X-Git-Url: https://git.whamcloud.com/?p=fs%2Flustre-release.git;a=blobdiff_plain;f=lustre%2Ftests%2Fsanity.sh;h=a40f77d1e1d716f1dcbd62c43b89c8bbb47c8763;hp=6c15e9ef4b917d949eecde8221d1bff28c2b2a55;hb=16ddb77519b26ea8df18b5abd84afb03d08c47dc;hpb=5e9479cd77214d1b85056edaca66ac38b784be6d diff --git a/lustre/tests/sanity.sh b/lustre/tests/sanity.sh old mode 100644 new mode 100755 index 6c15e9e..a40f77d --- a/lustre/tests/sanity.sh +++ b/lustre/tests/sanity.sh @@ -9,8 +9,8 @@ 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 68b 75 119d $SANITY_EXCEPT" +# bug number for skipped test: 13297 2108 9789 3637 9789 3561 12622 12653 12653 5188 16260 19742 +ALWAYS_EXCEPT=" 27u 42a 42b 42c 42d 45 51d 65a 65e 68b $SANITY_EXCEPT" # bug number for skipped test: 2108 9789 3637 9789 3561 5188/5749 1443 #ALWAYS_EXCEPT=${ALWAYS_EXCEPT:-"27m 42a 42b 42c 42d 45 68 76"} # UPDATE THE COMMENT ABOVE WITH BUG NUMBERS WHEN CHANGING ALWAYS_EXCEPT! @@ -20,9 +20,6 @@ CPU=`awk '/model/ {print $4}' /proc/cpuinfo` # buffer i/o errs sock spc runas [ "$CPU" = "UML" ] && EXCEPT="$EXCEPT 27m 27n 27o 27p 27q 27r 31d 54a 64b 99a 99b 99c 99d 99e 99f 101" -# test76 is not valid with FIDs because inode numbers are not reused -ALWAYS_EXCEPT="$ALWAYS_EXCEPT 76" - case `uname -r` in 2.4*) FSTYPE=${FSTYPE:-ext3} ;; 2.6*) FSTYPE=${FSTYPE:-ldiskfs} ;; @@ -47,7 +44,6 @@ MCREATE=${MCREATE:-mcreate} OPENFILE=${OPENFILE:-openfile} OPENUNLINK=${OPENUNLINK:-openunlink} READS=${READS:-"reads"} -TRUNCATE=${TRUNCATE:-truncate} MUNLINK=${MUNLINK:-munlink} SOCKETSERVER=${SOCKETSERVER:-socketserver} SOCKETCLIENT=${SOCKETCLIENT:-socketclient} @@ -72,8 +68,9 @@ LUSTRE=${LUSTRE:-$(cd $(dirname $0)/..; echo $PWD)} . $LUSTRE/tests/test-framework.sh init_test_env $@ . ${CONFIG:=$LUSTRE/tests/cfg/${NAME}.sh} +init_logging -[ "$SLOW" = "no" ] && EXCEPT_SLOW="24o 24v 27m 36f 36g 51b 51c 60c 63 64b 68 71 73 77f 78 101 103 115 120g 124b" +[ "$SLOW" = "no" ] && EXCEPT_SLOW="24o 24v 27m 36f 36g 36h 51b 51c 60c 63 64b 68 71 73 77f 78 101 103 115 120g 124b" SANITYLOG=${TESTSUITELOG:-$TMP/$(basename $0 .sh).log} FAIL_ON_ERROR=false @@ -129,7 +126,7 @@ 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_ID $RUNAS +check_runas_id $RUNAS_ID $RUNAS_GID $RUNAS build_test_filter @@ -303,16 +300,16 @@ test_6g() { $RUNAS mkdir $DIR/d6g/d || error chmod g+s $DIR/d6g/d || error mkdir $DIR/d6g/d/subdir - $CHECKSTAT -g \#$RUNAS_ID $DIR/d6g/d/subdir || error + $CHECKSTAT -g \#$RUNAS_GID $DIR/d6g/d/subdir || error } run_test 6g "Is new dir in sgid dir inheriting group?" test_6h() { # bug 7331 [ $RUNAS_ID -eq $UID ] && skip_env "RUNAS_ID = UID = $UID -- skipping" && return touch $DIR/f6h || error "touch failed" - chown $RUNAS_ID:$RUNAS_ID $DIR/f6h || error "initial chown failed" - $RUNAS -G$RUNAS_ID chown $RUNAS_ID:0 $DIR/f6h && error "chown worked" - $CHECKSTAT -t file -u \#$RUNAS_ID -g \#$RUNAS_ID $DIR/f6h || error + chown $RUNAS_ID:$RUNAS_GID $DIR/f6h || error "initial chown failed" + $RUNAS -G$RUNAS_GID chown $RUNAS_ID:0 $DIR/f6h && error "chown worked" + $CHECKSTAT -t file -u \#$RUNAS_ID -g \#$RUNAS_GID $DIR/f6h || error } run_test 6h "$RUNAS chown RUNAS_ID.0 .../f6h (should return error)" @@ -479,7 +476,7 @@ 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 + do_facet $SINGLEMDS lctl set_param fail_loc=0x80000141 touch $DIR/$tdir/$tfile || true } run_test 17h "create objects: lov_free_memmd() doesn't lbug" @@ -489,12 +486,23 @@ test_17i() { #bug 20018 local foo=$DIR/$tdir/$tfile ln -s $foo $foo || error "create symlink failed" #define OBD_FAIL_MDS_READLINK_EPROTO 0x143 - do_facet mds lctl set_param fail_loc=0x80000144 + do_facet $SINGLEMDS lctl set_param fail_loc=0x80000143 ls -l $foo && error "error not detected" return 0 } run_test 17i "don't panic on short symlink" +test_17k() { #bug 22301 + rsync --help | grep -q xattr || + skip_env "$(rsync --version| head -1) does not support xattrs" + mkdir -p $DIR/{$tdir,$tdir.new} + touch $DIR/$tdir/$tfile + ln -s $DIR/$tdir/$tfile $DIR/$tdir/$tfile.lnk + rsync -av -X $DIR/$tdir/ $DIR/$tdir.new || + error "rsync failed with xattrs enabled" +} +run_test 17k "symlinks: rsync with xattrs enabled =========================" + test_18() { touch $DIR/f ls $DIR || error @@ -553,13 +561,13 @@ run_test 21 "write to dangling link ============================" test_22() { WDIR=$DIR/$tdir mkdir -p $WDIR - chown $RUNAS_ID $WDIR + chown $RUNAS_ID:$RUNAS_GID $WDIR (cd $WDIR || error "cd $WDIR failed"; $RUNAS tar cf - /etc/hosts /etc/sysconfig/network | \ $RUNAS tar xf -) ls -lR $WDIR/etc || error "ls -lR $WDIR/etc failed" $CHECKSTAT -t dir $WDIR/etc || error "checkstat -t dir failed" - $CHECKSTAT -u \#$RUNAS_ID $WDIR/etc || error "checkstat -u failed" + $CHECKSTAT -u \#$RUNAS_ID -g \#$RUNAS_GID $WDIR/etc || error "checkstat -u failed" } run_test 22 "unpack tar archive as non-root user ===============" @@ -796,6 +804,17 @@ test_24v() { } run_test 24v "list directory with large files (handle hash collision, bug: 17560)" +test_24w() { # bug21506 + SZ1=234852 + dd if=/dev/zero of=$DIR/$tfile bs=1M count=1 seek=4096 || return 1 + dd if=/dev/zero bs=$SZ1 count=1 >> $DIR/$tfile || return 2 + dd if=$DIR/$tfile of=$DIR/${tfile}_left bs=1M skip=4097 || return 3 + SZ2=`ls -l $DIR/${tfile}_left | awk '{print $5}'` + [ "$SZ1" = "$SZ2" ] || \ + error "Error reading at the end of the file $tfile" +} +run_test 24w "Reading a file larger than 4Gb" + test_25a() { echo '== symlink sanity =============================================' @@ -983,22 +1002,30 @@ test_27m() { } run_test 27m "create file while OST0 was full ==================" -# osc's keep a NOSPC stick flag that gets unset with rmdir +sleep_maxage() { + local DELAY=$(do_facet $SINGLEMDS lctl get_param -n lov.*.qos_maxage | head -n 1 | awk '{print $1 * 2}') + sleep $DELAY +} + +# OSCs keep a NOSPC flag that will be reset after ~5s (qos_maxage) +# if the OST isn't full anymore. reset_enospc() { - local FAIL_LOC=${1:-0} - local OSTIDX=${2:-""} + local OSTIDX=${1:-""} - mkdir -p $DIR/d27/nospc - rmdir $DIR/d27/nospc local list=$(comma_list $(osts_nodes)) [ "$OSTIDX" ] && list=$(facet_host ost$((OSTIDX + 1))) - do_nodes $list lctl set_param fail_loc=$FAIL_LOC + do_nodes $list lctl set_param fail_loc=0 + sleep_maxage } exhaust_precreations() { local OSTIDX=$1 - local MDSIDX=$(get_mds_dir "$DIR/d27") + local FAILLOC=$2 + local FAILIDX=${3:-$OSTIDX} + + mkdir -p $DIR/$tdir + local MDSIDX=$(get_mds_dir "$DIR/$tdir") echo OSTIDX=$OSTIDX MDSIDX=$MDSIDX local OST=$(lfs osts | grep ${OSTIDX}": " | \ @@ -1007,28 +1034,32 @@ exhaust_precreations() { sed -e 's/_UUID$//;s/^.*-//') # on the mdt's osc - 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) + local mdtosc_proc1=$(get_mdtosc_proc_path mds${MDSIDX} $OST) + local last_id=$(do_facet mds${MDSIDX} lctl get_param -n \ + osc.$mdtosc_proc1.prealloc_last_id) + local next_id=$(do_facet mds${MDSIDX} lctl get_param -n \ + osc.$mdtosc_proc1.prealloc_next_id) - echo ${OST}-osc-${MDT_INDEX}.prealloc_last_id=$last_id - echo ${OST}-osc-${MDT_INDEX}.prealloc_next_id=$next_id + local mdtosc_proc2=$(get_mdtosc_proc_path mds${MDSIDX}) + do_facet mds${MDSIDX} lctl get_param osc.$mdtosc_proc2.prealloc* - mkdir -p $DIR/d27 - $SETSTRIPE $DIR/d27 -i $OSTIDX -c 1 + mkdir -p $DIR/$tdir/${OST} + $SETSTRIPE $DIR/$tdir/${OST} -i $OSTIDX -c 1 #define OBD_FAIL_OST_ENOSPC 0x215 + do_facet ost$((OSTIDX + 1)) lctl set_param fail_val=$FAILIDX do_facet ost$((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 $((last_id - next_id + 2)) - do_facet mds${MDSIDX} lctl get_param osc.*${OST}-osc-${MDT_INDEX}.prealloc* | grep '[0-9]' - reset_enospc $2 $OSTIDX + createmany -o $DIR/$tdir/${OST}/f $next_id $((last_id - next_id + 2)) + do_facet mds${MDSIDX} lctl get_param osc.$mdtosc_proc2.prealloc* + do_facet ost$((OSTIDX + 1)) lctl set_param fail_loc=$FAILLOC + sleep_maxage } exhaust_all_precreations() { local i for (( i=0; i < OSTCOUNT; i++ )) ; do - exhaust_precreations $i 0x215 + exhaust_precreations $i $1 -1 done - reset_enospc $1 } test_27n() { @@ -1037,11 +1068,11 @@ test_27n() { remote_ost_nodsh && skip "remote OST with nodsh" && return reset_enospc - rm -f $DIR/d27/f27n + rm -f $DIR/$tdir/$tfile exhaust_precreations 0 0x80000215 - $SETSTRIPE -c -1 $DIR/d27 - touch $DIR/d27/f27n || error - $GETSTRIPE $DIR/d27/f27n + $SETSTRIPE -c -1 $DIR/$tdir + touch $DIR/$tdir/$tfile || error + $GETSTRIPE $DIR/$tdir/$tfile reset_enospc } run_test 27n "create file with some full OSTs ==================" @@ -1052,14 +1083,13 @@ test_27o() { remote_ost_nodsh && skip "remote OST with nodsh" && return reset_enospc - rm -f $DIR/d27/f27o + rm -f $DIR/$tdir/$tfile exhaust_all_precreations 0x215 - sleep 5 - touch $DIR/d27/f27o && error "able to create $DIR/d27/f27o" + touch $DIR/$tdir/$tfile && error "able to create $DIR/$tdir/$tfile" reset_enospc - rm -rf $DIR/d27/* + rm -rf $DIR/$tdir/* } run_test 27o "create file with all full OSTs (should error) ====" @@ -1069,15 +1099,17 @@ test_27p() { remote_ost_nodsh && skip "remote OST with nodsh" && return reset_enospc - rm -f $DIR/d27/f27p + rm -f $DIR/$tdir/$tfile + mkdir -p $DIR/$tdir - $MCREATE $DIR/d27/f27p || error "mcreate failed" - $TRUNCATE $DIR/d27/f27p 80000000 || error "truncate failed" - $CHECKSTAT -s 80000000 $DIR/d27/f27p || error "checkstat failed" + $MCREATE $DIR/$tdir/$tfile || error "mcreate failed" + $TRUNCATE $DIR/$tdir/$tfile 80000000 || error "truncate failed" + $CHECKSTAT -s 80000000 $DIR/$tdir/$tfile || error "checkstat failed" exhaust_precreations 0 0x80000215 - echo foo >> $DIR/d27/f27p || error "append failed" - $CHECKSTAT -s 80000004 $DIR/d27/f27p || error "checkstat failed" + echo foo >> $DIR/$tdir/$tfile || error "append failed" + $CHECKSTAT -s 80000004 $DIR/$tdir/$tfile || error "checkstat failed" + $LFS getstripe $DIR/$tdir/$tfile reset_enospc } @@ -1089,16 +1121,16 @@ test_27q() { remote_ost_nodsh && skip "remote OST with nodsh" && return reset_enospc - rm -f $DIR/d27/f27q + rm -f $DIR/$tdir/$tfile - $MCREATE $DIR/d27/f27q || error "mcreate $DIR/d27/f27q failed" - $TRUNCATE $DIR/d27/f27q 80000000 ||error "truncate $DIR/d27/f27q failed" - $CHECKSTAT -s 80000000 $DIR/d27/f27q || error "checkstat failed" + $MCREATE $DIR/$tdir/$tfile || error "mcreate $DIR/$tdir/$tfile failed" + $TRUNCATE $DIR/$tdir/$tfile 80000000 ||error "truncate $DIR/$tdir/$tfile failed" + $CHECKSTAT -s 80000000 $DIR/$tdir/$tfile || error "checkstat failed" exhaust_all_precreations 0x215 - echo foo >> $DIR/d27/f27q && error "append succeeded" - $CHECKSTAT -s 80000000 $DIR/d27/f27q || error "checkstat 2 failed" + echo foo >> $DIR/$tdir/$tfile && error "append succeeded" + $CHECKSTAT -s 80000000 $DIR/$tdir/$tfile || error "checkstat 2 failed" reset_enospc } @@ -1110,10 +1142,10 @@ test_27r() { remote_ost_nodsh && skip "remote OST with nodsh" && return reset_enospc - rm -f $DIR/d27/f27r + rm -f $DIR/$tdir/$tfile exhaust_precreations 0 0x80000215 - $SETSTRIPE $DIR/d27/f27r -i 0 -c 2 # && error + $SETSTRIPE $DIR/$tdir/$tfile -i 0 -c 2 # && error reset_enospc } @@ -1146,14 +1178,14 @@ test_27u() { # bug 4900 #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 + mkdir -p $DIR/$tdir + createmany -o $DIR/$tdir/t- 1000 do_facet $SINGLEMDS lctl set_param fail_loc=0 TLOG=$DIR/$tfile.getstripe - $GETSTRIPE $DIR/d27u > $TLOG + $GETSTRIPE $DIR/$tdir > $TLOG OBJS=`awk -vobj=0 '($1 == 0) { obj += 1 } END { print obj;}' $TLOG` - unlinkmany $DIR/d27u/t- 1000 + unlinkmany $DIR/$tdir/t- 1000 [ $OBJS -gt 0 ] && \ error "$OBJS objects created on OST-0. See $TLOG" || pass } @@ -1164,7 +1196,8 @@ test_27v() { # bug 4900 remote_mds_nodsh && skip "remote MDS with nodsh" && return remote_ost_nodsh && skip "remote OST with nodsh" && return - exhaust_all_precreations + exhaust_all_precreations 0x215 + reset_enospc mkdir -p $DIR/$tdir $SETSTRIPE $DIR/$tdir -c 1 # 1 stripe / file @@ -1178,8 +1211,6 @@ test_27v() { # bug 4900 local START=`date +%s` createmany -o $DIR/$tdir/$tfile 32 - reset_enospc - local FINISH=`date +%s` local TIMEOUT=`lctl get_param -n timeout` [ $((FINISH - START)) -ge $((TIMEOUT / 2)) ] && \ @@ -1190,17 +1221,19 @@ test_27v() { # bug 4900 run_test 27v "skip object creation on slow OST =================" test_27w() { # bug 10997 - mkdir -p $DIR/d27w || error "mkdir failed" - $LSTRIPE $DIR/d27w/f0 -s 65536 || error "lstripe failed" - size=`$GETSTRIPE $DIR/d27w/f0 -qs` + mkdir -p $DIR/$tdir || error "mkdir failed" + $LSTRIPE $DIR/$tdir/f0 -s 65536 || error "lstripe failed" + size=`$GETSTRIPE $DIR/$tdir/f0 -s` [ $size -ne 65536 ] && error "stripe size $size != 65536" || true + gsdir=$($LFS getstripe -d $DIR/$tdir) + [ $(echo $gsdir | grep -c stripe_count) -ne 1 ] && error "$LFS getstripe -d $DIR/$tdir failed" [ "$OSTCOUNT" -lt "2" ] && skip_env "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 $offset failed" - count=`$GETSTRIPE -qc $DIR/d27w/f$i` - index=`$GETSTRIPE -qo $DIR/d27w/f$i` + $LSTRIPE $DIR/$tdir/f$i -c $i -i $offset || error "lstripe -c $i -i $offset failed" + count=`$GETSTRIPE -c $DIR/$tdir/f$i` + index=`$GETSTRIPE -o $DIR/$tdir/f$i` [ $count -ne $i ] && error "stripe count $count != $i" || true [ $index -ne $offset ] && error "stripe offset $index != $offset" || true done @@ -1209,21 +1242,20 @@ run_test 27w "check lfs setstripe -c -s -i options =============" test_27x() { [ "$OSTCOUNT" -lt "2" ] && skip_env "$OSTCOUNT < 2 OSTs" && return - DELAY=$(do_facet mds lctl get_param -n lov.*.qos_maxage | awk '{print $1 + 2}') - OFFSET=$(($OSTCOUNTi - 1)) + OFFSET=$(($OSTCOUNT - 1)) OSTIDX=0 local OST=$(lfs osts | awk '/'${OSTIDX}': / { print $2 }' | sed -e 's/_UUID$//') mkdir -p $DIR/$tdir $SETSTRIPE $DIR/$tdir -c 1 # 1 stripe per file - do_facet ost$OSTIDX lctl set_param -n obdfilter.$OST.degraded 1 - sleep $DELAY + do_facet ost$((OSTIDX + 1)) lctl set_param -n obdfilter.$OST.degraded 1 + sleep_maxage createmany -o $DIR/$tdir/$tfile $OSTCOUNT for i in `seq 0 $OFFSET`; do [ `$GETSTRIPE $DIR/$tdir/$tfile$i | grep -A 10 obdidx | awk '{print $1}' | grep -w "$OSTIDX"` ] && error "OST0 was degraded but new created file still use it" done - do_facet ost$OSTIDX lctl set_param -n obdfilter.$OST.degraded 0 + do_facet ost$((OSTIDX + 1)) lctl set_param -n obdfilter.$OST.degraded 0 } run_test 27x "create files while OST0 is degraded" @@ -1231,8 +1263,16 @@ test_27y() { [ "$OSTCOUNT" -lt "2" ] && skip_env "$OSTCOUNT < 2 OSTs -- skipping" && return remote_mds_nodsh && skip "remote MDS with nodsh" && return - MDS_OSCS=`do_facet mds lctl dl | awk '/[oO][sS][cC].*md[ts]/ { print $4 }'` - DELAY=$(do_facet mds lctl get_param -n lov.*.qos_maxage | awk '{print $1 + 2}') + local mdtosc=$(get_mdtosc_proc_path $SINGLEMDS $FSNAME-OST0000) + local last_id=$(do_facet $SINGLEMDS lctl get_param -n \ + osc.$mdtosc.prealloc_last_id) + local next_id=$(do_facet $SINGLEMDS lctl get_param -n \ + osc.$mdtosc.prealloc_next_id) + local fcount=$((last_id - next_id)) + [ $fcount -eq 0 ] && skip "not enough space on OST0" && return + [ $fcount -gt $OSTCOUNT ] && fcount=$OSTCOUNT + + MDS_OSCS=`do_facet $SINGLEMDS lctl dl | awk '/[oO][sS][cC].*md[ts]/ { print $4 }'` OFFSET=$(($OSTCOUNT-1)) OST=-1 for OSC in $MDS_OSCS; do @@ -1240,7 +1280,7 @@ test_27y() { OST=`osc_to_ost $OSC` } else { echo $OSC "is Deactivate:" - do_facet mds lctl --device %$OSC deactivate + do_facet $SINGLEMDS lctl --device %$OSC deactivate } fi done @@ -1249,8 +1289,8 @@ test_27y() { $SETSTRIPE $DIR/$tdir -c 1 # 1 stripe / file do_facet ost$OSTIDX lctl set_param -n obdfilter.$OST.degraded 1 - sleep $DELAY - createmany -o $DIR/$tdir/$tfile $OSTCOUNT + sleep_maxage + createmany -o $DIR/$tdir/$tfile $fcount do_facet ost$OSTIDX lctl set_param -n obdfilter.$OST.degraded 0 for i in `seq 0 $OFFSET`; do @@ -1260,12 +1300,94 @@ test_27y() { for OSC in $MDS_OSCS; do [ `osc_to_ost $OSC` != $OST ] && { echo $OSC "is activate" - do_facet mds lctl --device %$OSC activate + do_facet $SINGLEMDS lctl --device %$OSC activate } done } run_test 27y "create files while OST0 is degraded and the rest inactive" +check_seq_oid() +{ + echo check file $1 + local old_ifs="$IFS" + IFS=$'\t\n :' + lmm=($($GETSTRIPE -v $1)) + + IFS=$'[:]' + fid=($($LFS path2fid $1)) + IFS="$old_ifs" + + # compare lmm_seq and lu_fid->f_seq + [ ${lmm[4]} = ${fid[1]} ] || { error "SEQ mismatch"; return 1; } + # compare lmm_object_id and lu_fid->oid + [ ${lmm[6]} = ${fid[2]} ] || { error "OID mismatch"; return 2; } + + echo -e "\tseq ${fid[1]}, oid ${fid[2]} ver ${fid[3]}\n\tstripe count: ${lmm[8]}" + + [ "$FSTYPE" != "ldiskfs" ] && skip "can not check trusted.fid FSTYPE=$FSTYPE" && return 0 + + # check the trusted.fid attribute of the OST objects of the file + for (( i=0, j=19; i < ${lmm[8]}; i++, j+=4 )); do + local obdidx=${lmm[$j]} + local devnum=$((obdidx + 1)) + local objid=${lmm[$((j+1))]} + local group=${lmm[$((j+3))]} + local dev=$(ostdevname $devnum) + local dir=${MOUNT%/*}/ost$devnum + do_facet ost$devnum mount -t $FSTYPE $dev $dir $OST_MOUNT_OPTS || + { error "mounting $dev as $FSTYPE failed"; return 3; } + + obj_filename=$(do_facet ost$devnum find $dir/O/$group -name $objid) + local ff=$(do_facet ost$devnum $LL_DECODE_FILTER_FID $obj_filename) + IFS=$'/= [:]' + ff=($(echo $ff)) + IFS="$old_ifs" + + # compare lmm_seq and filter_fid->ff_parent.f_seq + [ ${ff[11]} = ${lmm[4]} ] || { error "parent SEQ mismatch"; return 4; } + # compare lmm_object_id and filter_fid->ff_parent.f_oid + [ ${ff[12]} = ${lmm[6]} ] || { error "parent OID mismatch"; return 5; } + let stripe=${ff[13]} + [ $stripe -eq $i ] || { error "stripe mismatch"; return 6; } + + echo -e "\t\tost $obdidx, objid $objid, group $group" + do_facet ost$devnum umount -d $dev + done +} + +test_27z() { + mkdir -p $DIR/$tdir + $SETSTRIPE $DIR/$tdir/$tfile-1 -c 1 -o 0 -s 1m || + { error "setstripe -c -1 failed"; return 1; } + dd if=/dev/zero of=$DIR/$tdir/$tfile-1 bs=1M count=1 || + { error "dd 1 mb failed"; return 2; } + $SETSTRIPE $DIR/$tdir/$tfile-2 -c -1 -o $(($OSTCOUNT - 1)) -s 1m || + { error "setstripe -c 1 failed"; return 3; } + dd if=/dev/zero of=$DIR/$tdir/$tfile-2 bs=1M count=$OSTCOUNT || + { error "dd $OSTCOUNT mb failed"; return 4; } + sync + + check_seq_oid $DIR/$tdir/$tfile-1 || return 5 + check_seq_oid $DIR/$tdir/$tfile-2 || return 6 +} +run_test 27z "check SEQ/OID on the MDT and OST filesystems" + +test_27A() { # b=19102 + local restore_size=`$GETSTRIPE -s $MOUNT` + local restore_count=`$GETSTRIPE -c $MOUNT` + local restore_offset=`$GETSTRIPE -o $MOUNT` + $SETSTRIPE -c 0 -o -1 -s 0 $MOUNT + local default_size=`$GETSTRIPE -s $MOUNT` + local default_count=`$GETSTRIPE -c $MOUNT` + local default_offset=`$GETSTRIPE -o $MOUNT` + local dsize=$((1024 * 1024)) + [ $default_size -eq $dsize ] || error "stripe size $default_size != $dsize" + [ $default_count -eq 1 ] || error "stripe count $default_count != 1" + [ $default_offset -eq -1 ] || error "stripe offset $default_offset != -1" + $SETSTRIPE -c $restore_count -o $restore_offset -s $restore_size $MOUNT +} +run_test 27A "check filesystem-wide default LOV EA values" + # createtest also checks that device nodes are created and # then visible correctly (#2091) test_28() { # bug 2091 @@ -1381,7 +1503,7 @@ test_31f() { # bug 4554 $SETSTRIPE $DIR/d31f -s 1048576 -c 1 cp /etc/hosts $DIR/d31f ls -l $DIR/d31f - $DIR/d31f/hosts + $GETSTRIPE $DIR/d31f/hosts multiop_bg_pause $DIR/d31f D_c || return 1 MULTIPID2=$! @@ -1711,6 +1833,70 @@ test_33b() { } run_test 33b "test open file with malformed flags (No panic and return error)" +test_33c() { + local ostnum + local ostname + local write_bytes + local all_zeros + + all_zeros=: + rm -fr $DIR/d33 + mkdir -p $DIR/d33 + # Read: 0, Write: 4, create/destroy: 2/0, stat: 1, punch: 0 + + sync + for ostnum in $(seq $OSTCOUNT); do + # test-framework's OST numbering is one-based, while Lustre's + # is zero-based + ostname=$(printf "lustre-OST%.4d" $((ostnum - 1))) + # Parsing llobdstat's output sucks; we could grep the /proc + # path, but that's likely to not be as portable as using the + # llobdstat utility. So we parse lctl output instead. + write_bytes=$(do_facet ost$ostnum lctl get_param -n \ + obdfilter/$ostname/stats | + awk '/^write_bytes/ {print $7}' ) + echo "baseline_write_bytes@$OSTnum/$ostname=$write_bytes" + if (( ${write_bytes:-0} > 0 )) + then + all_zeros=false + break; + fi + done + + $all_zeros || return 0 + + # Write four bytes + echo foo > $DIR/d33/bar + # Really write them + sync + + # Total up write_bytes after writing. We'd better find