "swapoff failed" + fi + $LCTL blockdev_detach $LLOOP || error "detach failed" rm -f $LLOOP unset LLOOP fi - rm -f $DIR/f68 + rm -f $DIR/f68* } meminfo() { @@ -3061,10 +3163,29 @@ swap_used() { swapon -s | awk '($1 == "'$1'") { print $4 }' } +# test case for lloop driver, basic function +test_68a() { + [ "$UID" != 0 ] && skip "must run as root" && return + + grep -q llite_lloop /proc/modules + [ $? -ne 0 ] && skip "can't find module llite_lloop" && return + + LLOOP=$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" + + trap cleanup_68 EXIT + + directio rdwr $LLOOP 0 1024 4096 || error "direct write failed" + directio rdwr $LLOOP 0 1025 4096 && error "direct write should fail" + + cleanup_68 +} +run_test 68a "lloop driver - basic test ========================" # excercise swapping to lustre by adding a high priority swapfile entry # and then consuming memory until it is used. -test_68() { +test_68b() { # was test_68 [ "$UID" != 0 ] && skip "must run as root" && return lctl get_param -n devices | grep -q obdfilter && \ skip "local OST" && return @@ -3080,10 +3201,10 @@ test_68() { [[ $NR_BLOCKS -le 2048 ]] && NR_BLOCKS=2048 LLOOP=$TMP/lloop.`date +%s`.`date +%N` - dd if=/dev/zero of=$DIR/f68 bs=64k seek=$NR_BLOCKS count=1 - mkswap $DIR/f68 + dd if=/dev/zero of=$DIR/f68b bs=64k seek=$NR_BLOCKS count=1 + mkswap $DIR/f68b - $LCTL blockdev_attach $DIR/f68 $LLOOP || error "attach failed" + $LCTL blockdev_attach $DIR/f68b $LLOOP || error "attach failed" trap cleanup_68 EXIT @@ -3098,7 +3219,7 @@ test_68() { [ $SWAPUSED -eq 0 ] && echo "no swap used???" || true } -run_test 68 "support swapping to Lustre ========================" +run_test 68b "support swapping to Lustre ========================" # bug5265, obdfilter oa2dentry return -ENOENT # #define OBD_FAIL_OST_ENOENT 0x217 @@ -3129,7 +3250,8 @@ test_69() { run_test 69 "verify oa2dentry return -ENOENT doesn't LBUG ======" test_71() { - sh rundbench -C -D $DIR 2 || error "dbench failed!" + mkdir -p $DIR/$tdir + sh rundbench -C -D $DIR/$tdir 2 || error "dbench failed!" } run_test 71 "Running dbench on lustre (don't segment fault) ====" @@ -3138,7 +3260,7 @@ test_72() { # bug 5695 - Test that on 2.6 remove_suid works properly [ "$RUNAS_ID" = "$UID" ] && skip "RUNAS_ID = UID = $UID -- skipping" && return # Check that testing environment is properly set up. Skip if not - FAIL_ON_ERROR=false check_runas_id_ret $RUNAS_ID $RUNAS || { + FAIL_ON_ERROR=false check_runas_id_ret $RUNAS_ID $RUNAS_ID $RUNAS || { skip "User $RUNAS_ID does not exist - skipping" return 0 } @@ -3154,6 +3276,7 @@ test_72() { # bug 5695 - Test that on 2.6 remove_suid works properly cancel_lru_locks mdc test -u $DIR/f72 -o -g $DIR/f72 && error "S/gid is not dropped on MDS" true + rm -f $DIR/f72 } run_test 72 "Test that remove suid works properly (bug5695) ====" @@ -3196,6 +3319,7 @@ test_74a() { # bug 6149, 6184 ls $DIR/f74a lctl set_param fail_loc=0 true + rm -f $DIR/f74a } run_test 74a "ldlm_enqueue freed-export error path, ls (shouldn't LBUG)" @@ -3209,6 +3333,7 @@ test_74b() { # bug 13310 touch $DIR/f74b lctl set_param fail_loc=0 true + rm -f $DIR/f74b } run_test 74b "ldlm_enqueue freed-export error path, touch (shouldn't LBUG)" @@ -3229,9 +3354,6 @@ test75_prep() { } test_75a() { -# skipped temporarily: we do not have join file currently -# please remove this when ready - huanghua - return test75_prep cp -p ${F128k} ${FHEAD} @@ -3249,9 +3371,6 @@ test_75a() { run_test 75a "TEST join file ====================================" test_75b() { -# skipped temporarily: we do not have join file currently -# please remove this when ready - huanghua - return test75_prep cp -p ${F128k} ${FTAIL} @@ -3265,9 +3384,6 @@ test_75b() { run_test 75b "TEST join file 2 ==================================" test_75c() { -# skipped temporarily: we do not have join file currently -# please remove this when ready - huanghua - return test75_prep cp -p ${F128k} ${FTAIL} @@ -3281,9 +3397,6 @@ test_75c() { run_test 75c "TEST join file 3 ==================================" test_75d() { -# skipped temporarily: we do not have join file currently -# please remove this when ready - huanghua - return test75_prep cp -p ${F128k} ${FHEAD} @@ -3298,9 +3411,6 @@ test_75d() { run_test 75d "TEST join file 4 ==================================" test_75e() { -# skipped temporarily: we do not have join file currently -# please remove this when ready - huanghua - return test75_prep rm -rf ${FHEAD} || "delete join file error" @@ -3308,9 +3418,6 @@ test_75e() { run_test 75e "TEST join file 5 (remove joined file) =============" test_75f() { -# skipped temporarily: we do not have join file currently -# please remove this when ready - huanghua - return test75_prep cp -p ${F128k} ${F75}_join_10_compare @@ -3328,9 +3435,6 @@ test_75f() { run_test 75f "TEST join file 6 (join 10 files) ==================" test_75g() { -# skipped temporarily: we do not have join file currently -# please remove this when ready - huanghua - return [ ! -f ${F75}_join_10 ] && echo "${F75}_join_10 missing" && return $LFS getstripe ${F75}_join_10 @@ -3397,14 +3501,17 @@ setup_f77() { } test_77a() { # bug 10889 + $GSS && skip "could not run with gss" && return [ ! -f $F77_TMP ] && setup_f77 set_checksums 1 dd if=$F77_TMP of=$DIR/$tfile bs=1M count=$F77SZ || error "dd error" set_checksums 0 + rm -f $DIR/$tfile } run_test 77a "normal checksum read/write operation =============" test_77b() { # bug 10889 + $GSS && skip "could not run with gss" && return [ ! -f $F77_TMP ] && setup_f77 #define OBD_FAIL_OSC_CHECKSUM_SEND 0x409 lctl set_param fail_loc=0x80000409 @@ -3417,6 +3524,7 @@ test_77b() { # bug 10889 run_test 77b "checksum error on client write ====================" test_77c() { # bug 10889 + $GSS && skip "could not run with gss" && return [ ! -f $DIR/f77b ] && skip "requires 77b - skipping" && return set_checksums 1 for algo in $CKSUM_TYPES; do @@ -3429,10 +3537,12 @@ test_77c() { # bug 10889 done set_checksums 0 set_checksum_type $ORIG_CSUM_TYPE + rm -f $DIR/f77b } run_test 77c "checksum error on client read ===================" test_77d() { # bug 10889 + $GSS && skip "could not run with gss" && return #define OBD_FAIL_OSC_CHECKSUM_SEND 0x409 lctl set_param fail_loc=0x80000409 set_checksums 1 @@ -3444,6 +3554,7 @@ test_77d() { # bug 10889 run_test 77d "checksum error on OST direct write ===============" test_77e() { # bug 10889 + $GSS && skip "could not run with gss" && return [ ! -f $DIR/f77 ] && skip "requires 77d - skipping" && return #define OBD_FAIL_OSC_CHECKSUM_RECEIVE 0x408 lctl set_param fail_loc=0x80000408 @@ -3457,6 +3568,7 @@ test_77e() { # bug 10889 run_test 77e "checksum error on OST direct read ================" test_77f() { # bug 10889 + $GSS && skip "could not run with gss" && return set_checksums 1 for algo in $CKSUM_TYPES; do cancel_lru_locks osc @@ -3473,6 +3585,7 @@ test_77f() { # bug 10889 run_test 77f "repeat checksum error on write (expect error) ====" test_77g() { # bug 10889 + $GSS && skip "could not run with gss" && return remote_ost_nodsh && skip "remote OST with nodsh" && return [ ! -f $F77_TMP ] && setup_f77 @@ -3489,6 +3602,7 @@ test_77g() { # bug 10889 run_test 77g "checksum error on OST write ======================" test_77h() { # bug 10889 + $GSS && skip "could not run with gss" && return remote_ost_nodsh && skip "remote OST with nodsh" && return [ ! -f $DIR/f77g ] && skip "requires 77g - skipping" && return @@ -3503,6 +3617,7 @@ test_77h() { # bug 10889 run_test 77h "checksum error on OST read =======================" test_77i() { # bug 13805 + $GSS && skip "could not run with gss" && return #define OBD_FAIL_OSC_CONNECT_CKSUM 0x40b lctl set_param fail_loc=0x40b remount_client $MOUNT @@ -3517,6 +3632,7 @@ test_77i() { # bug 13805 run_test 77i "client not supporting OSD_CONNECT_CKSUM ==========" test_77j() { # bug 13805 + $GSS && skip "could not run with gss" && return #define OBD_FAIL_OSC_CKSUM_ADLER_ONLY 0x40c lctl set_param fail_loc=0x40c remount_client $MOUNT @@ -3543,7 +3659,12 @@ test_78() { # bug 10901 echo "MemTotal: $MEMTOTAL" # reserve 256MB of memory for the kernel and other running processes, # and then take 1/2 of the remaining memory for the read/write buffers. - MEMTOTAL=$(((MEMTOTAL - 256 ) / 2)) + if [ $MEMTOTAL -gt 512 ] ;then + MEMTOTAL=$(((MEMTOTAL - 256 ) / 2)) + else + # for those poor memory-starved high-end