X-Git-Url: https://git.whamcloud.com/?p=fs%2Flustre-release.git;a=blobdiff_plain;f=lustre%2Ftests%2Fsanity.sh;h=b0220e205abdda1c0c308b4901e9bea2e270c20f;hp=55bd93a8f8ff1f8ef9eeb5ccd8746cea6e4a1cf2;hb=db68565b462d954dc50b0f638fea820991f54209;hpb=1df3357a1937481a19b23f5f9c4acf581b1ab5c2 diff --git a/lustre/tests/sanity.sh b/lustre/tests/sanity.sh index 55bd93a..b0220e2 100644 --- a/lustre/tests/sanity.sh +++ b/lustre/tests/sanity.sh @@ -447,6 +447,10 @@ test_17g() { [ $(lustre_version_code $SINGLEMDS) -le $(version_code 2.3.55) ] && TESTS="4094 4095" + # skip long symlink name for rhel6.5. + # rhel6.5 has a limit (PATH_MAX - sizeof(struct filename)) + grep -q '6.5' /etc/redhat-release && TESTS="59 60 61 4062 4063" + for i in $TESTS; do local SYMNAME=$(str_repeat 'x' $i) ln -s $SYMNAME $DIR/$tdir/f$i || error "failed $i-char symlink" @@ -571,7 +575,7 @@ test_17m() { echo "stop and checking mds${mds_index}: $cmd" # e2fsck should not return error - stop mds${mds_index} -f + stop mds${mds_index} do_facet mds${mds_index} $cmd || rc=$? start mds${mds_index} $devname $MDS_MOUNT_OPTS @@ -588,7 +592,9 @@ check_fs_consistency_17n() { local cmd local rc=0 - for mdt_index in $(seq 1 $MDSCOUNT); do + # create/unlink in 17n only change 2 MDTs(MDT1/MDT2), + # so it only check MDT1/MDT2 instead of all of MDTs. + for mdt_index in $(seq 1 2); do devname=$(mdsdevname $mdt_index) cmd="$E2FSCK -fnvd $devname" @@ -638,6 +644,9 @@ test_17n() { run_test 17n "run e2fsck against master/slave MDT which contains remote dir" test_17o() { + [[ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.3.64) ]] && + skip "Need MDS version at least 2.3.64" && return + local WDIR=$DIR/${tdir}o local mdt_index local mdtdevname @@ -785,7 +794,7 @@ run_test 24c "rename directory to non-existent target" test_24d() { test_mkdir $DIR/$tdir test_mkdir $DIR/$tdir/d$testnum.1 - test_mkdir $DIR/$tdir/d$ttestnum.2 + test_mkdir $DIR/$tdir/d$testnum.2 mrename $DIR/$tdir/d$testnum.1 $DIR/$tdir/d$testnum.2 $CHECKSTAT -a $DIR/$tdir/d$testnum.1 || error "d$testnum.1 exists" $CHECKSTAT -t dir $DIR/$tdir/d$testnum.2 || error "d$testnum.2 not dir" @@ -987,7 +996,7 @@ max_pages_per_rpc() { test_24v() { local NRFILES=100000 - local FREE_INODES=$(lfs_df -i | grep "summary" | awk '{print $4}') + local FREE_INODES=$(mdt_free_inodes 0) [ $FREE_INODES -lt $NRFILES ] && \ skip "not enough free inodes $FREE_INODES required $NRFILES" && return @@ -995,6 +1004,9 @@ test_24v() { [ $PARALLEL == "yes" ] && skip "skip parallel run" && return trap simple_cleanup_common EXIT + # Performance issue on ZFS see LU-4072 (c.f. LU-2887) + [ $(facet_fstype $SINGLEMDS) = "zfs" ] && NRFILES=10000 + mkdir -p $DIR/$tdir createmany -m $DIR/$tdir/$tfile $NRFILES @@ -1047,6 +1059,9 @@ test_24x() { mkdir -p $remote_dir/tgt_dir touch $remote_dir/tgt_file + mrename $remote_dir $DIR/ && + error "rename dir cross MDT works!" + mrename $DIR/$tdir/src_dir $remote_dir/tgt_dir && error "rename dir cross MDT works!" @@ -1117,6 +1132,7 @@ run_test 24z "rename one remote dir to another remote dir should fail" test_24A() { # LU-3182 local NFILES=5000 + rm -rf $DIR/$tdir mkdir -p $DIR/$tdir createmany -m $DIR/$tdir/$tfile $NFILES local t=`ls $DIR/$tdir | wc -l` @@ -1274,10 +1290,10 @@ test_27k() { # bug 2844 [ ! -d $DIR/d27 ] && test_mkdir -p $DIR d27 $SETSTRIPE -S 67108864 $FILE || error "setstripe failed" BLKSIZE=`stat $FILE | awk '/IO Block:/ { print $7 }'` - [ $BLKSIZE -le $LL_MAX_BLKSIZE ] || error "$BLKSIZE > $LL_MAX_BLKSIZE" + [ $BLKSIZE -le $LL_MAX_BLKSIZE ] || error "1:$BLKSIZE > $LL_MAX_BLKSIZE" dd if=/dev/zero of=$FILE bs=4k count=1 BLKSIZE=`stat $FILE | awk '/IO Block:/ { print $7 }'` - [ $BLKSIZE -le $LL_MAX_BLKSIZE ] || error "$BLKSIZE > $LL_MAX_BLKSIZE" + [ $BLKSIZE -le $LL_MAX_BLKSIZE ] || error "2:$BLKSIZE > $LL_MAX_BLKSIZE" } run_test 27k "limit i_blksize for broken user apps =============" @@ -1352,9 +1368,9 @@ exhaust_precreations() { # on the mdt's osc local mdtosc_proc1=$(get_mdtosc_proc_path mds${MDSIDX} $OST) local last_id=$(do_facet mds${MDSIDX} lctl get_param -n \ - osc.