non-zeros. + for ostnum in $(seq $OSTCOUNT); do + ostname=$(printf "lustre-OST%.4d" $((ostnum - 1))) + write_bytes=$(do_facet ost$ostnum lctl get_param -n \ + obdfilter/$ostname/stats | + awk '/^write_bytes/ {print $7}' ) + echo "write_bytes@$OSTnum/$ostname=$write_bytes" + if (( ${write_bytes:-0} > 0 )) + then + all_zeros=false + break; + fi + done + + if $all_zeros + then + for ostnum in $(seq $OSTCOUNT); do + ostname=$(printf "lustre-OST%.4d" $((ostnum - 1))) + echo "Check that write_bytes is present in obdfilter/*/stats:" + do_facet ost$ostnum lctl get_param -n \ + obdfilter/$ostname/stats + done + error "OST not keeping write_bytes stats (b22312)" + fi +} +run_test 33c "test llobdstat and write_bytes" + TEST_34_SIZE=${TEST_34_SIZE:-2000000000000} test_34a() { rm -f $DIR/f34 @@ -1772,6 +1958,23 @@ test_34f() { # bug 6242, 6243 } run_test 34f "read from a file with no objects until EOF =======" +test_34g() { + dd if=/dev/zero of=$DIR/$tfile bs=1 count=100 seek=$TEST_34_SIZE || error + $TRUNCATE $DIR/$tfile $((TEST_34_SIZE / 2))|| error + $CHECKSTAT -s $((TEST_34_SIZE / 2)) $DIR/$tfile || error "truncate failed" + cancel_lru_locks osc + $CHECKSTAT -s $((TEST_34_SIZE / 2)) $DIR/$tfile || \ + error "wrong size after lock cancel" + + $TRUNCATE $DIR/$tfile $TEST_34_SIZE || error + $CHECKSTAT -s $TEST_34_SIZE $DIR/$tfile || \ + error "expanding truncate failed" + cancel_lru_locks osc + $CHECKSTAT -s $TEST_34_SIZE $DIR/$tfile || \ + error "wrong expanded size after lock cancel" +} +run_test 34g "truncate long file ===============================" + test_35a() { cp /bin/sh $DIR/f35a chmod 444 $DIR/f35a @@ -1817,13 +2020,15 @@ test_36e() { } run_test 36e "utime on non-owned file (should return error) ====" -test_36f() { +subr_36fh() { + local fl="$1" + local LANG_SAVE=$LANG + local LC_LANG_SAVE=$LC_LANG export LANG=C LC_LANG=C # for date language DATESTR="Dec 20 2000" mkdir -p $DIR/$tdir - #define OBD_FAIL_OST_BRW_PAUSE_BULK 0x214 - lctl set_param fail_loc=0x80000214 + lctl set_param fail_loc=$fl date; date +%s cp /etc/hosts $DIR/$tdir/$tfile sync & # write RPC generated with "current" inode timestamp, but delayed @@ -1838,6 +2043,13 @@ test_36f() { echo "AFTER : $LS_AFTER" && \ echo "WANT : $DATESTR" && \ error "$DIR/$tdir/$tfile timestamps changed" || true + + export LANG=$LANG_SAVE LC_LANG=$LC_LANG_SAVE +} + +test_36f() { + #define OBD_FAIL_OST_BRW_PAUSE_BULK 0x214 + subr_36fh "0x80000214" } run_test 36f "utime on file racing with OST BRW write ==========" @@ -1856,6 +2068,12 @@ test_36g() { } run_test 36g "filter mod data cache expiry =====================" +test_36h() { + #define OBD_FAIL_OST_BRW_PAUSE_BULK2 0x227 + subr_36fh "0x80000227" +} +run_test 36h "utime on file racing with OST BRW write ==========" + test_37() { mkdir -p $DIR/$tdir echo f > $DIR/$tdir/fbugfile @@ -1887,16 +2105,310 @@ test_39() { $OPENFILE -f O_CREAT:O_TRUNC:O_WRONLY $DIR/${tfile}2 if [ ! $DIR/${tfile}2 -nt $DIR/$tfile ]; then echo "mtime" - ls -l $DIR/$tfile $DIR/${tfile}2 + ls -l --full-time $DIR/$tfile $DIR/${tfile}2 echo "atime" - ls -lu $DIR/$tfile $DIR/${tfile}2 + ls -lu --full-time $DIR/$tfile $DIR/${tfile}2 echo "ctime" - ls -lc $DIR/$tfile $DIR/${tfile}2 + ls -lc --full-time $DIR/$tfile $DIR/${tfile}2 error "O_TRUNC didn't change timestamps" fi } run_test 39 "mtime changed on create ===========================" +test_39b() { + mkdir -p $DIR/$tdir + cp -p /etc/passwd $DIR/$tdir/fopen + cp -p /etc/passwd $DIR/$tdir/flink + cp -p /etc/passwd $DIR/$tdir/funlink + cp -p /etc/passwd $DIR/$tdir/frename + ln $DIR/$tdir/funlink $DIR/$tdir/funlink2 + + sleep 1 + echo "aaaaaa" >> $DIR/$tdir/fopen + echo "aaaaaa" >> $DIR/$tdir/flink + echo "aaaaaa" >> $DIR/$tdir/funlink + echo "aaaaaa" >> $DIR/$tdir/frename + + local open_new=`stat -c %Y $DIR/$tdir/fopen` + local link_new=`stat -c %Y $DIR/$tdir/flink` + local unlink_new=`stat -c %Y $DIR/$tdir/funlink` + local rename_new=`stat -c %Y $DIR/$tdir/frename` + + cat $DIR/$tdir/fopen > /dev/null + ln $DIR/$tdir/flink $DIR/$tdir/flink2 + rm -f $DIR/$tdir/funlink2 + mv -f $DIR/$tdir/frename $DIR/$tdir/frename2 + + for (( i=0; i < 2; i++ )) ; do + local open_new2=`stat -c %Y $DIR/$tdir/fopen` + local link_new2=`stat -c %Y $DIR/$tdir/flink` + local unlink_new2=`stat -c %Y $DIR/$tdir/funlink` + local rename_new2=`stat -c %Y $DIR/$tdir/frename2` + + [ $open_new2 -eq $open_new ] || error "open file reverses mtime" + [ $link_new2 -eq $link_new ] || error "link file reverses mtime" + [ $unlink_new2 -eq $unlink_new ] || error "unlink file reverses mtime" + [ $rename_new2 -eq $rename_new ] || error "rename file reverses mtime" + + cancel_lru_locks osc + if [ $i = 0 ] ; then echo "repeat after cancel_lru_locks"; fi + done +} +run_test 39b "mtime change on open, link, unlink, rename ======" + +# this should be set to past +TEST_39_MTIME=`date -d "1 year ago" +%s` + +# bug 11063 +test_39c() { + touch $DIR1/$tfile + sleep 2 + local mtime0=`stat -c %Y $DIR1/$tfile` + + touch -m -d @$TEST_39_MTIME $DIR1/$tfile + local mtime1=`stat -c %Y $DIR1/$tfile` + [ "$mtime1" = $TEST_39_MTIME ] || \ + error "mtime is not set to past: $mtime1, should be $TEST_39_MTIME" + + local d1=`date +%s` + echo hello >> $DIR1/$tfile + local d2=`date +%s` + local mtime2=`stat -c %Y $DIR1/$tfile` + [ "$mtime2" -ge "$d1" ] && [ "$mtime2" -le "$d2" ] || \ + error "mtime is not updated on write: $d1 <= $mtime2 <= $d2" + + mv $DIR1/$tfile $DIR1/$tfile-1 + + for (( i=0; i < 2; i++ )) ; do + local mtime3=`stat -c %Y $DIR1/$tfile-1` + [ "$mtime2" = "$mtime3" ] || \ + error "mtime ($mtime2) changed (to $mtime3) on rename" + + cancel_lru_locks osc + if [ $i = 0 ] ; then echo "repeat after cancel_lru_locks"; fi + done +} +run_test 39c "mtime change on rename ===========================" + +# bug 21114 +test_39d() { + touch $DIR1/$tfile + + touch -m -d @$TEST_39_MTIME $DIR1/$tfile + + for (( i=0; i < 2; i++ )) ; do + local mtime=`stat -c %Y $DIR1/$tfile` + [ $mtime = $TEST_39_MTIME ] || \ + error "mtime($mtime) is not set to $TEST_39_MTIME" + + cancel_lru_locks osc + if [ $i = 0 ] ; then echo "repeat after cancel_lru_locks"; fi + done +} +run_test 39d "create, utime, stat ==============================" + +# bug 21114 +test_39e() { + touch $DIR1/$tfile + local mtime1=`stat -c %Y $DIR1/$tfile` + + touch -m -d @$TEST_39_MTIME $DIR1/$tfile + + for (( i=0; i < 2; i++ )) ; do + local mtime2=`stat -c %Y $DIR1/$tfile` + [ $mtime2 = $TEST_39_MTIME ] || \ + error "mtime($mtime2) is not set to $TEST_39_MTIME" + + cancel_lru_locks osc + if [ $i = 0 ] ; then echo "repeat after cancel_lru_locks"; fi + done +} +run_test 39e "create, stat, utime, stat ========================" + +# bug 21114 +test_39f() { + touch $DIR1/$tfile + mtime1=`stat -c %Y $DIR1/$tfile` + + sleep 2 + touch -m -d @$TEST_39_MTIME $DIR1/$tfile + + for (( i=0; i < 2; i++ )) ; do + local mtime2=`stat -c %Y $DIR1/$tfile` + [ $mtime2 = $TEST_39_MTIME ] || \ + error "mtime($mtime2) is not set to $TEST_39_MTIME" + + cancel_lru_locks osc + if [ $i = 0 ] ; then echo "repeat after cancel_lru_locks"; fi + done +} +run_test 39f "create, stat, sleep, utime, stat =================" + +# bug 11063 +test_39g() { + echo hello >> $DIR1/$tfile + local mtime1=`stat -c %Y $DIR1/$tfile` + + sleep 2 + chmod o+r $DIR1/$tfile + + for (( i=0; i < 2; i++ )) ; do + local mtime2=`stat -c %Y $DIR1/$tfile` + [ "$mtime1" = "$mtime2" ] || \ + error "lost mtime: $mtime2, should be $mtime1" + + cancel_lru_locks osc + if [ $i = 0 ] ; then echo "repeat after cancel_lru_locks"; fi + done +} +run_test 39g "write, chmod, stat ===============================" + +# bug 11063 +test_39h() { + touch $DIR1/$tfile + sleep 1 + + local d1=`date` + echo hello >> $DIR1/$tfile + local mtime1=`stat -c %Y $DIR1/$tfile` + + touch -m -d @$TEST_39_MTIME $DIR1/$tfile + local d2=`date` + if [ "$d1" != "$d2" ]; then + echo "write and touch not within one second" + else + for (( i=0; i < 2; i++ )) ; do + local mtime2=`stat -c %Y $DIR1/$tfile` + [ "$mtime2" = $TEST_39_MTIME ] || \ + error "lost mtime: $mtime2, should be $TEST_39_MTIME" + + cancel_lru_locks osc + if [ $i = 0 ] ; then echo "repeat after cancel_lru_locks"; fi + done + fi +} +run_test 39h "write, utime within one second, stat =============" + +test_39i() { + touch $DIR1/$tfile + sleep 1 + + echo hello >> $DIR1/$tfile + local mtime1=`stat -c %Y $DIR1/$tfile` + + mv $DIR1/$tfile $DIR1/$tfile-1 + + for (( i=0; i < 2; i++ )) ; do + local mtime2=`stat -c %Y $DIR1/$tfile-1` + + [ "$mtime1" = "$mtime2" ] || \ + error "lost mtime: $mtime2, should be $mtime1" + + cancel_lru_locks osc + if [ $i = 0 ] ; then echo "repeat after cancel_lru_locks"; fi + done +} +run_test 39i "write, rename, stat ==============================" + +test_39j() { + touch $DIR1/$tfile + sleep 1 + + multiop_bg_pause $DIR1/$tfile oO_RDWR:w2097152_c || error "multiop failed" + local multipid=$! + local mtime1=`stat -c %Y $DIR1/$tfile` + + mv $DIR1/$tfile $DIR1/$tfile-1 + + kill -USR1 $multipid + wait $multipid || error "multiop close failed" + + for (( i=0; i < 2; i++ )) ; do + local mtime2=`stat -c %Y $DIR1/$tfile-1` + [ "$mtime1" = "$mtime2" ] || \ + error "mtime is lost on close: $mtime2, should be $mtime1" + + cancel_lru_locks osc + if [ $i = 0 ] ; then echo "repeat after cancel_lru_locks"; fi + done +} +run_test 39j "write, rename, close, stat =======================" + +test_39k() { + touch $DIR1/$tfile + sleep 1 + + multiop_bg_pause $DIR1/$tfile oO_RDWR:w2097152_c || error "multiop failed" + local multipid=$! + local mtime1=`stat -c %Y $DIR1/$tfile` + + touch -m -d @$TEST_39_MTIME $DIR1/$tfile + + kill -USR1 $multipid + wait $multipid || error "multiop close failed" + + for (( i=0; i < 2; i++ )) ; do + local mtime2=`stat -c %Y $DIR1/$tfile` + + [ "$mtime2" = $TEST_39_MTIME ] || \ + error "mtime is lost on close: $mtime2, should be $TEST_39_MTIME" + + cancel_lru_locks osc + if [ $i = 0 ] ; then echo "repeat after cancel_lru_locks"; fi + done +} +run_test 39k "write, utime, close, stat ========================" + +# this should be set to future +TEST_39_ATIME=`date -d "1 year" +%s` + +test_39l() { + local atime_diff=$(do_facet $SINGLEMDS lctl get_param -n mdd.