clusters... + MEMTOTAL=$((MEMTOTAL / 2)) + fi echo "Mem to use for directio: $MEMTOTAL" [ $F78SIZE -gt $MEMTOTAL ] && F78SIZE=$MEMTOTAL [ $F78SIZE -gt 512 ] && F78SIZE=512 @@ -3553,11 +3674,12 @@ test_78() { # bug 10901 [ $SMALLESTOST -lt 10240 ] && \ skip "too small OSTSIZE, useless to run large O_DIRECT test" && return 0 - [ $F78SIZE -gt $((SMALLESTOST * $OSTCOUNT / 1024 - 5)) ] && \ - F78SIZE=$((SMALLESTOST * $OSTCOUNT / 1024 - 5)) + [ $F78SIZE -gt $((SMALLESTOST * $OSTCOUNT / 1024 - 80)) ] && \ + F78SIZE=$((SMALLESTOST * $OSTCOUNT / 1024 - 80)) + [ "$SLOW" = "no" ] && NSEQ=1 && [ $F78SIZE -gt 32 ] && F78SIZE=32 echo "File size: $F78SIZE" - $SETSTRIPE $DIR/$tfile -c -1 || error "setstripe failed" + $SETSTRIPE $DIR/$tfile -c $OSTCOUNT || error "setstripe failed" for i in `seq 1 $NSEQ` do FSIZE=$(($F78SIZE / ($NSEQ - $i + 1))) @@ -3602,32 +3724,28 @@ run_test 79 "df report consistency check =======================" test_80() { # bug 10718 dd if=/dev/zero of=$DIR/$tfile bs=1M count=1 seek=1M sync; sleep 1; sync - BEFORE=`date +%s` - cancel_lru_locks OSC - AFTER=`date +%s` - DIFF=$((AFTER-BEFORE)) + local BEFORE=`date +%s` + cancel_lru_locks osc + local AFTER=`date +%s` + local DIFF=$((AFTER-BEFORE)) if [ $DIFF -gt 1 ] ; then error "elapsed for 1M@1T = $DIFF" fi true + rm -f $DIR/$tfile } run_test 80 "Page eviction is equally fast at high offsets too ====" -# on the LLNL clusters, runas will still pick up root's $TMP settings, -# which will not be writable for the runas user, and then you get a CVS -# error message with a corrupt path string (CVS bug) and panic. -# We're not using much space, so just stick it in /tmp, which is safe. -OLDTMPDIR=$TMPDIR -OLDTMP=$TMP -TMPDIR=/tmp -TMP=/tmp -OLDHOME=$HOME -[ $RUNAS_ID -ne $UID ] && HOME=/tmp - test_99a() { + [ -z "$(which cvs 2>/dev/null)" ] && skip "could not find cvs" && \ + return mkdir -p $DIR/d99cvsroot chown $RUNAS_ID $DIR/d99cvsroot + local oldPWD=$PWD # bug 13584, use $TMP as working dir + cd $TMP + $RUNAS cvs -d $DIR/d99cvsroot init || error + cd $oldPWD } run_test 99a "cvs init =========================================" @@ -3676,22 +3794,34 @@ test_99f() { [ ! -d $DIR/d99cvsroot ] && test_99d cd $DIR/d99reposname $RUNAS cvs commit -m 'nomsg' foo99 + rm -fr $DIR/d99cvsroot } run_test 99f "cvs commit =======================================" test_100() { - netstat -tna | while read PROT SND RCV LOCAL REMOTE STAT; do + [ "$NETTYPE" = tcp ] || \ + { skip "TCP secure port test, not useful for NETTYPE=$NETTYPE" && \ + return ; } + + remote_ost_nodsh && skip "remote OST with nodsh" && return + remote_mds_nodsh && skip "remote MDS with nodsh" && return + remote_servers || \ + { skip "useless for local single node setup" && return; } + + netstat -tna | ( rc=1; while read PROT SND RCV LOCAL REMOTE STAT; do [ "$PROT" != "tcp" ] && continue - RPORT=`echo $REMOTE | cut -d: -f2` + RPORT=$(echo $REMOTE | cut -d: -f2) [ "$RPORT" != "$ACCEPTOR_PORT" ] && continue + + rc=0 LPORT=`echo $LOCAL | cut -d: -f2` if [ $LPORT -ge 1024 ]; then echo "bad: $PROT $SND $RCV $LOCAL $REMOTE $STAT" netstat -tna - error "local: $LPORT > 1024, remote: $RPORT" + error_exit "local: $LPORT > 1024, remote: $RPORT" fi done - true + [ "$rc" = 0 ] || error_exit "privileged port not found" ) } run_test 100 "check local port using privileged port ===========" @@ -3771,6 +3901,7 @@ cleanup_test101() { [ "$SETUP_TEST101" = "yes" ] || return trap 0 rm -rf $DIR/$tdir + rm -f $DIR/$tfile SETUP_TEST101=no } @@ -3824,41 +3955,87 @@ test_101b() { } run_test 101b "check stride-io mode read-ahead =================" +set_read_ahead() { + lctl get_param -n llite.*.max_read_ahead_mb | head -n 1 + lctl set_param -n llite.*.max_read_ahead_mb $1 > /dev/null 2>&1 +} + +test_101d() { + local file=$DIR/$tfile + local size=${FILESIZE_101c:-500} + local ra_MB=${READAHEAD_MB:-40} + + local space=$(df -P $DIR | tail -n 1 | awk '{ print $4 }') + [ $space -gt $((size / 1024)) ] || + { skip "Need free space ${size}M, have $space" && return; } + + echo Creating ${size}M test file $file + dd if=/dev/zero of=$file bs=1M count=$size + echo Cancel LRU locks on lustre client to flush the client cache + cancel_lru_locks osc + + echo Disable read-ahead + local old_READAHEAD=$(set_read_ahead 0) + + echo Reading the test file $file with read-ahead disabled + time_ra_OFF=$(do_and_time "dd if=$file of=/dev/null bs=1M count=$size") + + echo Cancel LRU locks on lustre client to flush the client cache + cancel_lru_locks osc + echo Enable read-ahead with ${ra_MB}MB + set_read_ahead $ra_MB + + echo Reading the test file $file with read-ahead enabled + time_ra_ON=$(do_and_time "dd if=$file of=/dev/null bs=1M count=$size") + + echo read-ahead disabled time read $time_ra_OFF + echo read-ahead enabled time read $time_ra_ON + + set_read_ahead $old_READAHEAD + rm -f $file + + [ $time_ra_ON -lt $time_ra_OFF ] || + error "read-ahead enabled time read (${time_ra_ON}s) is more than + read-ahead disabled time read (${time_ra_OFF}s) filesize ${size}M" +} +run_test 101d "file read with and without read-ahead enabled =================" + export SETUP_TEST102=no setup_test102() { [ "$SETUP_TEST102" = "yes" ] && return mkdir -p $DIR/$tdir + chown $RUNAS_ID $DIR/$tdir STRIPE_SIZE=65536 - STRIPE_COUNT=4 - STRIPE_OFFSET=2 + STRIPE_OFFSET=1 + STRIPE_COUNT=$OSTCOUNT + [ $OSTCOUNT -gt 4 ] && STRIPE_COUNT=4 trap cleanup_test102 EXIT cd $DIR - $SETSTRIPE $tdir -s $STRIPE_SIZE -i $STRIPE_OFFSET -c $STRIPE_COUNT + $1 $SETSTRIPE $tdir -s $STRIPE_SIZE -i $STRIPE_OFFSET -c $STRIPE_COUNT cd $DIR/$tdir for num in 1 2 3 4 do - for count in 1 2 3 4 + for count in `seq 1 $STRIPE_COUNT` do - for offset in 0 1 2 3 + for offset in `seq 0 $[$STRIPE_COUNT - 1]` do local stripe_size=`expr $STRIPE_SIZE \* $num` local file=file"$num-$offset-$count" - $SETSTRIPE $file -s $stripe_size -i $offset -c $count + $1 $SETSTRIPE $file -s $stripe_size -i $offset -c $count done done done cd $DIR - star -c f=$TMP/f102.tar $tdir + $1 $TAR cf $TMP/f102.tar $tdir --xattrs SETUP_TEST102=yes } cleanup_test102() { - [ "$SETUP_TEST102" = "yes" ] || return trap 0 + [ "$SETUP_TEST102" = "yes" ] || return 0 rm -f $TMP/f102.tar - rm -rf $DIR/$tdir SETUP_TEST102=no } @@ -3916,22 +4093,23 @@ test_102b() { echo "get/set/list trusted.lov xattr ..." [ "$OSTCOUNT" -lt "2" ] && skip "skipping 2-stripe test" && return local testfile=$DIR/$tfile - $SETSTRIPE $testfile -s 65536 -i 1 -c 2 + $SETSTRIPE -s 65536 -i 1 -c 2 $testfile || error "setstripe failed" getfattr -d -m "^trusted" $testfile 2> /dev/null | \ grep "trusted.lov" || error "can't get trusted.lov from $testfile" local testfile2=${testfile}2 local value=`getfattr -n trusted.lov $testfile 2> /dev/null | \ grep "trusted.lov" |sed -e 's/[^=]\+=//'` - + $MCREATE $testfile2 - setfattr -n trusted.lov -v $value $testfile2 + setfattr -n trusted.lov -v $value $testfile2 local tmp_file=${testfile}3 $GETSTRIPE -v $testfile2 > $tmp_file local stripe_size=`grep "size" $tmp_file| awk '{print $2}'` local stripe_count=`grep "count" $tmp_file| awk '{print $2}'` [ "$stripe_size" -eq 65536 ] || error "stripe size $stripe_size != 65536" [ "$stripe_count" -eq 2 ] || error "stripe count $stripe_count != 2" + rm -f $DIR/$tfile } run_test 102b "getfattr/setfattr for trusted.lov EAs ============" @@ -3942,16 +4120,16 @@ test_102c() { mkdir -p $DIR/$tdir chown $RUNAS_ID $DIR/$tdir local testfile=$DIR/$tdir/$tfile - $RUNAS $SETSTRIPE $testfile -s 65536 -i 1 -c 2 + $RUNAS $SETSTRIPE -s 65536 -i 1 -c 2 $testfile||error "setstripe failed" $RUNAS getfattr -d -m "^lustre" $testfile 2> /dev/null | \ grep "lustre.lov" || error "can't get lustre.lov from $testfile" local testfile2=${testfile}2 local value=`getfattr -n lustre.lov $testfile 2> /dev/null | \ grep "lustre.lov" |sed -e 's/[^=]\+=//' ` - + $RUNAS $MCREATE $testfile2 - $RUNAS setfattr -n lustre.lov -v $value $testfile2 + $RUNAS setfattr -n lustre.lov -v $value $testfile2 local tmp_file=${testfile}3 $RUNAS $GETSTRIPE -v $testfile2 > $tmp_file local stripe_size=`grep "size" $tmp_file| awk '{print $2}'` @@ -3961,135 +4139,88 @@ test_102c() { } run_test 102c "non-root getfattr/setfattr for lustre.lov EAs ===========" -get_stripe_info() { - stripe_size=0 - stripe_count=0 - stripe_offset=0 - local lines=`sed -n '/obdidx/=' $1` - stripe_size=`awk '{if($1~/size/) print $2}' $1` - stripe_count=`awk '{if($1~/count/) print $2}' $1` - lines=`expr $lines + 1` - stripe_offset=`sed -n ${lines}p $1 |awk '{print $1}'` -} - compare_stripe_info1() { + local stripe_index_all_zero=1 + for num in 1 2 3 4 do - for count in 1 2 3 4 + for count in `seq 1 $STRIPE_COUNT` do - for offset in 0 1 2 3 + for offset in `seq 0 $[$STRIPE_COUNT - 1]` do local size=`expr $STRIPE_SIZE \* $num` local file=file"$num-$offset-$count" - local tmp_file=out - $GETSTRIPE -v $file > $tmp_file - get_stripe_info $tmp_file - if test $stripe_size -ne $size - then - error "$file: different stripe size" && return + get_stripe_info client $PWD/$file "$1" + if [ $stripe_size -ne $size ]; then + error "$file: different stripe size $stripe_size, expected $size" && return fi - if test $stripe_count -ne $count - then - error "$file: different stripe count" && return + if [ $stripe_count -ne $count ]; then + error "$file: different stripe count $stripe_count, expected $count" && return fi - if test $stripe_offset -ne 0 - then - error "$file: different stripe offset" && return + if [ $stripe_index -ne 0 ]; then + stripe_index_all_zero=0 fi - rm -f $tmp_file done done done + [ $stripe_index_all_zero -eq 1 ] && error "all files are being extracted starting from OST index 0" + return 0 } compare_stripe_info2() { for num in 1 2 3 4 do - for count in 1 2 3 4 + for count in `seq 1 $STRIPE_COUNT` do - for offset in 0 1 2 3 + for offset in `seq 0 $[$STRIPE_COUNT - 1]` do local size=`expr $STRIPE_SIZE \* $num` local file=file"$num-$offset-$count" - local tmp_file=out - $GETSTRIPE -v $file > $tmp_file - get_stripe_info $tmp_file - if test $stripe_size -ne $size - then - error "$file: different stripe size" && return + get_stripe_info client $PWD/$file + if [ $stripe_size -ne $size ]; then + error "$file: different stripe size $stripe_size, expected $size" && return fi - if test $stripe_count -ne $count - then - error "$file: different stripe count" && return + if [ $stripe_count -ne $count ]; then + error "$file: different stripe count $stripe_count, expected $count" && return fi - if test $stripe_offset -ne $offset - then - error "$file: different stripe offset" && return + if [ $stripe_index -ne $offset ]; then + error "$file: different stripe offset $stripe_index, expected $offset" && return fi - rm -f $tmp_file done done done } +find_lustre_tar() { + [ -n "$(which tar 2>/dev/null)" ] && strings $(which tar) | grep -q lustre && echo tar +} + test_102d() { - # b10930: star test for trusted.lov xattr - star --xhelp 2>&1 | grep -q nolustre - if [ $? -ne 0 ] - then - skip "being skipped because a lustre-aware star is not installed." && return - fi - [ "$OSTCOUNT" -lt "4" ] && skip "skipping 4-stripe test" && return + # b10930: tar test for trusted.lov xattr + TAR=$(find_lustre_tar) + [ -z "$TAR" ] && skip "lustre-aware tar is not installed" && return + [ "$OSTCOUNT" -lt "2" ] && skip "skipping N-stripe test" && return setup_test102 mkdir -p $DIR/d102d - star -x f=$TMP/f102.tar -C $DIR/d102d + $TAR xf $TMP/f102.tar -C $DIR/d102d --xattrs cd $DIR/d102d/$tdir compare_stripe_info1 - } -run_test 102d "star restore stripe info from tarfile,not keep osts ===========" - -test_102e() { - # b10930: star test for trusted.lov xattr - star --xhelp 2>&1 | grep -q nolustre - [ $? -ne 0 ] && skip "lustre-aware star is not installed" && return - [ "$OSTCOUNT" -lt "4" ] && skip "skipping 4-stripe test" && return - setup_test102 - mkdir -p $DIR/d102e - star -x -preserve-osts f=$TMP/f102.tar -C $DIR/d102e - cd $DIR/d102e/$tdir - compare_stripe_info2 -} -run_test 102e "star restore stripe info from tarfile, keep osts ===========" +run_test 102d "tar restore stripe info from tarfile,not keep osts ===========" test_102f() { - # b10930: star test for trusted.lov xattr - star --xhelp 2>&1 | grep -q nolustre - [ $? -ne 0 ] && skip "lustre-aware star is not installed" && return - [ "$OSTCOUNT" -lt "4" ] && skip "skipping 4-stripe test" && return + # b10930: tar test for trusted.lov xattr + TAR=$(find_lustre_tar) + [ -z "$TAR" ] && skip "lustre-aware tar is not installed" && return + [ "$OSTCOUNT" -lt "2" ] && skip "skipping N-stripe test" && return setup_test102 mkdir -p $DIR/d102f cd $DIR - star -copy $tdir $DIR/d102f + $TAR cf - --xattrs $tdir | $TAR xf - --xattrs -C $DIR/d102f cd $DIR/d102f/$tdir compare_stripe_info1 } -run_test 102f "star copy files, not keep osts ===========" - -test_102g() { - # b10930: star test for trusted.lov xattr - star --xhelp 2>&1 | grep -q nolustre - [ $? -ne 0 ] && skip "lustre-aware star is not installed" && return - [ "$OSTCOUNT" -lt "4" ] && skip "skipping 4-stripe test" && return - setup_test102 - mkdir -p $DIR/d102g - cd $DIR - star -copy -preserve-osts $tdir $DIR/d102g - cd $DIR/d102g/$tdir - compare_stripe_info2 - cleanup_test102 -} -run_test 102g "star copy files, keep osts ===========" +run_test 102f "tar copy files, not keep osts ===========" test_102h() { # bug 15777 [ -z $(lctl get_param -n mdc.*.connect_flags | grep xattr) ] && @@ -4132,6 +4263,7 @@ test_102h() { # bug 15777 error "$XBIG different after growing $XSML" fi log "$XBIG still valid after growing $XSML" + rm -f $file } run_test 102h "grow xattr from inside inode to external block" @@ -4144,6 +4276,21 @@ test_102i() { # bug 17038 } run_test 102i "lgetxattr test on symbolic link ============" +test_102j() { + TAR=$(find_lustre_tar) + [ -z "$TAR" ] && skip "lustre-aware tar is not installed" && return + [ "$OSTCOUNT" -lt "2" ] && skip "skipping N-stripe test" && return + setup_test102 "$RUNAS" + mkdir -p $DIR/d102j + chown $RUNAS_ID $DIR/d102j + $RUNAS $TAR xf $TMP/f102.tar -C $DIR/d102j --xattrs + cd $DIR/d102j/$tdir + compare_stripe_info1 "$RUNAS" +} +run_test 102j "non-root tar restore stripe info from tarfile, not keep osts ===" + +cleanup_test102 + run_acl_subtest() { $LUSTRE/tests/acl/run $LUSTRE/tests/acl/$1.test @@ -4169,19 +4316,19 @@ test_103 () { echo "performing cp ..." run_acl_subtest cp || error echo "performing getfacl-noacl..." - run_acl_subtest getfacl-noacl || error + run_acl_subtest getfacl-noacl || error "getfacl-noacl test failed" echo "performing misc..." - run_acl_subtest misc || error + run_acl_subtest misc || error "misc test failed" echo "performing permissions..." - run_acl_subtest permissions || error + run_acl_subtest permissions || error "permissions failed" echo "performing setfacl..." - run_acl_subtest setfacl || error + run_acl_subtest setfacl || error "setfacl test failed" # inheritance test got from HP echo "performing inheritance..." - cp $LUSTRE/tests/acl/make-tree . || error - chmod +x make-tree || error - run_acl_subtest inheritance || error + cp $LUSTRE/tests/acl/make-tree . || error "cannot copy make-tree" + chmod +x make-tree || error "chmod +x failed" + run_acl_subtest inheritance || error "inheritance test failed" rm -f make-tree cd $SAVE_PWD @@ -4203,12 +4350,13 @@ test_104() { lfs df -i $DIR || error "lfs df -i $DIR failed" lfs df $DIR/$tfile || error "lfs df $DIR/$tfile failed" lfs df -ih $DIR/$tfile || error "lfs df -ih $DIR/$tfile failed" - + OSC=`lctl get_param -n devices | awk '/-osc-/ {print $4}' | head -n 1` lctl --device %$OSC deactivate lfs df || error "lfs df with deactivated OSC failed" lctl --device %$OSC recover lfs df || error "lfs df with reactivated OSC failed" + rm -f $DIR/$tfile } run_test 104 "lfs df [-ih] [path] test =========================" @@ -4221,6 +4369,7 @@ test_105a() { else flocks_test 1 off -f $DIR/$tfile || error "fail flock off" fi + rm -f $DIR/$tfile } run_test 105a "flock when mounted without -o flock test ========" @@ -4232,6 +4381,7 @@ test_105b() { else flocks_test 1 off -c $DIR/$tfile || error "fail flock off" fi + rm -f $DIR/$tfile } run_test 105b "fcntl when mounted without -o flock test ========" @@ -4243,6 +4393,7 @@ test_105c() { else flocks_test 1 off -l $DIR/$tfile || error "fail flock off" fi + rm -f $DIR/$tfile } run_test 105c "lockf when mounted without -o flock test ========" @@ -4304,6 +4455,7 @@ test_110() { touch $DIR/d110/yyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyy && error ""create with 256 char should fail, but not ls -l $DIR/d110 + rm -fr $DIR/d110 } run_test 110 "filename length checking" @@ -4370,7 +4522,7 @@ test_116() { declare -i FILL FILL=$(($MINV / 4)) echo "Filling 25% remaining space in OST${MINI} with ${FILL}Kb" - $SETSTRIPE $DIR/$tdir/OST${MINI} -i $MINI -c 1 + $SETSTRIPE -i $MINI -c 1 $DIR/$tdir/OST${MINI}||error "setstripe failed" i=0 while [ $FILL -gt 0 ]; do i=$(($i + 1)) @@ -4435,6 +4587,8 @@ test_116() { echo "$MAXC files created on larger OST $MAXI1" [ $MINC -gt 0 ] && echo "Wrote $(($MAXC * 100 / $MINC - 100))% more files to larger OST $MAXI1" [ $MAXC -gt $MINC ] || error_ignore "stripe QOS didn't balance free space" + + rm -rf $DIR/$tdir/OST${MINI} } run_test 116 "stripe QOS: free space balance ===================" @@ -4446,6 +4600,7 @@ test_117() # bug 10891 > $DIR/$tfile || error "truncate failed" lctl set_param fail_loc=0 echo "Truncate succeeded." + rm -f $DIR/$tfile } run_test 117 "verify fsfilt_extend ==========" @@ -4464,7 +4619,7 @@ reset_async() { FILE=$DIR/reset_async # Ensure all OSCs are cleared - $LSTRIPE $FILE 0 -1 -1 + $LSTRIPE -c -1 $FILE dd if=/dev/zero of=$FILE bs=64k count=$OSTCOUNT sync rm $FILE @@ -4473,7 +4628,7 @@ reset_async() { test_118a() #bug 11710 { reset_async - + multiop $DIR/$tfile oO_CREAT:O_RDWR:O_SYNC:w4096c DIRTY=$(lctl get_param -n llite.*.dump_page_cache | grep -c dirty) WRITEBACK=$(lctl get_param -n llite.*.dump_page_cache | grep -c writeback) @@ -4482,6 +4637,7 @@ test_118a() #bug 11710 error "Dirty pages not flushed to disk, dirty=$DIRTY, writeback=$WRITEBACK" return 1; fi + rm -f $DIR/$tfile } run_test 118a "verify O_SYNC works ==========" @@ -4516,7 +4672,7 @@ test_118b() # until a subsequent RPC completes successfully without error. multiop $DIR/$tfile Ow4096yc rm -f $DIR/$tfile - + return 0 } run_test 118b "Reclaim dirty pages on fatal error ==========" @@ -4558,7 +4714,7 @@ test_118c() if [[ $DIRTY -ne 0 || $WRITEBACK -ne 0 ]]; then error "Dirty pages not flushed to disk, dirty=$DIRTY, writeback=$WRITEBACK" fi - + rm -f $DIR/$tfile echo "Dirty pages flushed via fsync on EROFS" return 0 @@ -4574,7 +4730,7 @@ test_118d() #define OBD_FAIL_OST_BRW_PAUSE_BULK set_nodes_failloc "$(osts_nodes)" 0x214 # multiop should block due to fsync until pages are written - multiop $DIR/$tfile oO_CREAT:O_RDWR:O_SYNC:w4096c & + multiop $DIR/$tfile oO_CREAT:O_RDWR:O_SYNC:w4096c & MULTIPID=$! sleep 1 @@ -4616,7 +4772,7 @@ test_118f() { if [[ $RC -eq 0 ]]; then error "Must return error due to dropped pages, rc=$RC" fi - + lctl set_param fail_loc=0x0 LOCKED=$(lctl get_param -n llite.*.dump_page_cache | grep -c locked) @@ -4648,7 +4804,7 @@ test_118g() { # simulate local -ENOMEM multiop $DIR/$tfile oO_CREAT:O_RDWR:O_SYNC:w4096c RC=$? - + lctl set_param fail_loc=0 if [[ $RC -eq 0 ]]; then error "Must return error due to dropped pages, rc=$RC" @@ -4661,7 +4817,7 @@ test_118g() { if [[ $LOCKED -ne 0 ]]; then error "Locked pages remain in cache, locked=$LOCKED" fi - + if [[ $DIRTY -ne 0 || $WRITEBACK -ne 0 ]]; then error "Dirty pages not flushed to disk, dirty=$DIRTY, writeback=$WRITEBACK" fi @@ -4684,7 +4840,7 @@ test_118h() { # Should simulate ENOMEM error which is recoverable and should be handled by timeout multiop $DIR/$tfile oO_CREAT:O_RDWR:O_SYNC:w4096c RC=$? - + set_nodes_failloc "$(osts_nodes)" 0 if [[ $RC -eq 0 ]]; then error "Must return error due to dropped pages, rc=$RC" @@ -4697,7 +4853,7 @@ test_118h() { if [[ $LOCKED -ne 0 ]]; then error "Locked pages remain in cache, locked=$LOCKED" fi - + if [[ $DIRTY -ne 0 || $WRITEBACK -ne 0 ]]; then error "Dirty pages not flushed to disk, dirty=$DIRTY, writeback=$WRITEBACK" fi @@ -4716,13 +4872,13 @@ test_118i() { #define OBD_FAIL_OST_BRW_WRITE_BULK 0x20e set_nodes_failloc "$(osts_nodes)" 0x20e - + # Should simulate ENOMEM error which is recoverable and should be handled by timeout multiop $DIR/$tfile oO_CREAT:O_RDWR:O_SYNC:w4096c & PID=$! sleep 5 set_nodes_failloc "$(osts_nodes)" 0 - + wait $PID RC=$? if [[ $RC -ne 0 ]]; then @@ -4735,7 +4891,7 @@ test_118i() { if [[ $LOCKED -ne 0 ]]; then error "Locked pages remain in cache, locked=$LOCKED" fi - + if [[ $DIRTY -ne 0 || $WRITEBACK -ne 0 ]]; then error "Dirty pages not flushed to disk, dirty=$DIRTY, writeback=$WRITEBACK" fi @@ -4769,7 +4925,7 @@ test_118j() { if [[ $LOCKED -ne 0 ]]; then error "Locked pages remain in cache, locked=$LOCKED" fi - + # in recoverable error on OST we want resend and stay until it finished if [[ $DIRTY -ne 0 || $WRITEBACK -ne 0 ]]; then error "Dirty pages not flushed to disk, dirty=$DIRTY, writeback=$WRITEBACK" @@ -4825,7 +4981,7 @@ test_119b() # bug 11737 { [ "$OSTCOUNT" -lt "2" ] && skip "skipping 2-stripe test" && return - $SETSTRIPE $DIR/$tfile -c 2 + $SETSTRIPE -c 2 $DIR/$tfile || error "setstripe failed" dd if=/dev/zero of=$DIR/$tfile bs=1M count=1 seek=1 || error "dd failed" sync multiop $DIR/$tfile oO_RDONLY:O_DIRECT:r$((2048 * 1024)) || \ @@ -5051,11 +5207,11 @@ run_test 120g "Early Lock Cancel: performance test" test_121() { #bug #10589 rm -rf $DIR/$tfile - writes=$(LANG=C dd if=/dev/zero of=$DIR/$tfile count=1 2>&1 | awk -F '+' '/out/ {print $1}') + writes=$(LANG=C dd if=/dev/zero of=$DIR/$tfile count=1 2>&1 | awk -F '+' '/out$/ {print $1}') #define OBD_FAIL_LDLM_CANCEL_RACE 0x310 lctl set_param fail_loc=0x310 cancel_lru_locks osc > /dev/null - reads=$(LANG=C dd if=$DIR/$tfile of=/dev/null 2>&1 | awk -F '+' '/in/ {print $1}') + reads=$(LANG=C dd if=$DIR/$tfile of=/dev/null 2>&1 | awk -F '+' '/in$/ {print $1}') lctl set_param fail_loc=0 [ "$reads" -eq "$writes" ] || error "read" $reads "blocks, must be" $writes } @@ -5068,45 +5224,83 @@ test_123a() { # was test 123, statahead(bug 11401) SLOWOK=1 fi - remount_client $MOUNT + rm -rf $DIR/$tdir mkdir -p $DIR/$tdir - error=0 NUMFREE=`df -i -P $DIR | tail -n 1 | awk '{ print $4 }'` [ $NUMFREE -gt 100000 ] && NUMFREE=100000 || NUMFREE=$((NUMFREE-1000)) MULT=10 - for ((i=1, j=0; i<=$NUMFREE; j=$i, i=$((i * MULT)) )); do + for ((i=100, j=0; i<=$NUMFREE; j=$i, i=$((i * MULT)) )); do createmany -o $DIR/$tdir/$tfile $j $((i - j)) - lctl get_param -n llite.*.statahead_max | grep '[0-9]' - cancel_lru_locks mdc - cancel_lru_locks osc - stime=`date +%s` - ls -l $DIR/$tdir > /dev/null - etime=`date +%s` - delta_sa=$((etime - stime)) - log "ls $i files with statahead: $delta_sa sec" - lctl get_param -n llite.*.statahead_stats - max=`lctl get_param -n llite.*.statahead_max | head -n 1` lctl set_param -n llite.*.statahead_max 0 lctl get_param llite.*.statahead_max cancel_lru_locks mdc cancel_lru_locks osc stime=`date +%s` - ls -l $DIR/$tdir > /dev/null + time ls -l $DIR/$tdir | wc -l etime=`date +%s` delta=$((etime - stime)) log "ls $i files without statahead: $delta sec" - lctl set_param llite.*.statahead_max=$max - if [ $delta_sa -gt $(($delta + 2)) ]; then - log "ls $i files is slower with statahead!" - error=1 + + swrong=`lctl get_param -n llite.*.statahead_stats | grep "statahead wrong:" | awk '{print $3}'` + lctl get_param -n llite.*.statahead_max | grep '[0-9]' + cancel_lru_locks mdc + cancel_lru_locks osc + stime=`date +%s` + time ls -l $DIR/$tdir | wc -l + etime=`date +%s` + delta_sa=$((etime - stime)) + log "ls $i files with statahead: $delta_sa sec" + lctl get_param -n llite.*.statahead_stats + ewrong=`lctl get_param -n llite.*.statahead_stats | grep "statahead wrong:" | awk '{print $3}'` + + if [ $swrong -lt $ewrong ]; then + log "statahead was stopped, maybe too many locks held!" + fi + + [ $delta -eq 0 -o $delta_sa -eq 0 ] && continue + + if [ $((delta_sa * 100)) -gt $((delta * 105)) -a $delta_sa -gt $((delta + 2)) ]; then + if [ $SLOWOK -eq 0 ]; then + error "ls $i files is slower with statahead!" + debugsave + + lctl set_param debug=-1 + max=`lctl get_param -n llite.*.statahead_max | head -n 1` + lctl set_param -n llite.*.statahead_max 0 + lctl get_param llite.*.statahead_max + cancel_lru_locks mdc + cancel_lru_locks osc + $LCTL clear + stime=`date +%s` + time ls -l $DIR/$tdir | wc -l + etime=`date +%s` + error "ls $i files (again) without statahead: $((etime - stime)) sec" + + lctl set_param debug=-1 + lctl set_param llite.*.statahead_max=$max + lctl get_param -n llite.*.statahead_max | grep '[0-9]' + cancel_lru_locks mdc + cancel_lru_locks osc + $LCTL clear + stime=`date +%s` + time ls -l $DIR/$tdir | wc -l + etime=`date +%s` + error "ls $i files (again) with statahead: $((etime - stime)) sec" + lctl get_param -n llite.*.statahead_stats + + debugrestore + else + log "ls $i files is slower with statahead!" + fi + break fi [ $delta -gt 20 ] && break [ $delta -gt 8 ] && MULT=$((50 / delta)) - [ "$SLOW" = "no" -a $delta -gt 3 ] && break + [ "$SLOW" = "no" -a $delta -gt 5 ] && break done log "ls done" @@ -5118,17 +5312,13 @@ test_123a() { # was test 123, statahead(bug 11401) log "rm -r $DIR/$tdir/: $delta seconds" log "rm done" lctl get_param -n llite.