*.atime_diff) + + mkdir -p $DIR/$tdir + + # test setting directory atime to future + touch -a -d @$TEST_39_ATIME $DIR/$tdir + local atime=$(stat -c %X $DIR/$tdir) + [ "$atime" = $TEST_39_ATIME ] || \ + error "atime is not set to future: $atime, should be $TEST_39_ATIME" + + # test setting directory atime from future to now + local d1=$(date +%s) + ls $DIR/$tdir + local d2=$(date +%s) + + cancel_lru_locks mdc + atime=$(stat -c %X $DIR/$tdir) + [ "$atime" -ge "$d1" -a "$atime" -le "$d2" ] || \ + error "atime is not updated from future: $atime, should be $d1 dir atime + atime_diff + d1=$(date +%s) + ls $DIR/$tdir + d2=$(date +%s) + cancel_lru_locks mdc + atime=$(stat -c %X $DIR/$tdir) + [ "$atime" -ge "$d1" -a "$atime" -le "$d2" ] || \ + error "atime is not updated : $atime, should be $d2" + + do_facet $SINGLEMDS lctl set_param -n mdd.*.atime_diff=60 + sleep 3 + + # test not setting directory atime when now < dir atime + atime_diff + ls $DIR/$tdir + cancel_lru_locks mdc + atime=$(stat -c %X $DIR/$tdir) + [ "$atime" -ge "$d1" -a "$atime" -le "$d2" ] || \ + error "atime is updated to $atime, should remain $d1/dev/null + done + $LCTL get_param $proc_osc0/cur_dirty_bytes + $LCTL get_param $proc_osc0/cur_grant_bytes + + # perform the real test + $LCTL set_param $proc_osc0/rpc_stats 0 + for ((;i<$files; i++)); do + [ $($LFS getstripe -i $TDIR/f$i) -eq 0 ] || continue + dd if=/dev/zero of=$TDIR/f$i bs=$pagesz count=$pages 2>/dev/null + done + sync + $LCTL get_param $proc_osc0/rpc_stats + + $LCTL get_param $proc_osc0/rpc_stats | + while read PPR RRPC RPCT RCUM BAR WRPC WPCT WCUM; do + [ "$PPR" != "16:" ] && continue + [ $WPCT -lt 85 ] && error "$pages-page write RPCs only $WPCT% < 85%" + break # we only want the "pages per rpc" stat + done + rm -rf $TDIR +} +run_test 42e "verify sub-RPC writes are not done synchronously" + test_43() { mkdir -p $DIR/$tdir cp -p /bin/ls $DIR/$tdir/$tfile @@ -2086,7 +2663,7 @@ test_43b() { cp -p `which multiop` $DIR/d43/multiop || cp -p multiop $DIR/d43/multiop MULTIOP_PROG=$DIR/d43/multiop multiop_bg_pause $TMP/test43.junk O_c || return 1 MULTIOP_PID=$! - truncate $DIR/d43/multiop 0 && error "expected error, got success" + $TRUNCATE $DIR/d43/multiop 0 && error "expected error, got success" kill -USR1 $MULTIOP_PID || return 2 wait $MULTIOP_PID || return 3 rm $TMP/test43.junk @@ -2181,10 +2758,6 @@ test_45() { } run_test 45 "osc io page accounting ============================" -page_size() { - getconf PAGE_SIZE -} - # in a 2 stripe file (lov.sh), page 1023 maps to page 511 in its object. this # test tickles a bug where re-dirtying a page was failing to be mapped to the # objects offset and an assert hit when an rpc was built with 1023's mapped @@ -2454,7 +3027,7 @@ test_52a() { link $DIR/d52a/foo $DIR/d52a/foo_link 2>/dev/null && error "link worked" echo foo >> $DIR/d52a/foo || error "append foo failed" mrename $DIR/d52a/foo $DIR/d52a/foo_ren && error "rename worked" - lsattr $DIR/d52a/foo | egrep -q "^-+[ae]-+ $DIR/d52a/foo" || error "lsattr" + lsattr $DIR/d52a/foo | egrep -q "^-+a[-e]+ $DIR/d52a/foo" || error "lsattr" chattr -a $DIR/d52a/foo || error "chattr -a failed" cp -r $DIR/d52a /tmp/ rm -fr $DIR/d52a || error "cleanup rm failed" @@ -2474,7 +3047,7 @@ test_52b() { mrename $DIR/d52b/foo $DIR/d52b/foo_ren && error "rename worked" [ -f $DIR/d52b/foo ] || error [ -f $DIR/d52b/foo_ren ] && error - lsattr $DIR/d52b/foo | egrep -q "^-+[ie]-+ $DIR/d52b/foo" || error "lsattr" + lsattr $DIR/d52b/foo | egrep -q "^-+i[-e]+ $DIR/d52b/foo" || error "lsattr" chattr -i $DIR/d52b/foo || error "chattr failed" rm -fr $DIR/d52b || error @@ -2492,7 +3065,8 @@ test_53() { local ostnum # only test MDT0000 - for value in $(do_facet $SINGLEMDS lctl get_param osc.*-osc-MDT0000.prealloc_last_id) ; do + local mdtosc=$(get_mdtosc_proc_path $SINGLEMDS) + for value in $(do_facet $SINGLEMDS lctl get_param osc.$mdtosc.prealloc_last_id) ; do param=`echo ${value[0]} | cut -d "=" -f1` ostname=`echo $param | cut -d "." -f2 | cut -d - -f 1-2` mds_last=$(do_facet $SINGLEMDS lctl get_param -n $param) @@ -2576,21 +3150,10 @@ test_54e() { } run_test 54e "console/tty device works in lustre ======================" -check_fstype() { - grep -q $FSTYPE /proc/filesystems && return 1 - modprobe $FSTYPE - grep -q $FSTYPE /proc/filesystems && return 1 - insmod ../$FSTYPE/$FSTYPE.o - grep -q $FSTYPE /proc/filesystems && return 1 - insmod ../$FSTYPE/$FSTYPE.ko - grep -q $FSTYPE /proc/filesystems && return 1 - return 0 -} - test_55() { rm -rf $DIR/d55 mkdir $DIR/d55 - check_fstype && skip_env "can't find fs $FSTYPE" && return + client_only && skip "Not a server" && return mount -t $FSTYPE -o loop,iopen $EXT2_DEV $DIR/d55 || error "mounting" touch $DIR/d55/foo $IOPENTEST1 $DIR/d55/foo $DIR/d55 || error "running $IOPENTEST1" @@ -2641,12 +3204,14 @@ test_56a() { # was test_56 [ "$OSTCOUNT" -lt 2 ] && \ skip_env "skipping other lfs getstripe --obd test" && return - FILENUM=`$GETSTRIPE --recursive $DIR/d56 | sed -n '/^[ ]*1[ ]/p' | wc -l` - OBDUUID=`$GETSTRIPE --recursive $DIR/d56 | sed -n '/^[ ]*1:/p' | awk '{print $2}'` - FOUND=`$GETSTRIPE -r --obd $OBDUUID $DIR/d56 | wc -l` + OSTIDX=1 + OBDUUID=$(lfs osts | grep ${OSTIDX}": " | awk '{print $2}') + FILENUM=`$GETSTRIPE -ir $DIR/d56 | grep -x $OSTIDX | wc -l` + FOUND=`$GETSTRIPE -r --obd $OBDUUID $DIR/d56 | grep obdidx | wc -l` [ $FOUND -eq $FILENUM ] || \ error "lfs getstripe --obd wrong: found $FOUND, expected $FILENUM" - [ `$GETSTRIPE -r -v --obd $OBDUUID $DIR/d56 | sed '/^[ ]*1[ ]/d' |\ + [ `$GETSTRIPE -r -v --obd $OBDUUID $DIR/d56 | \ + sed '/^[ ]*'${OSTIDX}'[ ]/d' |\ sed -n '/^[ ]*[0-9][0-9]*[ ]/p' | wc -l` -eq 0 ] || \ error "lfs getstripe --obd wrong: should not show file on other obd" echo "lfs getstripe --obd passed." @@ -2726,8 +3291,8 @@ run_test 56h "check lfs find ! -name =============================" test_56i() { tdir=${tdir}i mkdir -p $DIR/$tdir - UUID=`$GETSTRIPE $DIR/$tdir | awk '/0: / { print $2 }'` - OUT="`$LFIND -ost $UUID $DIR/$tdir`" + UUID=$(ostuuid_from_index 0 $DIR/$tdir) + OUT=$($LFIND -obd $UUID $DIR/$tdir) [ "$OUT" ] && error "$LFIND returned directory '$OUT'" || true } run_test 56i "check 'lfs find -ost UUID' skips directories =======" @@ -2830,14 +3395,14 @@ test_56q() { setup_56 $NUMFILES $NUMDIRS - chgrp $RUNAS_ID $TDIR/file* || error "chown $DIR/${tdir}g/file$i failed" + chgrp $RUNAS_GID $TDIR/file* || error "chown $DIR/${tdir}g/file$i failed" EXPECTED=$NUMFILES - NUMS="`$LFIND -gid $RUNAS_ID $TDIR | wc -l`" + NUMS="`$LFIND -gid $RUNAS_GID $TDIR | wc -l`" [ $NUMS -eq $EXPECTED ] || \ error "lfs find -gid $TDIR wrong: found $NUMS, expected $EXPECTED" EXPECTED=$(( ($NUMFILES+1) * $NUMDIRS + 1)) - NUMS="`$LFIND ! -gid $RUNAS_ID $TDIR | wc -l`" + NUMS="`$LFIND ! -gid $RUNAS_GID $TDIR | wc -l`" [ $NUMS -eq $EXPECTED ] || \ error "lfs find ! -gid $TDIR wrong: found $NUMS, expected $EXPECTED" @@ -2845,11 +3410,43 @@ test_56q() { } run_test 56q "check lfs find -gid and ! -gid ===============================" +test_56r() { + setup_56 $NUMFILES $NUMDIRS + TDIR=$DIR/${tdir}g + + EXPECTED=12 + NUMS=`$LFIND -size 0 -t f $TDIR | wc -l` + [ $NUMS -eq $EXPECTED ] || \ + error "lfs find $TDIR -size 0 wrong: found $NUMS, expected $EXPECTED" + EXPECTED=0 + NUMS=`$LFIND ! -size 0 -t f $TDIR | wc -l` + [ $NUMS -eq $EXPECTED ] || \ + error "lfs find $TDIR ! -size 0 wrong: found $NUMS, expected $EXPECTED" + echo "test" > $TDIR/56r && sync + EXPECTED=1 + NUMS=`$LFIND -size 5 -t f $TDIR | wc -l` + [ $NUMS -eq $EXPECTED ] || \ + error "lfs find $TDIR -size 5 wrong: found $NUMS, expected $EXPECTED" + EXPECTED=1 + NUMS=`$LFIND -size +5 -t f $TDIR | wc -l` + [ $NUMS -eq $EXPECTED ] || \ + error "lfs find $TDIR -size +5 wrong: found $NUMS, expected $EXPECTED" + EXPECTED=13 + NUMS=`$LFIND -size +0 -t f $TDIR | wc -l` + [ $NUMS -eq $EXPECTED ] || \ + error "lfs find $TDIR -size +0 wrong: found $NUMS, expected $EXPECTED" + EXPECTED=0 + NUMS=`$LFIND ! -size -5 -t f $TDIR | wc -l` + [ $NUMS -eq $EXPECTED ] || \ + error "lfs find $TDIR ! -size -5 wrong: found $NUMS, expected $EXPECTED" +} + +run_test 56r "check lfs find -size works ==========================" + test_57a() { # note test will not do anything if MDS is not local remote_mds_nodsh && skip "remote MDS with nodsh" && return - - local MNTDEV="osd.*MDT*.mntdev" + local MNTDEV="osd*.*MDT*.mntdev" DEV=$(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 @@ -3182,7 +3779,7 @@ test_65k() { # bug11679 STRIPE_INDEX=`do_facet $SINGLEMDS lctl get_param -n lov.*md*.target_obd | grep $STRIPE_OST | awk -F: '{print $1}' | head -n 1` - [ -f $DIR/$tdir/${STRIPE_INDEX} ] && continue + [ -f $DIR/$tdir/${STRIPE_INDEX} ] && continue echo "$SETSTRIPE $DIR/$tdir/${STRIPE_INDEX} -i ${STRIPE_INDEX} -c 1" do_facet client $SETSTRIPE $DIR/$tdir/${STRIPE_INDEX} -i ${STRIPE_INDEX} -c 1 RC=$? @@ -3213,6 +3810,7 @@ test_66() { run_test 66 "update inode blocks count on client ===============" LLOOP= +LLITELOOPLOAD= cleanup_68() { trap 0 if [ ! -z "$LLOOP" ]; then @@ -3224,6 +3822,10 @@ cleanup_68() { rm -f $LLOOP unset LLOOP fi + if [ ! -z "$LLITELOOPLOAD" ]; then + rmmod llite_lloop + unset LLITELOOPLOAD + fi rm -f $DIR/f68* } @@ -3239,15 +3841,21 @@ swap_used() { test_68a() { [ "$UID" != 0 ] && skip_env "must run as root" && return - grep -q llite_lloop /proc/modules - [ $? -ne 0 ] && skip_env "can't find module llite_lloop" && return + trap cleanup_68 EXIT + + if ! module_loaded llite_lloop; then + if load_module llite/llite_lloop; then + LLITELOOPLOAD=yes + else + skip_env "can't find module llite_lloop" + return + fi + fi LLOOP=$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" - 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" @@ -3263,7 +3871,7 @@ test_68b() { # was test_68 skip "local OST" && return grep -q llite_lloop /proc/modules - [ $? -ne 0 ] && skip_env "can't find module llite_lloop" && return + [ $? -ne 0 ] && skip "can't find module llite_lloop" && return [ -z "`$LCTL list_nids | grep -v tcp`" ] && \ skip "can't reliably test swap with TCP" && return @@ -3304,7 +3912,7 @@ test_69() { $DIRECTIO write ${f}.2 0 1 || error "directio write error" do_facet ost1 lctl set_param fail_loc=0x217 - truncate $f 1 # vmtruncate() will ignore truncate() error. + $TRUNCATE $f 1 # vmtruncate() will ignore truncate() error. $DIRECTIO write $f 0 2 && error "write succeeded, expect -ENOENT" do_facet ost1 lctl set_param fail_loc=0 @@ -3332,7 +3940,7 @@ test_72() { # bug 5695 - Test that on 2.6 remove_suid works properly [ "$RUNAS_ID" = "$UID" ] && skip_env "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_ID $RUNAS || { + FAIL_ON_ERROR=false check_runas_id_ret $RUNAS_ID $RUNAS_GID $RUNAS || { skip_env "User $RUNAS_ID does not exist - skipping" return 0 } @@ -3409,121 +4017,21 @@ test_74b() { # bug 13310 } run_test 74b "ldlm_enqueue freed-export error path, touch (shouldn't LBUG)" -JOIN=${JOIN:-"lfs join"} -F75=$DIR/f75 -F128k=${F75}_128k -FHEAD=${F75}_head -FTAIL=${F75}_tail -export T75_PREP=no -test75_prep() { - [ $T75_PREP = "yes" ] && return - echo "using F75=$F75, F128k=$F128k, FHEAD=$FHEAD, FTAIL=$FTAIL" - - dd if=/dev/urandom of=${F75}_128k bs=128k count=1 || error "dd failed" - log "finished dd" - chmod 777 ${F128k} - T75_PREP=yes -} - -test_75a() { - test75_prep - - cp -p ${F128k} ${FHEAD} - log "finished cp to $FHEAD" - cp -p ${F128k} ${FTAIL} - log "finished cp to $FTAIL" - cat ${F128k} ${F128k} > ${F75}_sim_sim - - $JOIN ${FHEAD} ${FTAIL} || error "join ${FHEAD} ${FTAIL} error" - log "finished join $FHEAD to ${F75}_sim_sim" - cmp ${FHEAD} ${F75}_sim_sim || error "${FHEAD} ${F75}_sim_sim differ" - log "finished cmp $FHEAD to ${F75}_sim_sim" - $CHECKSTAT -a ${FTAIL} || error "tail ${FTAIL} still exist after join" -} -run_test 75a "TEST join file ====================================" - -test_75b() { - test75_prep - - cp -p ${F128k} ${FTAIL} - cat ${F75}_sim_sim >> ${F75}_join_sim - cat ${F128k} >> ${F75}_join_sim - $JOIN ${FHEAD} ${FTAIL} || error "join ${FHEAD} ${FTAIL} error" - cmp ${FHEAD} ${F75}_join_sim || \ - error "${FHEAD} ${F75}_join_sim are different" - $CHECKSTAT -a ${FTAIL} || error "tail ${FTAIL} exist after join" -} -run_test 75b "TEST join file 2 ==================================" - -test_75c() { - test75_prep - - cp -p ${F128k} ${FTAIL} - cat ${F128k} >> ${F75}_sim_join - cat ${F75}_join_sim >> ${F75}_sim_join - $JOIN ${FTAIL} ${FHEAD} || error "join error" - cmp ${FTAIL} ${F75}_sim_join || \ - error "${FTAIL} ${F75}_sim_join are different" - $CHECKSTAT -a ${FHEAD} || error "tail ${FHEAD} exist after join" -} -run_test 75c "TEST join file 3 ==================================" - -test_75d() { - test75_prep - - cp -p ${F128k} ${FHEAD} - cp -p ${F128k} ${FHEAD}_tmp - cat ${F75}_sim_sim >> ${F75}_join_join - cat ${F75}_sim_join >> ${F75}_join_join - $JOIN ${FHEAD} ${FHEAD}_tmp || error "join ${FHEAD} ${FHEAD}_tmp error" - $JOIN ${FHEAD} ${FTAIL} || error "join ${FHEAD} ${FTAIL} error" - cmp ${FHEAD} ${F75}_join_join ||error "${FHEAD} ${F75}_join_join differ" $CHECKSTAT -a ${FHEAD}_tmp || error "${FHEAD}_tmp exist after join" - $CHECKSTAT -a ${FTAIL} || error "tail ${FTAIL} exist after join (2)" -} -run_test 75d "TEST join file 4 ==================================" - -test_75e() { - test75_prep - - rm -rf ${FHEAD} || "delete join file error" -} -run_test 75e "TEST join file 5 (remove joined file) =============" - -test_75f() { - test75_prep - - cp -p ${F128k} ${F75}_join_10_compare - cp -p ${F128k} ${F75}_join_10 - for ((i = 0; i < 10; i++)); do - cat ${F128k} >> ${F75}_join_10_compare - cp -p ${F128k} ${FTAIL} - $JOIN ${F75}_join_10 ${FTAIL} || \ - error "join ${F75}_join_10 ${FTAIL} error" - $CHECKSTAT -a ${FTAIL} || error "tail file exist after join" - done - cmp ${F75}_join_10 ${F75}_join_10_compare || \ - error "files ${F75}_join_10 ${F75}_join_10_compare differ" -} -run_test 75f "TEST join file 6 (join 10 files) ==================" - -test_75g() { - [ ! -f ${F75}_join_10 ] && echo "${F75}_join_10 missing" && return - $LFS getstripe ${F75}_join_10 - - $OPENUNLINK ${F75}_join_10 ${F75}_join_10 || error "files unlink open" - - ls -l $F75* +test_74c() { +#define OBD_FAIL_LDLM_NEW_LOCK + lctl set_param fail_loc=0x80000319 + touch $DIR/$tfile && error "Touch successful" + true } -run_test 75g "TEST join file 7 (open unlink) ====================" +run_test 74c "ldlm_lock_create error path, (shouldn't LBUG)" num_inodes() { awk '/lustre_inode_cache/ {print $2; exit}' /proc/slabinfo } -test_76() { # bug 1443 - DETH=$(grep deathrow /proc/kallsyms /proc/ksyms 2> /dev/null | wc -l) - [ $DETH -eq 0 ] && skip "No _iget." && return 0 - BEFORE_INODES=`num_inodes` +test_76() { # Now for bug 20433, added originally in bug 1443 + cancel_lru_locks osc + BEFORE_INODES=`num_inodes` echo "before inodes: $BEFORE_INODES" local COUNT=1000 [ "$SLOW" = "no" ] && COUNT=100 @@ -3531,13 +4039,22 @@ test_76() { # bug 1443 touch $DIR/$tfile rm -f $DIR/$tfile done + cancel_lru_locks osc AFTER_INODES=`num_inodes` echo "after inodes: $AFTER_INODES" - [ $AFTER_INODES -gt $((BEFORE_INODES + 32)) ] && \ - error "inode slab grew from $BEFORE_INODES to $AFTER_INODES" - true + local wait=0 + while [ $AFTER_INODES -gt $BEFORE_INODES ]; do + sleep 2 + AFTER_INODES=`num_inodes` + wait=$((wait+2)) + echo "wait $wait seconds inodes: $AFTER_INODES" + if [ $wait -gt 30 ]; then + error "inode slab grew from $BEFORE_INODES to $AFTER_INODES" + fi + done } -run_test 76 "destroy duplicate inodes in client inode cache ====" +run_test 76 "confirm clients recycle inodes properly ====" + export ORIG_CSUM="" set_checksums() @@ -3724,7 +4241,9 @@ rm -f $F77_TMP unset F77_TMP test_78() { # bug 10901 - NSEQ=5 + remote_ost || { skip_env "local OST" && return; } + + NSEQ=5 F78SIZE=$(($(awk '/MemFree:/ { print $2 }' /proc/meminfo) / 1024)) echo "MemFree: $F78SIZE, Max file size: $MAXFREE" MEMTOTAL=$(($(awk '/MemTotal:/ { print $2 }' /proc/meminfo) / 1024)) @@ -3953,28 +4472,23 @@ test_101() { } run_test 101 "check read-ahead for random reads ================" -export SETUP_TEST101=no -setup_test101() { - [ "$SETUP_TEST101" = "yes" ] && return +setup_test101b() { mkdir -p $DIR/$tdir STRIPE_SIZE=1048576 STRIPE_COUNT=$OSTCOUNT STRIPE_OFFSET=0 - trap cleanup_test101 EXIT + trap cleanup_test101b EXIT # prepare the read-ahead file $SETSTRIPE $DIR/$tfile -s $STRIPE_SIZE -i $STRIPE_OFFSET -c $OSTCOUNT dd if=/dev/zero of=$DIR/$tfile bs=1024k count=100 2> /dev/null - SETUP_TEST101=yes } -cleanup_test101() { - [ "$SETUP_TEST101" = "yes" ] || return +cleanup_test101b() { trap 0 rm -rf $DIR/$tdir - rm -f $DIR/$tfile - SETUP_TEST101=no + rm -f $DIR/$tfile } calc_total() { @@ -4008,7 +4522,7 @@ test_101b() { local FILE_LENGTH=$((STRIPE_SIZE*100)) local ITERATION=$((FILE_LENGTH/STRIDE_SIZE)) # prepare the read-ahead file - setup_test101 + setup_test101b cancel_lru_locks osc for BIDX in 2 4 8 16 32 64 128 256 do @@ -4022,7 +4536,7 @@ test_101b() { cancel_lru_locks osc ra_check_101 $BSIZE done - cleanup_test101 + cleanup_test101b true } run_test 101b "check stride-io mode read-ahead =================" @@ -4072,9 +4586,7 @@ test_101d() { } run_test 101d "file read with and without read-ahead enabled =================" -export SETUP_TEST102=no setup_test102() { - [ "$SETUP_TEST102" = "yes" ] && return mkdir -p $DIR/$tdir chown $RUNAS_ID $DIR/$tdir STRIPE_SIZE=65536 @@ -4101,14 +4613,12 @@ setup_test102() { cd $DIR $1 $TAR cf $TMP/f102.tar $tdir --xattrs - SETUP_TEST102=yes } cleanup_test102() { trap 0 - [ "$SETUP_TEST102" = "yes" ] || return 0 rm -f $TMP/f102.tar - SETUP_TEST102=no + rm -rf $DIR/d0.sanity/d102 } test_102a() { @@ -4361,6 +4871,28 @@ test_102j() { } run_test 102j "non-root tar restore stripe info from tarfile, not keep osts ===" +test_102k() { + touch $DIR/$tfile + # b22187 just check that does not crash for regular file. + setfattr -n trusted.lov $DIR/$tfile + # b22187 'setfattr -n trusted.lov' should work as remove LOV EA for directories + local test_kdir=$DIR/d102k + mkdir $test_kdir + local default_size=`$GETSTRIPE -s $test_kdir` + local default_count=`$GETSTRIPE -c $test_kdir` + local default_offset=`$GETSTRIPE -o $test_kdir` + $SETSTRIPE -s 65536 -i 1 -c $OSTCOUNT $test_kdir || error 'dir setstripe failed' + setfattr -n trusted.lov $test_kdir + local stripe_size=`$GETSTRIPE -s $test_kdir` + local stripe_count=`$GETSTRIPE -c $test_kdir` + local stripe_offset=`$GETSTRIPE -o $test_kdir` + [ $stripe_size -eq $default_size ] || error "stripe size $stripe_size != $default_size" + [ $stripe_count -eq $default_count ] || error "stripe count $stripe_count != $default_count" + [ $stripe_offset -eq $default_offset ] || error "stripe offset $stripe_offset != $default_offset" + rm -rf $DIR/$tfile $test_kdir +} +run_test 102k "setfattr without parameter of value shouldn't cause a crash" + cleanup_test102 run_acl_subtest() @@ -4414,7 +4946,7 @@ test_103 () { } run_test 103 "acl test =========================================" -test_104() { +test_104a() { touch $DIR/$tfile lfs df || error "lfs df failed" lfs df -ih || error "lfs df -ih failed" @@ -4430,7 +4962,18 @@ test_104() { lfs df || error "lfs df with reactivated OSC failed" rm -f $DIR/$tfile } -run_test 104 "lfs df [-ih] [path] test =========================" +run_test 104a "lfs df [-ih] [path] test =========================" + +test_104b() { + [ $RUNAS_ID -eq $UID ] && skip_env "RUNAS_ID = UID = $UID -- skipping" && return + chmod 666 /dev/obd + denied_cnt=$((`$RUNAS $LFS check servers 2>&1 | grep "Permission denied" | wc -l`)) + if [ $denied_cnt -ne 0 ]; + then + error "lfs check servers test failed" + fi +} +run_test 104b "$RUNAS lfs check servers test ====================" test_105a() { # doesn't work on 2.4 kernels @@ -4583,7 +5126,6 @@ test_116() { echo -n "Free space priority " lctl get_param -n lov.*-clilov-*.qos_prio_free - DELAY=$(lctl get_param -n lov.