*.statahead_stats - # wait for commitment of removal - sleep 2 - [ $error -ne 0 -a $SLOWOK -eq 0 ] && error "statahead is slow!" - return 0 } run_test 123a "verify statahead work" test_123b () { # statahead(bug 15027) mkdir -p $DIR/$tdir createmany -o $DIR/$tdir/$tfile-%d 1000 - + cancel_lru_locks mdc cancel_lru_locks osc @@ -5147,10 +5337,9 @@ run_test 123b "not panic with network error in statahead enqueue (bug 15027)" test_124a() { [ -z "`lctl get_param -n mdc.*.connect_flags | grep lru_resize`" ] && \ skip "no lru resize on server" && return 0 - NR=2000 + local NR=2000 mkdir -p $DIR/$tdir || error "failed to create $DIR/$tdir" - # use touch to produce $NR new locks log "create $NR files at $DIR/$tdir" createmany -o $DIR/$tdir/f $NR || error "failed to create $NR files in $DIR/$tdir" @@ -5158,14 +5347,14 @@ test_124a() { cancel_lru_locks mdc ls -l $DIR/$tdir > /dev/null - NSDIR="" - LRU_SIZE=0 + local NSDIR="" + local LRU_SIZE=0 for VALUE in `lctl get_param ldlm.namespaces.*mdc-*.lru_size`; do - PARAM=`echo ${VALUE[0]} | cut -d "=" -f1` + local PARAM=`echo ${VALUE[0]} | cut -d "=" -f1` LRU_SIZE=$(lctl get_param -n $PARAM) if [ $LRU_SIZE -gt $(default_lru_size) ]; then NSDIR=$(echo $PARAM | cut -d "." -f1-3) - log "using $(basename $NSDIR) namespace" + log "NS=$(basename $NSDIR)" break fi done @@ -5174,40 +5363,53 @@ test_124a() { skip "Not enough cached locks created!" return 0 fi - log "created $LRU_SIZE lock(s)" - - # we want to sleep 30s to not make test too long - SLEEP=30 - SLEEP_ADD=2 - - # we know that lru resize allows one client to hold $LIMIT locks for 10h - MAX_HRS=10 - - # get the pool limit - LIMIT=`lctl get_param -n $NSDIR.pool.limit` - - # calculate lock volume factor taking into account data set size and the - # rule that number of locks will be getting smaller durring sleep interval - # and we need to additionally enforce LVF to take this into account. - # Use $LRU_SIZE_B here to take into account real number of locks created - # in the case of CMD, LRU_SIZE_B != $NR in most of cases - LVF=$(($MAX_HRS * 60 * 60 * $LIMIT / $SLEEP)) - LRU_SIZE_B=$LRU_SIZE - log "make client drop locks $LVF times faster so that ${SLEEP}s is enough to cancel $LRU_SIZE lock(s)" - OLD_LVF=`lctl get_param -n $NSDIR.pool.lock_volume_factor` + log "LRU=$LRU_SIZE" + + local SLEEP=30 + + # We know that lru resize allows one client to hold $LIMIT locks + # for 10h. After that locks begin to be killed by client. + local MAX_HRS=10 + local LIMIT=`lctl get_param -n $NSDIR.pool.limit` + + # Make LVF so higher that sleeping for $SLEEP is enough to _start_ + # killing locks. Some time was spent for creating locks. This means + # that up to the moment of sleep finish we must have killed some of + # them (10-100 locks). This depends on how fast ther were created. + # Many of them were touched in almost the same moment and thus will + # be killed in groups. + local LVF=$(($MAX_HRS * 60 * 60 / $SLEEP * $LIMIT / $LRU_SIZE)) + + # Use $LRU_SIZE_B here to take into account real number of locks + # created in the case of CMD, LRU_SIZE_B != $NR in most of cases + local LRU_SIZE_B=$LRU_SIZE + log "LVF=$LVF" + local OLD_LVF=`lctl get_param -n $NSDIR.pool.lock_volume_factor` lctl set_param -n $NSDIR.pool.lock_volume_factor $LVF - log "sleep for $((SLEEP+SLEEP_ADD))s" - sleep $((SLEEP+SLEEP_ADD)) + + # Let's make sure that we really have some margin. Client checks + # cached locks every 10 sec. + SLEEP=$((SLEEP+20)) + log "Sleep ${SLEEP} sec" + local SEC=0 + while ((SEC<$SLEEP)); do + echo -n "..." + sleep 5 + SEC=$((SEC+5)) + LRU_SIZE=`lctl get_param -n $NSDIR/lru_size` + echo -n "$LRU_SIZE" + done + echo "" lctl set_param -n $NSDIR.pool.lock_volume_factor $OLD_LVF - LRU_SIZE_A=`lctl get_param -n $NSDIR.lru_size` + local LRU_SIZE_A=`lctl get_param -n $NSDIR/lru_size` [ $LRU_SIZE_B -gt $LRU_SIZE_A ] || { - error "No locks dropped in "$((SLEEP+SLEEP_ADD))"s. LRU size: $LRU_SIZE_A" + error "No locks dropped in ${SLEEP}s. LRU size: $LRU_SIZE_A" unlinkmany $DIR/$tdir/f $NR return } - log "Dropped "$((LRU_SIZE_B-LRU_SIZE_A))" locks in "$((SLEEP+SLEEP_ADD))"s" + log "Dropped "$((LRU_SIZE_B-LRU_SIZE_A))" locks in ${SLEEP}s" log "unlink $NR files at $DIR/$tdir" unlinkmany $DIR/$tdir/f $NR } @@ -5307,6 +5509,8 @@ run_test 125 "don't return EPROTO when a dir has a non-default striping and ACLs test_126() { # bug 12829/13455 [ -z "$(lctl get_param -n llite.*.client_type | grep local)" ] && skip "must run as local client" && return [ "$UID" != 0 ] && echo "skipping $TESTNAME (must run as root)" && return + $GSS && skip "must run as gss disabled" && return + $RUNAS -u 0 -g 1 touch $DIR/$tfile || error "touch failed" gid=`ls -n $DIR/$tfile | awk '{print $4}'` rm -f $DIR/$tfile @@ -5315,7 +5519,7 @@ test_126() { # bug 12829/13455 run_test 126 "check that the fsgid provided by the client is taken into account" test_127() { # bug 15521 - $LSTRIPE -i 0 -c 1 $DIR/$tfile + $SETSTRIPE -i 0 -c 1 $DIR/$tfile || error "setstripe failed" $LCTL set_param osc.*.stats=0 FSIZE=$((2048 * 1024)) dd if=/dev/zero of=$DIR/$tfile bs=$FSIZE count=1 @@ -5327,7 +5531,7 @@ test_127() { # bug 15521 echo "got $COUNT $NAME" [ ! $MIN ] && error "Missing min value for $NAME proc entry" eval $NAME=$COUNT || error "Wrong proc format" - + case $NAME in read_bytes|write_bytes) [ $MIN -lt 4096 ] && error "min is too small: $MIN" @@ -5412,10 +5616,18 @@ test_129() { } run_test 129 "test directory size limit ========================" +OLDIFS="$IFS" +cleanup_130() { + trap 0 + IFS="$OLDIFS" +} + test_130a() { filefrag_op=$(filefrag -e 2>&1 | grep "invalid option") [ -n "$filefrag_op" ] && skip "filefrag does not support FIEMAP" && return + trap cleanup_130 EXIT RETURN + local fm_file=$DIR/$tfile lfs setstripe -s 65536 -c 1 $fm_file || error "setstripe failed on $fm_file" dd if=/dev/zero of=$fm_file bs=65536 count=1 || error "dd failed for $fm_file" @@ -5433,6 +5645,7 @@ test_130a() { frag_lun=`echo $line | cut -d: -f5` ext_len=`echo $line | cut -d: -f4` if (( $frag_lun != $lun )); then + cleanup_130 error "FIEMAP on 1-stripe file($fm_file) failed" return fi @@ -5440,9 +5653,13 @@ test_130a() { done if (( lun != frag_lun || start_blk != 0 || tot_len != 64 )); then + cleanup_130 error "FIEMAP on 1-stripe file($fm_file) failed;" return fi + + cleanup_130 + echo "FIEMAP on single striped file succeeded" } run_test 130a "FIEMAP (1-stripe file)" @@ -5453,6 +5670,8 @@ test_130b() { filefrag_op=$(filefrag -e 2>&1 | grep "invalid option") [ -n "$filefrag_op" ] && skip "filefrag does not support FIEMAP" && return + trap cleanup_130 EXIT RETURN + local fm_file=$DIR/$tfile lfs setstripe -s 65536 -c 2 $fm_file || error "setstripe failed on $fm_file" dd if=/dev/zero of=$fm_file bs=1M count=2 || error "dd failed on $fm_file" @@ -5471,6 +5690,7 @@ test_130b() { ext_len=`echo $line | cut -d: -f4` if (( $frag_lun != $last_lun )); then if (( tot_len != 1024 )); then + cleanup_130 error "FIEMAP on $fm_file failed; returned len $tot_len for OST $last_lun instead of 256" return else @@ -5482,10 +5702,13 @@ test_130b() { last_lun=$frag_lun done if (( num_luns != 2 || tot_len != 1024 )); then + cleanup_130 error "FIEMAP on $fm_file failed; returned wrong number of luns or wrong len for OST $last_lun" return fi + cleanup_130 + echo "FIEMAP on 2-stripe file succeeded" } run_test 130b "FIEMAP (2-stripe file)" @@ -5496,6 +5719,8 @@ test_130c() { filefrag_op=$(filefrag -e 2>&1 | grep "invalid option") [ -n "$filefrag_op" ] && skip "filefrag does not support FIEMAP" && return + trap cleanup_130 EXIT RETURN + local fm_file=$DIR/$tfile lfs setstripe -s 65536 -c 2 $fm_file || error "setstripe failed on $fm_file" dd if=/dev/zero of=$fm_file seek=1 bs=1M count=1 || error "dd failed on $fm_file" @@ -5515,10 +5740,12 @@ test_130c() { if (( $frag_lun != $last_lun )); then logical=`echo $line | cut -d: -f2 | cut -d. -f1` if (( logical != 512 )); then + cleanup_130 error "FIEMAP on $fm_file failed; returned logical start for lun $logical instead of 512" return fi if (( tot_len != 512 )); then + cleanup_130 error "FIEMAP on $fm_file failed; returned len $tot_len for OST $last_lun instead of 1024" return else @@ -5530,10 +5757,13 @@ test_130c() { last_lun=$frag_lun done if (( num_luns != 2 || tot_len != 512 )); then + cleanup_130 error "FIEMAP on $fm_file failed; returned wrong number of luns or wrong len for OST $last_lun" return fi + cleanup_130 + echo "FIEMAP on 2-stripe file with hole succeeded" } run_test 130c "FIEMAP (2-stripe file with hole)" @@ -5544,6 +5774,8 @@ test_130d() { filefrag_op=$(filefrag -e 2>&1 | grep "invalid option") [ -n "$filefrag_op" ] && skip "filefrag does not support FIEMAP" && return + trap cleanup_130 EXIT RETURN + local fm_file=$DIR/$tfile lfs setstripe -s 65536 -c $OSTCOUNT $fm_file || error "setstripe failed on $fm_file" dd if=/dev/zero of=$fm_file bs=1M count=$OSTCOUNT || error "dd failed on $fm_file" @@ -5562,6 +5794,7 @@ test_130d() { ext_len=`echo $line | cut -d: -f4` if (( $frag_lun != $last_lun )); then if (( tot_len != 1024 )); then + cleanup_130 error "FIEMAP on $fm_file failed; returned len $tot_len for OST $last_lun instead of 1024" return else @@ -5573,10 +5806,13 @@ test_130d() { last_lun=$frag_lun done if (( num_luns != OSTCOUNT || tot_len != 1024 )); then + cleanup_130 error "FIEMAP on $fm_file failed; returned wrong number of luns or wrong len for OST $last_lun" return fi + cleanup_130 + echo "FIEMAP on N-stripe file succeeded" } run_test 130d "FIEMAP (N-stripe file)" @@ -5587,6 +5823,8 @@ test_130e() { filefrag_op=$(filefrag -e 2>&1 | grep "invalid option") [ -n "$filefrag_op" ] && skip "filefrag does not support FIEMAP" && return + trap cleanup_130 EXIT RETURN + local fm_file=$DIR/$tfile lfs setstripe -s 65536 -c 2 $fm_file || error "setstripe failed on $fm_file" NUM_BLKS=512 @@ -5610,6 +5848,7 @@ test_130e() { ext_len=`echo $line | cut -d: -f4` if (( $frag_lun != $last_lun )); then if (( tot_len != $EXPECTED_LEN )); then + cleanup_130 error "FIEMAP on $fm_file failed; returned len $tot_len for OST $last_lun instead of $EXPECTED_LEN" return else @@ -5621,15 +5860,125 @@ test_130e() { last_lun=$frag_lun done if (( num_luns != 2 || tot_len != $EXPECTED_LEN )); then - echo "$num_luns $tot_len" + cleanup_130 error "FIEMAP on $fm_file failed; returned wrong number of luns or wrong len for OST $last_lun" return fi + cleanup_130 + echo "FIEMAP with continuation calls succeeded" } run_test 130e "FIEMAP (test continuation FIEMAP calls)" +# Test for writev/readv +test_131a() { + rwv -f $DIR/$tfile -w -n 3 524288 1048576 1572864 || \ + error "writev test failed" + rwv -f $DIR/$tfile -r -v -n 2 1572864 1048576 || \ + error "readv failed" + rm -f $DIR/$tfile +} +run_test 131a "test iov's crossing stripe boundary for writev/readv" + +test_131b() { + rwv -f $DIR/$tfile -w -a -n 3 524288 1048576 1572864 || \ + error "append writev test failed" + rwv -f $DIR/$tfile -w -a -n 2 1572864 1048576 || \ + error "append writev test failed" + rm -f $DIR/$tfile +} +run_test 131b "test append writev" + +test_131c() { + rwv -f $DIR/$tfile -w -d -n 1 1048576 || return 0 + error "NOT PASS" +} +run_test 131c "test read/write on file w/o objects" + +test_131d() { + rwv -f $DIR/$tfile -w -n 1 1572864 + NOB=`rwv -f $DIR/$tfile -r -n 3 524288 524288 1048576 | awk '/error/ {print $6}'` + if [ "$NOB" != 1572864 ]; then + error "Short read filed: read $NOB bytes instead of 1572864" + fi + rm -f $DIR/$tfile +} +run_test 131d "test short read" + +test_131e() { + rwv -f $DIR/$tfile -w -s 1048576 -n 1 1048576 + rwv -f $DIR/$tfile -r -z -s 0 -n 1 524288 || \ + error "read hitting hole failed" + rm -f $DIR/$tfile +} +run_test 131e "test read hitting hole" + +get_ost_param() { + local token=$1 + local gl_sum=0 + for node in $(osts_nodes); do + gl=$(do_node $node "$LCTL get_param -n ost.OSS.ost.stats" | awk '/'$token'/ {print $2}' | head -n 1) + [ x$gl = x"" ] && gl=0 + gl_sum=$((gl_sum + gl)) + done + echo $gl +} + +som_mode_switch() { + local som=$1 + local gl1=$2 + local gl2=$3 + + if [ x$som = x"enabled" ]; then + [ $((gl2 - gl1)) -gt 0 ] && error "no glimpse RPC is expected" + do_facet mgs "$LCTL conf_param $FSNAME.mdt.som=disabled" + else + [ $((gl2 - gl1)) -gt 0 ] || error "some glimpse RPC is expected" + do_facet mgs "$LCTL conf_param $FSNAME.mdt.som=enabled" + fi + + # do remount to make new mount-conf parameters actual + echo remounting... + sync + stopall + setupall +} + +test_132() { #1028, SOM + local num=$(get_mds_dir $DIR) + local mymds=mds${num} + + dd if=/dev/zero of=$DIR/$tfile count=1 2>/dev/null + cancel_lru_locks osc + + som1=$(do_facet $mymds "$LCTL get_param mdt.*.som" | awk -F= ' {print $2}' | head -n 1) + + gl1=$(get_ost_param "ldlm_glimpse_enqueue") + stat $DIR/$tfile >/dev/null + gl2=$(get_ost_param "ldlm_glimpse_enqueue") + echo "====> SOM is "$som1", "$((gl2 - gl1))" glimpse RPC occured" + cancel_lru_locks osc + som_mode_switch $som1 $gl1 $gl2 + + som2=$(do_facet $mymds "$LCTL get_param mdt.*.som" | awk -F= ' {print $2}' | head -n 1) + if [ $som1 == $som2 ]; then + error "som is still "$som2 + if [ x$som2 = x"enabled" ]; then + som2="disabled" + else + som2="enabled" + fi + fi + + gl1=$(get_ost_param "ldlm_glimpse_enqueue") + stat $DIR/$tfile >/dev/null + gl2=$(get_ost_param "ldlm_glimpse_enqueue") + echo "SOM is "$som2", "$((gl2 - gl1))" glimpse RPC occured" + som_mode_switch $som2 $gl1 $gl2 +} +run_test 132 "som avoids glimpse rpc" + test_140() { #bug-17379 mkdir -p $DIR/$tdir || error "Creating dir $DIR/$tdir" cd $DIR/$tdir || error "Changing to $DIR/$tdir" @@ -5655,8 +6004,8 @@ test_140() { #bug-17379 } done i=`expr $i - 1` - [ $i -eq 5 -o $i -eq 8 ] || error "Invalid symlink depth" echo "The symlink depth = $i" + [ $i -eq 4 -o $i -eq 8 ] || error "Invalid symlink depth" } run_test 140 "Check reasonable stack depth (shouldn't LBUG) ====" @@ -5691,61 +6040,66 @@ test_150() { run_test 150 "truncate/append tests" function roc_access() { - ACCNUM=`$LCTL get_param -n obdfilter.*.stats | \ - grep 'cache_access'| awk '{print $2}' | \ - awk '{sum=sum+$3} END{print sum}'` - echo $ACCNUM + local list=$(comma_list $(osts_nodes)) + ACCNUM=`do_nodes $list $LCTL get_param -n obdfilter.*.stats | \ + grep 'cache_access'| awk '{print $2}' | \ + awk '{sum=sum+$3} END{print sum}'` + echo $ACCNUM } function roc_hit() { - ACCNUM=`$LCTL get_param -n obdfilter.*.stats | \ - grep 'cache_hit'|awk '{print $2}' | \ - awk '{sum=sum+$1} END{print sum}'` - echo $ACCNUM + local list=$(comma_list $(osts_nodes)) + ACCNUM=`do_nodes $list $LCTL get_param -n obdfilter.*.stats | \ + grep 'cache_hit'|awk '{print $2}' | \ + awk '{sum=sum+$1} END{print sum}'` + echo $ACCNUM } test_151() { - local CPAGES=3 + remote_ost_nodsh && skip "remote OST with nodsh" && return - # check whether obdfilter is cache capable at all - if ! $LCTL get_param -n obdfilter.*.read_cache_enable; then - echo "not cache-capable obdfilter" - return 0 - fi + local CPAGES=3 + local list=$(comma_list $(osts_nodes)) - # check cache is enabled on all obdfilters - if $LCTL get_param -n obdfilter.*.read_cache_enable | grep 0 >&/dev/null; then - echo "oss cache is disabled" - return 0 - fi + # check whether obdfilter is cache capable at all + if ! do_nodes $list $LCTL get_param -n obdfilter.*.read_cache_enable > /dev/null; then + echo "not cache-capable obdfilter" + return 0 + fi - $LCTL set_param -n obdfilter.*.writethrough_cache_enable 1 + # check cache is enabled on all obdfilters + if do_nodes $list $LCTL get_param -n obdfilter.*.read_cache_enable | grep 0 >&/dev/null; then + echo "oss cache is disabled" + return 0 + fi - # pages should be in the case right after write + do_nodes $list $LCTL set_param -n obdfilter.*.writethrough_cache_enable 1 + + # pages should be in the case right after write dd if=/dev/urandom of=$DIR/$tfile bs=4k count=$CPAGES || error "dd failed" - BEFORE=`roc_hit` + local BEFORE=`roc_hit` cancel_lru_locks osc - cat $DIR/$tfile >/dev/null - AFTER=`roc_hit` - if ! let "AFTER - BEFORE == CPAGES"; then - error "NOT IN CACHE: before: $BEFORE, after: $AFTER" - fi + cat $DIR/$tfile >/dev/null + local AFTER=`roc_hit` + if ! let "AFTER - BEFORE == CPAGES"; then + error "NOT IN CACHE: before: $BEFORE, after: $AFTER" + fi - # the following read invalidates the cache + # the following read invalidates the cache cancel_lru_locks osc - $LCTL set_param -n obdfilter.*.read_cache_enable 0 - cat $DIR/$tfile >/dev/null + do_nodes $list $LCTL set_param -n obdfilter.