*-clilov-*.qos_maxage | head -1 | awk '{print $1}') declare -a AVAIL free_min_max [ $MINV -gt 960000 ] && skip "too much free space in OST$MINI, skip" &&\ @@ -4604,7 +5146,7 @@ test_116() { done FILL=$(($MINV / 4)) sync - sleep $DELAY + sleep_maxage free_min_max DIFF=$(($MAXV - $MINV)) @@ -4633,7 +5175,7 @@ test_116() { done echo "wrote $i 200k files" sync - sleep $DELAY + sleep_maxage echo "Note: free space may not be updated, so measurements might be off" free_min_max @@ -4660,7 +5202,7 @@ test_116() { [ $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} + rm -rf $DIR/$tdir } run_test 116 "stripe QOS: free space balance ===================" @@ -5018,16 +5560,17 @@ test_118k() set_nodes_failloc "$(osts_nodes)" 0x20e mkdir -p $DIR/$tdir - for ((i=0;i<10;i++)); do - (dd if=/dev/zero of=$DIR/$tdir/$tfile-$i bs=1M count=10 || \ + for ((i=0;i<10;i++)); do + (dd if=/dev/zero of=$DIR/$tdir/$tfile-$i bs=1M count=10 || \ error "dd to $DIR/$tdir/$tfile-$i failed" )& - SLEEPPID=$! - sleep 0.500s - kill $SLEEPPID - wait $SLEEPPID - done + SLEEPPID=$! + sleep 0.500s + kill $SLEEPPID + wait $SLEEPPID + done - set_nodes_failloc "$(osts_nodes)" 0 + set_nodes_failloc "$(osts_nodes)" 0 + rm -rf $DIR/$tdir } run_test 118k "bio alloc -ENOMEM and IO TERM handling =========" @@ -5292,8 +5835,8 @@ run_test 121 "read cancel race =========" test_123a() { # was test 123, statahead(bug 11401) SLOWOK=0 if [ -z "$(grep "processor.*: 1" /proc/cpuinfo)" ]; then - log "testing on UP system. Performance may be not as good as expected." - SLOWOK=1 + log "testing on UP system. Performance may be not as good as expected." + SLOWOK=1 fi rm -rf $DIR/$tdir @@ -5325,49 +5868,32 @@ test_123a() { # was test 123, statahead(bug 11401) etime=`date +%s` delta_sa=$((etime - stime)) log "ls $i files with statahead: $delta_sa sec" - lctl get_param -n llite.*.statahead_stats + 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 - + [ $swrong -lt $ewrong ] && log "statahead was stopped, maybe too many locks held!" [ $delta -eq 0 -o $delta_sa -eq 0 ] && continue if [ $((delta_sa * 100)) -gt $((delta * 105)) -a $delta_sa -gt $((delta + 2)) ]; then + 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` + time ls -l $DIR/$tdir | wc -l + etime=`date +%s` + delta=$((etime - stime)) + log "ls $i files again without statahead: $delta sec" + lctl set_param llite.*.statahead_max=$max + 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 fi [ $delta -gt 20 ] && break @@ -5642,49 +6168,48 @@ test_128() { # bug 15212 run_test 128 "interactive lfs for 2 consecutive find's" set_dir_limits () { - local mntdev - local node + local mntdev + local canondev + local node local LDPROC=/proc/fs/ldiskfs - for node in $(mdts_nodes); do - devs=$(do_node $node "lctl get_param -n devices" | awk '($3 ~ "mdt" && $4 ~ "MDT") { print $4 }') - for dev in $devs; do - mntdev=$(do_node $node "lctl get_param -n osd.$dev.mntdev") - do_node $node "echo $1 >$LDPROC/\\\$(basename $mntdev)/max_dir_size" - done + for facet in $(get_facets MDS); do + canondev=$(ldiskfs_canon *.$(convert_facet2label $facet).mntdev $facet) + do_facet $facet "test -e $LDPROC/$canondev/max_dir_size" || LDPROC=/sys/fs/ldiskfs + do_facet $facet "echo $1 >$LDPROC/$canondev/max_dir_size" done } test_129() { - [ "$FSTYPE" != "ldiskfs" ] && skip "not needed for FSTYPE=$FSTYPE" && return 0 - remote_mds_nodsh && skip "remote MDS with nodsh" && return + [ "$FSTYPE" != "ldiskfs" ] && skip "not needed for FSTYPE=$FSTYPE" && return 0 + remote_mds_nodsh && skip "remote MDS with nodsh" && return - EFBIG=27 - MAX=16384 + EFBIG=27 + MAX=16384 - set_dir_limits $MAX + set_dir_limits $MAX - mkdir -p $DIR/$tdir + mkdir -p $DIR/$tdir - I=0 - J=0 - while [ ! $I -gt $((MAX * MDSCOUNT)) ]; do - multiop $DIR/$tdir/$J Oc - rc=$? - if [ $rc -eq $EFBIG ]; then - set_dir_limits 0 - echo "return code $rc received as expected" - return 0 - elif [ $rc -ne 0 ]; then - set_dir_limits 0 - error_exit "return code $rc received instead of expected $EFBIG" - fi - J=$((J+1)) - I=$(stat -c%s "$DIR/$tdir") - done + I=0 + J=0 + while [ ! $I -gt $((MAX * MDSCOUNT)) ]; do + multiop $DIR/$tdir/$J Oc + rc=$? + if [ $rc -eq $EFBIG ]; then + set_dir_limits 0 + echo "return code $rc received as expected" + return 0 + elif [ $rc -ne 0 ]; then + set_dir_limits 0 + error_exit "return code $rc received instead of expected $EFBIG" + fi + J=$((J+1)) + I=$(stat -c%s "$DIR/$tdir") + done - error "exceeded dir size limit $MAX x $MDSCOUNT $((MAX * MDSCOUNT)) : $I bytes" - do_facet $SINGLEMDS "echo 0 >$LDPROC" + set_dir_limits 0 + error "exceeded dir size limit $MAX x $MDSCOUNT $((MAX * MDSCOUNT)) : $I bytes" } run_test 129 "test directory size limit ========================" @@ -6004,9 +6529,11 @@ som_mode_switch() { if [ x$som = x"enabled" ]; then [ $((gl2 - gl1)) -gt 0 ] && error "no glimpse RPC is expected" + MOUNTOPT=`echo $MOUNTOPT | sed 's/som_preview//g'` do_facet mgs "$LCTL conf_param $FSNAME.mdt.som=disabled" else [ $((gl2 - gl1)) -gt 0 ] || error "some glimpse RPC is expected" + MOUNTOPT="$MOUNTOPT,som_preview" do_facet mgs "$LCTL conf_param $FSNAME.mdt.som=enabled" fi @@ -6018,8 +6545,9 @@ som_mode_switch() { } test_132() { #1028, SOM - local num=$(get_mds_dir $DIR) - local mymds=mds${num} + local num=$(get_mds_dir $DIR) + local mymds=mds${num} + local MOUNTOPT_SAVE=$MOUNTOPT dd if=/dev/zero of=$DIR/$tfile count=1 2>/dev/null cancel_lru_locks osc @@ -6030,9 +6558,12 @@ test_132() { #1028, SOM 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 + rm $DIR/$tfile som_mode_switch $som1 $gl1 $gl2 + dd if=/dev/zero of=$DIR/$tfile count=1 2>/dev/null + cancel_lru_locks osc + som2=$(do_facet $mymds "$LCTL get_param mdt.*.som" | awk -F= ' {print $2}' | head -n 1) if [ $som1 == $som2 ]; then error "som is still "$som2 @@ -6046,8 +6577,9 @@ test_132() { #1028, SOM gl1=$(get_ost_param "ldlm_glimpse_enqueue") stat $DIR/$tfile >/dev/null gl2=$(get_ost_param "ldlm_glimpse_enqueue") - echo "SOM is "$som2", "$((gl2 - gl1))" glimpse RPC occured" + echo "====> SOM is "$som2", "$((gl2 - gl1))" glimpse RPC occured" som_mode_switch $som2 $gl1 $gl2 + MOUNTOPT=$MOUNTOPT_SAVE } run_test 132 "som avoids glimpse rpc" @@ -6089,6 +6621,7 @@ test_150() { cancel_lru_locks osc cmp $TF $DIR/$tfile || error "$TMP/$tfile $DIR/$tfile differ" remount_client $MOUNT + df -P $MOUNT cmp $TF $DIR/$tfile || error "$TF $DIR/$tfile differ (remount)" $TRUNCATE $TF 6000 @@ -6462,13 +6995,13 @@ test_160() { # verify contents echo "verifying target fid" fidc=$($LFS changelog $MDT0 | grep timestamp | grep "CREAT" | \ - tail -1 | awk '{print $5}') + tail -1 | awk '{print $6}') fidf=$($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}') + tail -1 | awk '{print $7}') fidf=$($LFS path2fid $DIR/$tdir/pics/zach) [ "$fidc" == "p=$fidf" ] || \ err17935 "pfid in changelog $fidc != dir fid $fidf" @@ -6532,7 +7065,7 @@ test_161() { rm $DIR/$tdir/$tfile # rename mv $DIR/$tdir/foo1/sofia $DIR/$tdir/foo2/maggie - if [ "$($LFS fid2path $DIR --link 1 $FID)" != "/$tdir/foo2/maggie" ] + if [ "$($LFS fid2path $FSNAME --link 1 $FID)" != "$tdir/foo2/maggie" ] then $LFS fid2path $DIR $FID err17935 "bad link rename" @@ -6566,42 +7099,47 @@ check_path() { err17935 "path looked up \"${path}\" instead of \"${expected}\"" return 2 fi - echo "fid $fid resolves to path $path" + echo "fid $fid resolves to path $path (expected $expected)" } 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 + # 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 # regular file - FID=$($LFS path2fid $DIR/$tdir/d2/$tfile | tr -d '[') - check_path "/$tdir/d2/$tfile" $DIR $FID --link 0 + FID=$($LFS path2fid $DIR/$tdir/d2/$tfile | tr -d '[]') + check_path "$tdir/d2/$tfile" $FSNAME $FID --link 0 # softlink - ln -s $DIR/$tdir/d2/$tfile $DIR/$tdir/d2/p/q/r/slink - FID=$($LFS path2fid $DIR/$tdir/d2/p/q/r/slink | tr -d '[') - check_path "/$tdir/d2/p/q/r/slink" $DIR $FID --link 0 + ln -s $DIR/$tdir/d2/$tfile $DIR/$tdir/d2/p/q/r/slink + FID=$($LFS path2fid $DIR/$tdir/d2/p/q/r/slink | tr -d '[]') + check_path "$tdir/d2/p/q/r/slink" $FSNAME $FID --link 0 + + # softlink to wrong file + ln -s /this/is/garbage $DIR/$tdir/d2/p/q/r/slink.wrong + FID=$($LFS path2fid $DIR/$tdir/d2/p/q/r/slink.wrong | tr -d '[]') + check_path "$tdir/d2/p/q/r/slink.wrong" $FSNAME $FID --link 0 # hardlink - 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 - - # hardlink count: check that there are 2 links - # Doesnt work with CMD yet: 17935 + 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 $FID --link 1 + check_path "$DIR/$tdir/d2/p/q/r/hlink" $DIR $FID --link 0 + + # hardlink count: check that there are 2 links + # Doesnt work with CMD yet: 17935 ${LFS} fid2path $DIR $FID | wc -l | grep -q 2 || \ err17935 "expected 2 links" # hardlink indexing: remove the first link - rm $DIR/$tdir/d2/p/q/r/hlink - check_path "/$tdir/d2/a/b/c/new_file" $DIR $FID --link 0 + rm $DIR/$tdir/d2/p/q/r/hlink + check_path "$tdir/d2/a/b/c/new_file" $FSNAME $FID --link 0 return 0 } @@ -6609,15 +7147,16 @@ run_test 162 "path lookup sanity" test_163() { remote_mds_nodsh && skip "remote MDS with nodsh" && return - copytool --test || { skip "copytool not runnable: $?" && return; } - copytool & + copytool --test $FSNAME || { skip "copytool not runnable: $?" && return; } + copytool $FSNAME & sleep 1 - local uuid=$($LCTL get_param -n mdc.lustre-MDT0000-mdc-*.uuid) + local uuid=$($LCTL get_param -n mdc.${FSNAME}-MDT0000-mdc-*.uuid) # this proc file is temporary and linux-only - do_facet mds lctl set_param mdt.lustre-MDT0000.mdccomm=$uuid || error "lnl send failed" - kill $! + do_facet $SINGLEMDS lctl set_param mdt.