*.read_cache_enable 0 + cat $DIR/$tfile >/dev/null - # now data shouldn't be found in the cache - BEFORE=`roc_hit` + # now data shouldn't be found in the cache + BEFORE=`roc_hit` cancel_lru_locks osc - cat $DIR/$tfile >/dev/null - AFTER=`roc_hit` - if ! let "AFTER - BEFORE == CPAGES"; then - error "IN CACHE: before: $BEFORE, after: $AFTER" - fi + cat $DIR/$tfile >/dev/null + AFTER=`roc_hit` + if let "AFTER - BEFORE != 0"; then + error "IN CACHE: before: $BEFORE, after: $AFTER" + fi - $LCTL set_param -n obdfilter.*.read_cache_enable 1 + do_nodes $list $LCTL set_param -n obdfilter.*.read_cache_enable 1 rm -f $DIR/$tfile } run_test 151 "test cache on oss and controls ===============================" @@ -5753,14 +6107,14 @@ run_test 151 "test cache on oss and controls ===============================" test_152() { local TF="$TMP/$tfile" - # simulate ENOMEM during write -#define OBD_FAIL_OST_NOMEM 0x226 + # simulate ENOMEM during write +#define OBD_FAIL_OST_NOMEM 0x226 lctl set_param fail_loc=0x80000226 dd if=/dev/urandom of=$TF bs=6096 count=1 || error "dd failed" cp $TF $DIR/$tfile sync || error "sync failed" lctl set_param fail_loc=0 - + # discard client's cache cancel_lru_locks osc @@ -5769,10 +6123,268 @@ test_152() { cmp $TF $DIR/$tfile || error "cmp failed" lctl set_param fail_loc=0 - rm -f $TF + rm -f $TF } run_test 152 "test read/write with enomem ============================" +test_153() { + multiop $DIR/$tfile Ow4096Ycu || error "multiop failed" +} +run_test 153 "test if fdatasync does not crash =======================" + +test_154() { + cp /etc/hosts $DIR/$tfile + + fid=$($LFS path2fid $DIR/$tfile) + rc=$? + [ $rc -ne 0 ] && error "error: could not get fid for $DIR/$tfile." + + echo "open fid $fid" + diff /etc/hosts $DIR/.lustre/fid/$fid || error "open by fid failed: did not find expected data in file." + + echo "Opening a file by FID succeeded" +} +run_test 154 "Opening a file by FID" + +#Changelogs +err17935 () { + if [ $MDSCOUNT -gt 1 ]; then + error_ignore 17935 $* + else + error $* + fi +} +test_160() { + USER=$(do_facet $SINGLEMDS lctl --device $MDT0 changelog_register -n) + echo "Registered as changelog user $USER" + do_facet $SINGLEMDS lctl get_param -n mdd.$MDT0.changelog_users | \ + grep -q $USER || error "User $USER not found in changelog_users" + + # change something + mkdir -p $DIR/$tdir/pics/2008/zachy + touch $DIR/$tdir/pics/2008/zachy/timestamp + cp /etc/hosts $DIR/$tdir/pics/2008/zachy/pic1.jpg + mv $DIR/$tdir/pics/2008/zachy $DIR/$tdir/pics/zach + ln $DIR/$tdir/pics/zach/pic1.jpg $DIR/$tdir/pics/2008/portland.jpg + ln -s $DIR/$tdir/pics/2008/portland.jpg $DIR/$tdir/pics/desktop.jpg + rm $DIR/$tdir/pics/desktop.jpg + + $LFS changelog $MDT0 | tail -5 + + echo "verifying changelog mask" + do_facet $SINGLEMDS lctl set_param mdd.$MDT0.changelog_mask="-mkdir" + mkdir -p $DIR/$tdir/pics/2009/sofia + do_facet $SINGLEMDS lctl set_param mdd.$MDT0.changelog_mask="+mkdir" + mkdir $DIR/$tdir/pics/2009/zachary + DIRS=$($LFS changelog $MDT0 | tail -5 | grep -c MKDIR) + [ $DIRS -eq 1 ] || err17935 "changelog mask count $DIRS != 1" + + # verify contents + echo "verifying target fid" + fidc=$($LFS changelog $MDT0 | grep timestamp | grep "CREAT" | \ + tail -1 | awk '{print $5}') + fidf=$($LFS path2fid $DIR/$tdir/pics/zach/timestamp) + [ "$fidc" == "t=$fidf" ] || \ + err17935 "fid in changelog $fidc != file fid $fidf" + echo "verifying parent fid" + fidc=$($LFS changelog $MDT0 | grep timestamp | grep "CREAT" | \ + tail -1 | awk '{print $6}') + fidf=$($LFS path2fid $DIR/$tdir/pics/zach) + [ "$fidc" == "p=$fidf" ] || \ + err17935 "pfid in changelog $fidc != dir fid $fidf" + + USER_REC1=$(do_facet $SINGLEMDS lctl get_param -n \ + mdd.$MDT0.changelog_users | grep $USER | awk '{print $2}') + $LFS changelog_clear $MDT0 $USER $(($USER_REC1 + 5)) + USER_REC2=$(do_facet $SINGLEMDS lctl get_param -n \ + mdd.$MDT0.changelog_users | grep $USER | awk '{print $2}') + echo "verifying user clear: $(( $USER_REC1 + 5 )) == $USER_REC2" + [ $USER_REC2 == $(($USER_REC1 + 5)) ] || \ + err17935 "user index should be $(($USER_REC1 + 5)); is $USER_REC2" + + MIN_REC=$(do_facet $SINGLEMDS lctl get_param mdd.$MDT0.changelog_users | \ + awk 'min == "" || $2 < min {min = $2}; END {print min}') + FIRST_REC=$($LFS changelog $MDT0 | head -1 | awk '{print $1}') + echo "verifying min purge: $(( $MIN_REC + 1 )) == $FIRST_REC" + [ $FIRST_REC == $(($MIN_REC + 1)) ] || \ + err17935 "first index should be $(($MIN_REC + 1)); is $FIRST_REC" + + echo "verifying user deregister" + do_facet $SINGLEMDS lctl --device $MDT0 changelog_deregister $USER + do_facet $SINGLEMDS lctl get_param -n mdd.$MDT0.changelog_users | \ + grep -q $USER && error "User $USER still found in changelog_users" + + USERS=$(( $(do_facet $SINGLEMDS lctl get_param -n \ + mdd.$MDT0.changelog_users | wc -l) - 2 )) + if [ $USERS -eq 0 ]; then + LAST_REC1=$(do_facet $SINGLEMDS lctl get_param -n \ + mdd.$MDT0.changelog_users | head -1 | awk '{print $3}') + touch $DIR/$tdir/chloe + LAST_REC2=$(do_facet $SINGLEMDS lctl get_param -n \ + mdd.$MDT0.changelog_users | head -1 | awk '{print $3}') + echo "verify changelogs are off if we were the only user: $LAST_REC1 == $LAST_REC2" + [ $LAST_REC1 == $LAST_REC2 ] || error "changelogs not off" + else + echo "$USERS other changelog users; can't verify off" + fi +} +run_test 160 "changelog sanity" + +test_161() { + mkdir -p $DIR/$tdir + cp /etc/hosts $DIR/$tdir/$tfile + mkdir $DIR/$tdir/foo1 + mkdir $DIR/$tdir/foo2 + ln $DIR/$tdir/$tfile $DIR/$tdir/foo1/sofia + ln $DIR/$tdir/$tfile $DIR/$tdir/foo2/zachary + ln $DIR/$tdir/$tfile $DIR/$tdir/foo1/luna + ln $DIR/$tdir/$tfile $DIR/$tdir/foo2/thor + local FID=$($LFS path2fid $DIR/$tdir/$tfile | tr -d '[') + if [ "$($LFS fid2path $DIR $FID | wc -l)" != "5" ]; then + $LFS fid2path $DIR $FID + err17935 "bad link ea" + fi + # middle + rm $DIR/$tdir/foo2/zachary + # last + rm $DIR/$tdir/foo2/thor + # first + rm $DIR/$tdir/$tfile + # rename + mv $DIR/$tdir/foo1/sofia $DIR/$tdir/foo2/maggie + if [ "$($LFS fid2path $DIR --link 1 $FID)" != "/$tdir/foo2/maggie" ] + then + $LFS fid2path $DIR $FID + err17935 "bad link rename" + fi + rm $DIR/$tdir/foo2/maggie + + # overflow the EA + local longname=filename_avg_len_is_thirty_two_ + createmany -l$DIR/$tdir/foo1/luna $DIR/$tdir/foo2/$longname 1000 || \ + error "failed to hardlink many files" + links=$($LFS fid2path $DIR $FID | wc -l) + echo -n "${links}/1000 links in link EA" + [ ${links} -gt 60 ] || err17935 "expected at least 60 links in link EA" + unlinkmany $DIR/$tdir/foo2/$longname 1000 || \ + error "failed to unlink many hardlinks" +} +run_test 161 "link ea sanity" + +check_path() { + local expected=$1 + shift + local fid=$2 + + local path=$(${LFS} fid2path $*) + RC=$? + + if [ $RC -ne 0 ]; then + err17935 "path looked up of $expected failed. Error $RC" + return $RC + elif [ "${path}" != "${expected}" ]; then + err17935 "path looked up \"${path}\" instead of \"${expected}\"" + return 2 + fi + echo "fid $fid resolves to path $path" +} + +test_162() { + # Make changes to filesystem + mkdir -p $DIR/$tdir/d2 + touch $DIR/$tdir/d2/$tfile + touch $DIR/$tdir/d2/x1 + touch $DIR/$tdir/d2/x2 + mkdir -p $DIR/$tdir/d2/a/b/c + mkdir -p $DIR/$tdir/d2/p/q/r + FID=$($LFS path2fid $DIR/$tdir/d2/$tfile | tr -d '[') + check_path "/$tdir/d2/$tfile" $DIR $FID --link 0 + ln $DIR/$tdir/d2/$tfile $DIR/$tdir/d2/p/q/r/hlink + mv $DIR/$tdir/d2/$tfile $DIR/$tdir/d2/a/b/c/new_file + FID=$($LFS path2fid $DIR/$tdir/d2/a/b/c/new_file | tr -d '[') + # fid2path dir/fsname should both work + check_path "/$tdir/d2/a/b/c/new_file" $FSNAME $FID --link 1 + check_path "/$tdir/d2/p/q/r/hlink" $DIR $FID --link 0 + # check that there are 2 links + ${LFS} fid2path $DIR $FID | wc -l | grep -q 2 || \ + err17935 "expected 2 links" + + rm $DIR/$tdir/d2/p/q/r/hlink + check_path "/$tdir/d2/a/b/c/new_file" $DIR $FID --link 0 + # Doesnt work with CMD yet: 17935 + return 0 +} +run_test 162 "path lookup sanity" + +test_169() { + # do directio so as not to populate the page cache + log "creating a 10 Mb file" + multiop $DIR/$tfile oO_CREAT:O_DIRECT:O_RDWR:w$((10*1048576))c || error "multiop failed while creating a file" + log "starting reads" + dd if=$DIR/$tfile of=/dev/null bs=4096 & + log "truncating the file" + multiop $DIR/$tfile oO_TRUNC:c || error "multiop failed while truncating the file" + log "killing dd" + kill %+ || true # reads might have finished + echo "wait until dd is finished" + wait + log "removing the temporary file" + rm -rf $DIR/$tfile || error "tmp file removal failed" +} +run_test 169 "parallel read and truncate should not deadlock ===" + +test_170() { + $LCTL clear # bug 18514 + $LCTL debug_daemon start $TMP/${tfile}_log_good + touch $DIR/$tfile + $LCTL debug_daemon stop + sed -e "s/^...../a/g" $TMP/${tfile}_log_good > $TMP/${tfile}_log_bad || + error "sed failed to read log_good" + + $LCTL debug_daemon start $TMP/${tfile}_log_good + rm -rf $DIR/$tfile + $LCTL debug_daemon stop + + $LCTL df $TMP/${tfile}_log_bad > $TMP/${tfile}_log_bad.out 2>&1 || + error "lctl df log_bad failed" + + local bad_line=$(tail -n 1 $TMP/${tfile}_log_bad.out | awk '{print $9}') + local good_line1=$(tail -n 1 $TMP/${tfile}_log_bad.out | awk '{print $5}') + + $LCTL df $TMP/${tfile}_log_good > $TMP/${tfile}_log_good.out 2>&1 + local good_line2=$(tail -n 1 $TMP/${tfile}_log_good.out | awk '{print $5}') + + [ "$bad_line" ] && [ "$good_line1" ] && [ "$good_line2" ] || + error "bad_line good_line1 good_line2 are empty" + + cat $TMP/${tfile}_log_good >> $TMP/${tfile}_logs_corrupt + cat $TMP/${tfile}_log_bad >> $TMP/${tfile}_logs_corrupt + cat $TMP/${tfile}_log_good >> $TMP/${tfile}_logs_corrupt + + $LCTL df $TMP/${tfile}_logs_corrupt > $TMP/${tfile}_log_bad.out 2>&1 + local bad_line_new=$(tail -n 1 $TMP/${tfile}_log_bad.out | awk '{print $9}') + local good_line_new=$(tail -n 1 $TMP/${tfile}_log_bad.out | awk '{print $5}') + + [ "$bad_line_new" ] && [ "$good_line_new" ] || + error "bad_line_new good_line_new are empty" + + local expected_good=$((good_line1 + good_line2*2)) + + rm -f $TMP/${tfile}* + if [ $bad_line -ne $bad_line_new ]; then + error "expected $bad_line bad lines, but got $bad_line_new" + return 1 + fi + + if [ $expected_good -ne $good_line_new ]; then + error "expected $expected_good good lines, but got $good_line_new" + return 2 + fi + true +} +run_test 170 "test lctl df to handle corrupted log =====================" + +# OST pools tests POOL=${POOL:-cea1} TGT_COUNT=$OSTCOUNT TGTPOOL_FIRST=1 @@ -5802,27 +6414,24 @@ check_file_in_pool() } test_200a() { - remote_mds_nodsh && skip "remote MDS with nodsh" && return + remote_mgs_nodsh && skip "remote MGS with nodsh" && return do_facet mgs $LCTL pool_new $FSNAME.$POOL - do_facet mgs $LCTL get_param -n lov.$FSNAME-MDT0000-mdtlov.pools.$POOL - [ $? == 0 ] || error "Pool creation of $POOL failed" + # get param should return err until pool is created + wait_update $HOSTNAME "lctl get_param -n lov.$FSNAME-*.pools.$POOL 2>/dev/null || echo foo" "" || error "Pool creation of $POOL failed" } run_test 200a "Create new pool ==========================================" test_200b() { - remote_mds_nodsh && skip "remote MDS with nodsh" && return - TGT=$(seq -f $FSNAME-OST%04g_UUID $TGTPOOL_FIRST $TGTPOOL_STEP \ - $TGTPOOL_MAX | tr '\n' ' ') + remote_mgs_nodsh && skip "remote MGS with nodsh" && return + TGT=$(for i in `seq $TGTPOOL_FIRST $TGTPOOL_STEP $TGTPOOL_MAX`; do printf "$FSNAME-OST%04x_UUID " $i; done) do_facet mgs $LCTL pool_add $FSNAME.$POOL \ - $FSNAME-OST[$TGTPOOL_FIRST-$TGTPOOL_MAX/$TGTPOOL_STEP]_UUID - res=$(do_facet mgs $LCTL get_param -n lov.$FSNAME-MDT0000-mdtlov.pools.$POOL | sort \ - | tr '\n' ' ') - [ "$res" = "$TGT" ] || error "Pool content ($res) do not match requested ($TGT)" + $FSNAME-OST[$TGTPOOL_FIRST-$TGTPOOL_MAX/$TGTPOOL_STEP] + wait_update $HOSTNAME "lctl get_param -n lov.$FSNAME-*.pools.$POOL | sort -u | tr '\n' ' ' " "$TGT" || error "Add to pool failed" } run_test 200b "Add targets to a pool ====================================" test_200c() { - remote_mds_nodsh && skip "remote MDS with nodsh" && return + remote_mgs_nodsh && skip "remote MGS with nodsh" && return mkdir -p $POOL_DIR $SETSTRIPE -c 2 -p $POOL $POOL_DIR [ $? = 0 ] || error "Cannot set pool $POOL to $POOL_DIR" @@ -5830,14 +6439,14 @@ test_200c() { run_test 200c "Set pool on a directory =================================" test_200d() { - remote_mds_nodsh && skip "remote MDS with nodsh" && return - res=$($GETSTRIPE $POOL_DIR | grep pool: | cut -f8 -d " ") - [ "$res" = $POOL ] || error "Pool on $POOL_DIR is not $POOL" + remote_mgs_nodsh && skip "remote MGS with nodsh" && return + res=$($GETSTRIPE --pool $POOL_DIR | awk '/^pool:/ {print $2}') + [ "$res" = $POOL ] || error "Pool on $POOL_DIR is $res, not $POOL" } run_test 200d "Check pool on a directory ===============================" test_200e() { - remote_mds_nodsh && skip "remote MDS with nodsh" && return + remote_mgs_nodsh && skip "remote MGS with nodsh" && return failed=0 for i in $(seq -w 1 $(($TGT_COUNT * 3))) do @@ -5854,7 +6463,7 @@ test_200e() { run_test 200e "Check files allocation from directory pool ==============" test_200f() { - remote_mds_nodsh && skip "remote MDS with nodsh" && return + remote_mgs_nodsh && skip "remote MGS with nodsh" && return mkdir -p $POOL_FILE failed=0 for i in $(seq -w 1 $(($TGT_COUNT * 3))) @@ -5872,36 +6481,71 @@ test_200f() { run_test 200f "Create files in a pool ===================================" test_200g() { - remote_mds_nodsh && skip "remote MDS with nodsh" && return - TGT=$(do_facet mgs $LCTL get_param -n lov.$FSNAME-MDT0000-mdtlov.pools.$POOL | head -1) + remote_mgs_nodsh && skip "remote MGS with nodsh" && return + TGT=$($LCTL get_param -n lov.$FSNAME-*.pools.$POOL | head -1) do_facet mgs $LCTL pool_remove $FSNAME.$POOL $TGT - res=$(do_facet mgs $LCTL get_param -n lov.$FSNAME-MDT0000-mdtlov.pools.$POOL | grep $TGT) - [ "$res" = "" ] || error "$TGT not removed from $FSNAME.$POOL" + wait_update $HOSTNAME "lctl get_param -n lov.$FSNAME-*.pools.$POOL | grep $TGT" "" || error "$TGT not removed from $FSNAME.$POOL" } run_test 200g "Remove a target from a pool =============================" test_200h() { - remote_mds_nodsh && skip "remote MDS with nodsh" && return - for TGT in $(do_facet mgs $LCTL get_param -n lov.$FSNAME-MDT0000-mdtlov.pools.$POOL) + remote_mgs_nodsh && skip "remote MGS with nodsh" && return + for TGT in $($LCTL get_param -n lov.$FSNAME-*.pools.$POOL | sort -u) do do_facet mgs $LCTL pool_remove $FSNAME.$POOL $TGT done - res=$(do_facet mgs $LCTL get_param -n lov.$FSNAME-MDT0000-mdtlov.pools.$POOL) - [ "$res" = "" ] || error "Pool $FSNAME.$POOL cannot be drained" + wait_update $HOSTNAME "lctl get_param -n lov.$FSNAME-*.pools.$POOL" ""\ + || error "Pool $FSNAME.$POOL cannot be drained" + # striping on an empty pool should fall back to "pool of everything" + $SETSTRIPE -p $POOL ${POOL_FILE}/$tfile || error "failed to create file with empty pool" } run_test 200h "Remove all targets from a pool ==========================" test_200i() { - remote_mds_nodsh && skip "remote MDS with nodsh" && return + remote_mgs_nodsh && skip "remote MGS with nodsh" && return do_facet mgs $LCTL pool_destroy $FSNAME.$POOL - res=$(do_facet mgs "$LCTL get_param -n lov.$FSNAME-MDT0000-mdtlov.pools.$POOL 2>/dev/null") - [ "$res" = "" ] || error "Pool $FSNAME.$POOL is not destroyed" + # get param should return err once pool is gone + wait_update $HOSTNAME "lctl get_param -n lov.$FSNAME-*.pools.$POOL 2>/dev/null || echo foo" "foo" && return 0 + error "Pool $FSNAME.$POOL is not destroyed" } run_test 200i "Remove a pool ============================================" -TMPDIR=$OLDTMPDIR -TMP=$OLDTMP -HOME=$OLDHOME +test_212() { + size=`date +%s` + size=$((size % 8192 + 1)) + dd if=/dev/urandom of=$DIR/f212 bs=1k count=$size + sendfile $DIR/f212 $DIR/f212.xyz || error "sendfile wrong" + rm -f $DIR/f212 $DIR/f212.xyz +} +run_test 212 "Sendfile test ============================================" + +test_213() { + dd if=/dev/zero of=$DIR/$tfile bs=4k count=4 + cancel_lru_locks osc + lctl set_param fail_loc=0x8000040f + # generate a read lock + cat $DIR/$tfile > /dev/null + # write to the file, it will try to cancel the above read lock. + cat /etc/hosts >> $DIR/$tfile +} +run_test 213 "OSC lock completion and cancel race don't crash - bug 18829" + +# +# tests that do cleanup/setup should be run at the end +# + +test_900() { + local ls + #define OBD_FAIL_MGC_PAUSE_PROCESS_LOG 0x903 + $LCTL set_param fail_loc=0x903 + # cancel_lru_locks mgc - does not work due to lctl set_param syntax + for ls in /proc/fs/lustre/ldlm/namespaces/MGC*/lru_size; do + echo "clear" > $ls + done + FAIL_ON_ERROR=true cleanup + FAIL_ON_ERROR=true setup +} +run_test 900 "umount should not race with any mgc requeue thread" log "cleanup: ======================================================" check_and_cleanup_lustre