${FSNAME}-MDT0000.mdccomm=$uuid ||\ + error "kernel->userspace send failed" + kill -INT $! } -run_test 163 "LustreNetLink kernelcomms" +run_test 163 "kernel <-> userspace comms" test_169() { # do directio so as not to populate the page cache @@ -6687,6 +7226,100 @@ test_170() { } run_test 170 "test lctl df to handle corrupted log =====================" +test_171() { # bug20592 +#define OBD_FAIL_PTLRPC_DUMP_LOG 0x50e + $LCTL set_param fail_loc=0x50e + $LCTL set_param fail_val=3000 + multiop_bg_pause $DIR/$tfile Os || true + # cause log dump + sleep 3 + if dmesg | grep "recursive fault"; then + error "caught a recursive fault" + fi + $LCTL set_param fail_loc=0 + true +} +run_test 171 "test libcfs_debug_dumplog_thread stuck in do_exit() ======" + +# it would be good to share it with obdfilter-survey/libecho code +setup_obdecho_osc () { + local rc=0 + local ost_nid=$1 + local obdfilter_name=$2 + echo "Creating new osc for $obdfilter_name on $ost_nid" + [ $rc -eq 0 ] && { $LCTL attach osc ${obdfilter_name}_osc \ + ${obdfilter_name}_osc_UUID || rc=2; } + [ $rc -eq 0 ] && { $LCTL --device ${obdfilter_name}_osc setup \ + ${obdfilter_name}_UUID $ost_nid || rc=3; } + return $rc +} + +cleanup_obdecho_osc () { + local obdfilter_name=$1 + $LCTL --device ${obdfilter_name}_osc cleanup >/dev/null + $LCTL --device ${obdfilter_name}_osc detach >/dev/null + return 0 +} + +obdecho_create_test() { + local OBD=$1 + local node=$2 + local rc=0 + local id + do_facet $node "$LCTL attach echo_client ec ec_uuid" || rc=1 + [ $rc -eq 0 ] && { do_facet $node "$LCTL --device ec setup $OBD" || + rc=2; } + if [ $rc -eq 0 ]; then + id=$(do_facet $node "$LCTL --device ec create 1" | awk '/object id/ {print $6}') + [ ${PIPESTATUS[0]} -eq 0 -a -n "$id" ] || rc=3 + fi + echo "New object id is $id" + [ $rc -eq 0 ] && { do_facet $node "$LCTL --device ec test_brw 10 w v 64 $id" || + rc=4; } + [ $rc -eq 0 -o $rc -gt 2 ] && { do_facet $node "$LCTL --device ec " \ + "cleanup" || rc=5; } + [ $rc -eq 0 -o $rc -gt 1 ] && { do_facet $node "$LCTL --device ec " \ + "detach" || rc=6; } + [ $rc -ne 0 ] && echo "obecho_create_test failed: $rc" + return $rc +} + +test_180a() { + local rc=0 + local rmmod_local=0 + + if ! module_loaded obdecho; then + load_module obdecho/obdecho + rmmod_local=1 + fi + + local osc=$($LCTL dl | grep -v mdt | awk '$3 == "osc" {print $4; exit}') + local host=$(awk '/current_connection:/ {print $2}' /proc/fs/lustre/osc/$osc/import) + local target=$(awk '/target:/ {print $2}' /proc/fs/lustre/osc/$osc/import) + target=${target%_UUID} + + [[ -n $target ]] && { setup_obdecho_osc $host $target || rc=1; } || rc=1 + [ $rc -eq 0 ] && { obdecho_create_test ${target}_osc client || rc=2; } + [[ -n $target ]] && cleanup_obdecho_osc $target + [ $rmmod_local -eq 1 ] && rmmod obdecho + return $rc +} +run_test 180a "test obdecho on osc" + +test_180b() { + local rc=0 + local rmmod_remote=0 + + do_facet ost "lsmod | grep -q obdecho || " \ + "{ insmod ${LUSTRE}/obdecho/obdecho.ko || " \ + "modprobe obdecho; }" && rmmod_remote=1 + target=$(do_facet ost $LCTL dl | awk '/obdfilter/ {print $4;exit}') + [[ -n $target ]] && { obdecho_create_test $target ost || rc=1; } + [ $rmmod_remote -eq 1 ] && do_facet ost "rmmod obdecho" + return $rc +} +run_test 180b "test obdecho directly on obdfilter" + # OST pools tests POOL=${POOL:-cea1} TGT_COUNT=$OSTCOUNT @@ -6695,7 +7328,8 @@ TGTPOOL_MAX=$(($TGT_COUNT - 1)) TGTPOOL_STEP=2 TGTPOOL_LIST=`seq $TGTPOOL_FIRST $TGTPOOL_STEP $TGTPOOL_MAX` POOL_ROOT=${POOL_ROOT:-$DIR/d200.pools} -POOL_DIR=$POOL_ROOT/dir_tst +POOL_DIR_NAME=dir_tst +POOL_DIR=$POOL_ROOT/$POOL_DIR_NAME POOL_FILE=$POOL_ROOT/file_tst check_file_in_pool() @@ -6716,24 +7350,27 @@ check_file_in_pool() return 0 } +trap "cleanup_pools $FSNAME" EXIT + test_200a() { remote_mgs_nodsh && skip "remote MGS with nodsh" && return - do_facet mgs $LCTL pool_new $FSNAME.$POOL - # 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" - [ $($LFS pool_list $FSNAME | grep -c $POOL) -eq 1 ] || error "$POOL not in lfs pool_list" + create_pool $FSNAME.$POOL || return $? + [ $($LFS pool_list $FSNAME | grep -c $POOL) -eq 1 ] || + error "$POOL not in lfs pool_list" } run_test 200a "Create new pool ==========================================" test_200b() { 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 " $i; done) + TGT=$(for i in $TGTPOOL_LIST; do printf "$FSNAME-OST%04x_UUID " $i; done) do_facet mgs $LCTL pool_add $FSNAME.$POOL \ $FSNAME-OST[$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" + wait_update $HOSTNAME "lctl get_param -n lov.$FSNAME-*.pools.$POOL | sort -u | tr '\n' ' ' " "$TGT" || + error "Add to pool failed" local lfscount=$($LFS pool_list $FSNAME.$POOL | grep -c "\-OST") local addcount=$((($TGTPOOL_MAX - $TGTPOOL_FIRST) / $TGTPOOL_STEP + 1)) - [ $lfscount -eq $addcount ] || error "lfs pool_list bad ost count $lfscount != $addcount" + [ $lfscount -eq $addcount ] || + error "lfs pool_list bad ost count $lfscount != $addcount" } run_test 200b "Add targets to a pool ====================================" @@ -6742,13 +7379,25 @@ test_200c() { mkdir -p $POOL_DIR $SETSTRIPE -c 2 -p $POOL $POOL_DIR [ $? = 0 ] || error "Cannot set pool $POOL to $POOL_DIR" + # b-19919 test relative path works well + mkdir -p $POOL_DIR/$POOL_DIR_NAME + cd $POOL_DIR + $SETSTRIPE -c 2 -p $POOL $POOL_DIR_NAME + [ $? = 0 ] || error "Cannot set pool $POOL to $POOL_DIR/$POOL_DIR_NAME" + $SETSTRIPE -c 2 -p $POOL ./$POOL_DIR_NAME + [ $? = 0 ] || error "Cannot set pool $POOL to $POOL_DIR/./$POOL_DIR_NAME" + $SETSTRIPE -c 2 -p $POOL ../$POOL_DIR_NAME + [ $? = 0 ] || error "Cannot set pool $POOL to $POOL_DIR/../$POOL_DIR_NAME" + $SETSTRIPE -c 2 -p $POOL ../$POOL_DIR_NAME/$POOL_DIR_NAME + [ $? = 0 ] || error "Cannot set pool $POOL to $POOL_DIR/../$POOL_DIR_NAME/$POOL_DIR_NAME" + rm -rf $POOL_DIR_NAME; cd - } run_test 200c "Set pool on a directory =================================" test_200d() { remote_mgs_nodsh && skip "remote MGS with nodsh" && return - res=$($GETSTRIPE --pool $POOL_DIR | awk '/^pool:/ {print $2}') - [ "$res" = $POOL ] || error "Pool on $POOL_DIR is $res, not $POOL" + res=$($GETSTRIPE --pool $POOL_DIR) + [ $res = $POOL ] || error "Pool on $POOL_DIR is $res, not $POOL" } run_test 200d "Check pool on a directory ===============================" @@ -6788,43 +7437,62 @@ test_200f() { run_test 200f "Create files in a pool ===================================" test_200g() { - remote_mgs_nodsh && skip "remote MGS with nodsh" && return - TGT=$($LCTL get_param -n lov.$FSNAME-*.pools.$POOL | head -1) - res=$($LFS df --pool $FSNAME.$POOL | awk '{print $1}' | grep "$FSNAME-OST ") - [ "$res" = "$TGT" ] || echo "Pools OSTS $TGT is not $res that lfs df reports" + remote_mgs_nodsh && skip "remote MGS with nodsh" && return + TGT=$($LCTL get_param -n lov.$FSNAME-clilov-*.pools.$POOL | tr '\n' ' ') + res=$($LFS df --pool $FSNAME.$POOL | awk '{print $1}' | grep "$FSNAME-OST" | tr '\n' ' ') + [ "$res" = "$TGT" ] || error "Pools OSTs '$TGT' is not '$res' that lfs df reports" } run_test 200g "lfs df a pool ============================================" -test_201a() { # was 200g +test_201a() { 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 - wait_update $HOSTNAME "lctl get_param -n lov.$FSNAME-*.pools.$POOL | grep $TGT" "" || error "$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 201a "Remove a target from a pool =============================" -test_201b() { # was 200h +test_201b() { 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 - 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" + wait_update $HOSTNAME "lctl get_param -n lov.$FSNAME-*.pools.$POOL" "" || + error "Pool $FSNAME.$POOL cannot be drained" + # striping on an empty/nonexistant pool should fall back to "pool of everything" + touch ${POOL_DIR}/$tfile || error "failed to use fallback striping for empty pool" + # setstripe on an empty pool should fail + $SETSTRIPE -p $POOL ${POOL_FILE}/$tfile 2>/dev/null && \ + error "expected failure when creating file with empty pool" + return 0 } run_test 201b "Remove all targets from a pool ==========================" -test_201c() { # was 200i +test_201c() { remote_mgs_nodsh && skip "remote MGS with nodsh" && return do_facet mgs $LCTL pool_destroy $FSNAME.$POOL + + sleep 2 + # striping on an empty/nonexistant pool should fall back to "pool of everything" + touch ${POOL_DIR}/$tfile || error "failed to use fallback striping for missing pool" + # setstripe on an empty pool should fail + $SETSTRIPE -p $POOL ${POOL_FILE}/$tfile 2>/dev/null && \ + error "expected failure when creating file with missing pool" + # 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 + if wait_update $HOSTNAME "lctl get_param -n lov.$FSNAME-*.pools.$POOL 2>/dev/null || + echo foo" "foo"; then + remove_pool_from_list $FSNAME.$POOL + return 0 + fi error "Pool $FSNAME.$POOL is not destroyed" } run_test 201c "Remove a pool ============================================" +cleanup_pools $FSNAME + test_212() { size=`date +%s` size=$((size % 8192 + 1)) @@ -6858,100 +7526,200 @@ test_214() { # for bug 20133 } run_test 214 "hash-indexed directory test - bug 20133" -test_215() { # for bug 18102 - # /proc/sys/lnet/stats should look as 11 space-separated numbers - cat /proc/sys/lnet/stats >$TMP/lnet_stats.out - sysctl lnet.stats |sed 's/^lnet.stats\ =\ //g' >$TMP/lnet_stats.sys - STATS_LINES_OUT=$(cat $TMP/lnet_stats.out|wc -l) - [ "$STATS_LINES_OUT" = 1 ] || error "/proc/sys/lnet/stats has more than 1 line: $STATS" - STATS_LINES_SYS=$(cat $TMP/lnet_stats.sys|wc -l) - [ "$STATS_LINES_SYS" = 1 ] || error "lnet.stats has more than 1 line: $STATS" - STATS_REG='^[0-9]\+ [0-9]\+ [0-9]\+ [0-9]\+ [0-9]\+ [0-9]\+ [0-9]\+ [0-9]\+ [0-9]\+ [0-9]\+ [0-9]\+$' - grep "$STATS_REG" $TMP/lnet_stats.out || (cat $TMP/lnet_stats.out && - error "/proc/sys/lnet/stats misformatted") - grep "$STATS_REG" $TMP/lnet_stats.sys || (cat $TMP/lnet_stats.sys && - error "lnet.stats misformatted") - rm -f $TMP/lnet_stats.out $TMP/lnet_stats.sys - - # /proc/sys/lnet/routes should look exactly as expected - cat /proc/sys/lnet/routes >$TMP/lnet_routes.out - sysctl lnet.routes |sed 's/^lnet.routes\ =\ //g' >$TMP/lnet_routes.sys - echo "Routing disabled" >$TMP/lnet_routes.expected - echo "net hops state router" >>$TMP/lnet_routes.expected - diff $TMP/lnet_routes.expected $TMP/lnet_routes.out || - error "/proc/sys/lnet/routes does not look as expected" - diff $TMP/lnet_routes.expected $TMP/lnet_routes.sys || - error "lnet.routes does not look as expected" - rm -f $TMP/lnet_routes.expected $TMP/lnet_routes.out $TMP/lnet_routes.sys - - # /proc/sys/lnet/routers should look exactly as expected - cat /proc/sys/lnet/routers >$TMP/lnet_routers.out - sysctl lnet.routers |sed 's/^lnet.routers\ =\ //g' >$TMP/lnet_routers.sys - echo "ref rtr_ref alive_cnt state last_ping router" >$TMP/lnet_routers.expected - diff $TMP/lnet_routers.expected $TMP/lnet_routers.out || - error "/proc/sys/lnet/routers does not look as expected" - diff $TMP/lnet_routers.expected $TMP/lnet_routers.sys || - error "lnet.routers does not look as expected" - rm -f $TMP/lnet_routers.expected $TMP/lnet_routers.out $TMP/lnet_routers.sys - - # fisrt line of /proc/sys/lnet/peers should look exactly as expected - cat /proc/sys/lnet/peers >$TMP/lnet_peers.out - sysctl lnet.peers |sed 's/^lnet.peers\ =\ //g' >$TMP/lnet_peers.sys - head -1 $TMP/lnet_peers.out > $TMP/lnet_peers1.out - echo "nid refs state max rtr min tx min queue" >$TMP/lnet_peers1.expected - diff $TMP/lnet_peers1.expected $TMP/lnet_peers1.out || - error "first line of /proc/sys/lnet/peers does not look as expected" - rm -f $TMP/lnet_peers1.expected $TMP/lnet_peers1.out - # other lines should look as a nid followed by 1 number, a word, 6 numbers, e.g.: - # 0@lo 1 NA 0 0 0 0 0 0 - TOTAL_LINES=$(cat $TMP/lnet_peers.out |wc -l) - OTHER_LINES=$(($TOTAL_LINES - 1)) - MATCHED_LINES=$(cat $TMP/lnet_peers.out |tail -$TOTAL_LINES | - grep -c "^[0-9.]\+@[a-z0-9]\+ *[0-9]\+ *[a-Z]\+ *[0-9]\+ *[0-9]\+ *-\?[0-9]\+ *[0-9]\+ *-\?[0-9]\+ *[0-9]\+$") - [ "$MATCHED_LINES" = "$OTHER_LINES" ] || (cat $TMP/lnet_peers.out && - error "/proc/sys/lnet/peers misformatted") - diff $TMP/lnet_peers.out $TMP/lnet_peers.sys || - error "lnet.peers does not look as expected" - rm -f $TMP/lnet_peers.out $TMP/lnet_peers.sys - - # /proc/sys/lnet/buffers should look exactly as expected - cat /proc/sys/lnet/buffers >$TMP/lnet_buffers.out - sysctl lnet.buffers |sed 's/^lnet.buffers\ =\ //g' >$TMP/lnet_buffers.sys - echo "pages count credits min" >$TMP/lnet_buffers.expected - echo " 0 0 0 0" >>$TMP/lnet_buffers.expected - echo " 1 0 0 0" >>$TMP/lnet_buffers.expected - echo " 256 0 0 0" >>$TMP/lnet_buffers.expected - diff $TMP/lnet_buffers.expected $TMP/lnet_buffers.out || - error "/proc/sys/lnet/buffers does not look as expected" - diff $TMP/lnet_buffers.expected $TMP/lnet_buffers.sys || - error "lnet.buffers does not look as expected" - rm -f $TMP/lnet_buffers.expected $TMP/lnet_buffers.out $TMP/lnet_buffers.sys - - # fisrt line of /proc/sys/lnet/nis should look exactly as expected - cat /proc/sys/lnet/nis >$TMP/lnet_nis.out - sysctl lnet.nis |sed 's/^lnet.nis\ =\ //g' >$TMP/lnet_nis.sys - head -1 $TMP/lnet_nis.out > $TMP/lnet_nis1.out - echo "nid refs peer rtr max tx min" >$TMP/lnet_nis1.expected - diff $TMP/lnet_nis1.expected $TMP/lnet_nis1.out || - error "first line of /proc/sys/lnet/nis does not look as expected" - rm -f $TMP/lnet_nis1.expected $TMP/lnet_nis1.out - # other lines should look as a nid followed by 6 numbers, e.g.: - # 0@lo 3 0 0 0 0 0 - TOTAL_LINES=$(cat $TMP/lnet_nis.out |wc -l) - OTHER_LINES=$(($TOTAL_LINES - 1)) - MATCHED_LINES=$(cat $TMP/lnet_nis.out |tail -$TOTAL_LINES | - grep -c "^[0-9.]\+@[a-z0-9]\+ *[0-9]\+ *[0-9]\+ *[0-9]\+ *[0-9]\+ *[0-9]\+ *[0-9]\+$") - [ "$MATCHED_LINES" = "$OTHER_LINES" ] || (cat $TMP/lnet_nis.out && - error "/proc/sys/lnet/nis misformatted") - diff $TMP/lnet_nis.out $TMP/lnet_nis.sys || - error "lnet.nis does not look as expected" - rm -f $TMP/lnet_nis.out $TMP/lnet_nis.sys +# having "abc" as 1st arg, creates $TMP/lnet_abc.out and $TMP/lnet_abc.sys +create_lnet_proc_files() { + cat /proc/sys/lnet/$1 >$TMP/lnet_$1.out || error "cannot read /proc/sys/lnet/$1" + sysctl lnet.$1 >$TMP/lnet_$1.sys_tmp || error "cannot read lnet.$1" + + sed "s/^lnet.$1\ =\ //g" "$TMP/lnet_$1.sys_tmp" >$TMP/lnet_$1.sys + rm -f "$TMP/lnet_$1.sys_tmp" +} + +# counterpart of create_lnet_proc_files +remove_lnet_proc_files() { + rm -f $TMP/lnet_$1.out $TMP/lnet_$1.sys +} + +# uses 1st arg as trailing part of filename, 2nd arg as description for reports, +# 3rd arg as regexp for body +check_lnet_proc_stats() { + local l=$(cat "$TMP/lnet_$1" |wc -l) + [ $l = 1 ] || (cat "$TMP/lnet_$1" && error "$2 is not of 1 line: $l") + + grep -E "$3" "$TMP/lnet_$1" || (cat "$TMP/lnet_$1" && error "$2 misformatted") +} + +# uses 1st arg as trailing part of filename, 2nd arg as description for reports, +# 3rd arg as regexp for body, 4th arg as regexp for 1st line, 5th arg is +# optional and can be regexp for 2nd line (lnet.routes case) +check_lnet_proc_entry() { + local blp=2 # blp stands for 'position of 1st line of body' + [ "$5" = "" ] || blp=3 # lnet.routes case + + local l=$(cat "$TMP/lnet_$1" |wc -l) + # subtracting one from $blp because the body can be empty + [ "$l" -ge "$(($blp - 1))" ] || (cat "$TMP/lnet_$1" && error "$2 is too short: $l") + + sed -n '1 p' "$TMP/lnet_$1" |grep -E "$4" >/dev/null || + (cat "$TMP/lnet_$1" && error "1st line of $2 misformatted") + + [ "$5" = "" ] || sed -n '2 p' "$TMP/lnet_$1" |grep -E "$5" >/dev/null || + (cat "$TMP/lnet_$1" && error "2nd line of $2 misformatted") + + # bail out if any unexpected line happened + sed -n "$blp~1 p" "$TMP/lnet_$1" |grep -Ev "$3" + [ "$?" != 0 ] || error "$2 misformatted" +} + +test_215() { # for bugs 18102, 21079, 21517 + local N='(0|[1-9][0-9]*)' # non-negative numeric + local P='[1-9][0-9]*' # positive numeric + local I='(0|-?[1-9][0-9]*)' # any numeric (0 | >0 | <0) + local NET='[a-z][a-z0-9]*' # LNET net like o2ib2 + local ADDR='[0-9.]+' # LNET addr like 10.0.0.1 + local NID="$ADDR@$NET" # LNET nid like 10.0.0.1@o2ib2 + + local L1 # regexp for 1st line + local L2 # regexp for 2nd line (optional) + local BR # regexp for the rest (body) + + # /proc/sys/lnet/stats should look as 11 space-separated non-negative numerics + BR="^$N $N $N $N $N $N $N $N $N $N $N$" + create_lnet_proc_files "stats" + check_lnet_proc_stats "stats.out" "/proc/sys/lnet/stats" "$BR" + check_lnet_proc_stats "stats.sys" "lnet.stats" "$BR" + remove_lnet_proc_files "stats" + + # /proc/sys/lnet/routes should look like this: + # Routing disabled/enabled + # net hops state router + # where net is a string like tcp0, hops >= 0, state is up/down, + # router is a string like 192.168.1.1@tcp2 + L1="^Routing (disabled|enabled)$" + L2="^net +hops +state +router$" + BR="^$NET +$N +(up|down) +$NID$" + create_lnet_proc_files "routes" + check_lnet_proc_entry "routes.out" "/proc/sys/lnet/routes" "$BR" "$L1" "$L2" + check_lnet_proc_entry "routes.sys" "lnet.routes" "$BR" "$L1" "$L2" + remove_lnet_proc_files "routes" + + # /proc/sys/lnet/routers should look like this: + # ref rtr_ref alive_cnt state last_ping ping_sent deadline down_ni router + # where ref > 0, rtr_ref > 0, alive_cnt >= 0, state is up/down, + # last_ping >= 0, ping_sent is boolean (0/1), deadline and down_ni are + # numeric (0 or >0 or <0), router is a string like 192.168.1.1@tcp2 + L1="^ref +rtr_ref +alive_cnt +state +last_ping +ping_sent +deadline +down_ni +router$" + BR="^$P +$P +$N +(up|down) +$N +(0|1) +$I +$I +$NID$" + create_lnet_proc_files "routers" + check_lnet_proc_entry "routers.out" "/proc/sys/lnet/routers" "$BR" "$L1" + check_lnet_proc_entry "routers.sys" "lnet.routers" "$BR" "$L1" + remove_lnet_proc_files "routers" + + # /proc/sys/lnet/peers should look like this: + # nid refs state max rtr min tx min queue + # where nid is a string like 192.168.1.1@tcp2, refs > 0, + # state is up/down/NA, max >= 0. rtr, min, tx, min are + # numeric (0 or >0 or <0), queue >= 0. + L1="^nid +refs +state +max +rtr +min +tx +min +queue$" + BR="^$NID +$P +(up|down|NA) +$N +$I +$I +$I +$I +$N$" + create_lnet_proc_files "peers" + check_lnet_proc_entry "peers.out" "/proc/sys/lnet/peers" "$BR" "$L1" + check_lnet_proc_entry "peers.sys" "lnet.peers" "$BR" "$L1" + remove_lnet_proc_files "peers" + + # /proc/sys/lnet/buffers should look like this: + # pages count credits min + # where pages >=0, count >=0, credits and min are numeric (0 or >0 or <0) + L1="^pages +count +credits +min$" + BR="^ +$N +$N +$I +$I$" + create_lnet_proc_files "buffers" + check_lnet_proc_entry "buffers.out" "/proc/sys/lnet/buffers" "$BR" "$L1" + check_lnet_proc_entry "buffers.sys" "lnet.buffers" "$BR" "$L1" + remove_lnet_proc_files "buffers" + + # /proc/sys/lnet/nis should look like this: + # nid status alive refs peer rtr max tx min + # where nid is a string like 192.168.1.1@tcp2, status is up/down, + # alive is numeric (0 or >0 or <0), refs > 0, peer >= 0, + # rtr >= 0, max >=0, tx and min are numeric (0 or >0 or <0). + L1="^nid +status +alive +refs +peer +rtr +max +tx +min$" + BR="^$NID +(up|down) +$I +$P +$N +$N +$N +$I +$I$" + create_lnet_proc_files "nis" + check_lnet_proc_entry "nis.out" "/proc/sys/lnet/nis" "$BR" "$L1" + check_lnet_proc_entry "nis.sys" "lnet.nis" "$BR" "$L1" + remove_lnet_proc_files "nis" # can we successfully write to /proc/sys/lnet/stats? echo "0" >/proc/sys/lnet/stats || error "cannot write to /proc/sys/lnet/stats" sysctl -w lnet.stats=0 || error "cannot write to lnet.stats" } -run_test 215 "/proc/sys/lnet exists and has proper content - bug 18102" +run_test 215 "/proc/sys/lnet exists and has proper content - bugs 18102, 21079, 21517" + +test_216() { # bug 20317 + local node + local p="$TMP/sanityN-$TESTNAME.parameters" + save_lustre_params $HOSTNAME "osc.*.contention_seconds" > $p + for node in $(osts_nodes); do + save_lustre_params $node "ldlm.namespaces.filter-*.max_nolock_bytes" >> $p + save_lustre_params $node "ldlm.namespaces.filter-*.contended_locks" >> $p + save_lustre_params $node "ldlm.namespaces.filter-*.contention_seconds" >> $p + done + clear_osc_stats + + # agressive lockless i/o settings + for node in $(osts_nodes); do + do_node $node 'lctl set_param -n ldlm.namespaces.filter-*.max_nolock_bytes 2000000; lctl set_param -n ldlm.namespaces.filter-*.contended_locks 0; lctl set_param -n ldlm.namespaces.filter-*.contention_seconds 60' + done + lctl set_param -n osc.*.contention_seconds 60 + + $DIRECTIO write $DIR/$tfile 0 10 4096 + $CHECKSTAT -s 40960 $DIR/$tfile + + # disable lockless i/o + for node in $(osts_nodes); do + do_node $node 'lctl set_param -n ldlm.namespaces.filter-*.max_nolock_bytes 0; lctl set_param -n ldlm.namespaces.filter-*.contended_locks 32; lctl set_param -n ldlm.namespaces.filter-*.contention_seconds 0' + done + lctl set_param -n osc.*.contention_seconds 0 + clear_osc_stats + + dd if=/dev/zero of=$DIR/$tfile count=0 + $CHECKSTAT -s 0 $DIR/$tfile + + restore_lustre_params <$p + rm -f $p + rm $DIR/$tfile +} +run_test 216 "check lockless direct write works and updates file size and kms correctly" + +test_217() { # bug 22430 + local node + for node in $(nodes_list); do + if [[ $node = *-* ]] ; then + echo "lctl ping $node@$NETTYPE" + lctl ping $node@$NETTYPE + else + echo "skipping $node (no hiphen detected)" + fi + done +} +run_test 217 "check lctl ping for hostnames with hiphen ('-')" + +test_218() { + # 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 218 "parallel read and truncate should not deadlock =======================" # # tests that do cleanup/setup should be run at the end