$mdtosc_proc1.prealloc_last_id) + osc.$mdtosc_proc1.prealloc_last_id) local next_id=$(do_facet mds${MDSIDX} lctl get_param -n \ - osc.$mdtosc_proc1.prealloc_next_id) + osc.$mdtosc_proc1.prealloc_next_id) local mdtosc_proc2=$(get_mdtosc_proc_path mds${MDSIDX}) do_facet mds${MDSIDX} lctl get_param osc.$mdtosc_proc2.prealloc* @@ -1708,26 +1724,7 @@ check_seq_oid() local dev=$(ostdevname $ost) local oid_hex - if [ $(facet_fstype ost$ost) != ldiskfs ]; then - echo "Currently only works with ldiskfs-based OSTs" - continue - fi - - log "want: stripe:$stripe_nr ost:$obdidx oid:$oid/$hex seq:$seq" - - #don't unmount/remount the OSTs if we don't need to do that - # LU-2577 changes filter_fid to be smaller, so debugfs needs - # update too, until that use mount/ll_decode_filter_fid/mount - local dir=$(facet_mntpt ost$ost) - local opts=${OST_MOUNT_OPTS} - - if ! do_facet ost$ost test -b ${dev}; then - opts=$(csa_add "$opts" -o loop) - fi - - stop ost$ost - do_facet ost$ost mount -t $(facet_fstype ost$ost) $opts $dev $dir || - { error "mounting $dev as $FSTYPE failed"; return 3; } + log "want: stripe:$stripe_nr ost:$obdidx oid:$oid/$hex seq:$seq" seq=$(echo $seq | sed -e "s/^0x//g") if [ $seq == 0 ]; then @@ -1735,25 +1732,48 @@ check_seq_oid() else oid_hex=$(echo $hex | sed -e "s/^0x//g") fi - local obj_file=$(do_facet ost$ost find $dir/O/$seq -name $oid_hex) - local ff=$(do_facet ost$ost $LL_DECODE_FILTER_FID $obj_file) - do_facet ost$ost umount -d $dir - start ost$ost $dev $OST_MOUNT_OPTS + local obj_file="O/$seq/d$((oid %32))/$oid_hex" - # re-enable when debugfs will understand new filter_fid - #local ff=$(do_facet ost$ost "$DEBUGFS -c -R 'stat $obj_file' \ - # $dev 2>/dev/null" | grep "parent=") + local ff + # + # Don't unmount/remount the OSTs if we don't need to do that. + # LU-2577 changes filter_fid to be smaller, so debugfs needs + # update too, until that use mount/ll_decode_filter_fid/mount. + # Re-enable when debugfs will understand new filter_fid. + # + if false && [ $(facet_fstype ost$ost) == ldiskfs ]; then + ff=$(do_facet ost$ost "$DEBUGFS -c -R 'stat $obj_file' \ + $dev 2>/dev/null" | grep "parent=") + else + stop ost$ost + mount_fstype ost$ost + ff=$(do_facet ost$ost $LL_DECODE_FILTER_FID \ + $(facet_mntpt ost$ost)/$obj_file) + unmount_fstype ost$ost + start ost$ost $dev $OST_MOUNT_OPTS + fi - [ -z "$ff" ] && error "$obj_file: no filter_fid info" + [ -z "$ff" ] && error "$obj_file: no filter_fid info" - echo "$ff" | sed -e 's#.*objid=#got: objid=#' + echo "$ff" | sed -e 's#.*objid=#got: objid=#' - # /mnt/O/0/d23/23: objid=23 seq=0 parent=[0x200000400:0x1e:0x1] - # fid: objid=23 seq=0 parent=[0x200000400:0x1e:0x0] stripe=1 - local ff_parent=$(echo $ff|sed -e 's/.*parent=.//') - local ff_pseq=$(echo $ff_parent | cut -d: -f1) - local ff_poid=$(echo $ff_parent | cut -d: -f2) - local ff_pstripe=$(echo $ff_parent | sed -e 's/.*stripe=//') + # /mnt/O/0/d23/23: objid=23 seq=0 parent=[0x200000400:0x1e:0x1] + # fid: objid=23 seq=0 parent=[0x200000400:0x1e:0x0] stripe=1 + local ff_parent=$(echo $ff|sed -e 's/.*parent=.//') + local ff_pseq=$(echo $ff_parent | cut -d: -f1) + local ff_poid=$(echo $ff_parent | cut -d: -f2) + local ff_pstripe + if echo $ff_parent | grep -q 'stripe='; then + ff_pstripe=$(echo $ff_parent | sed -e 's/.*stripe=//') + else + # + # $LL_DECODE_FILTER_FID does not print "stripe="; look + # into f_ver in this case. See the comment on + # ff_parent. + # + ff_pstripe=$(echo $ff_parent | cut -d: -f3 | + sed -e 's/\]//') + fi # compare lmm_seq and filter_fid->ff_parent.f_seq [ $ff_pseq = $lmm_seq ] || @@ -1761,7 +1781,7 @@ check_seq_oid() # compare lmm_object_id and filter_fid->ff_parent.f_oid [ $ff_poid = $lmm_oid ] || error "FF parent OID $ff_poid != $lmm_oid" - [ $ff_pstripe = $stripe_nr ] || + (($ff_pstripe == $stripe_nr)) || error "FF stripe $ff_pstripe != $stripe_nr" stripe_nr=$((stripe_nr + 1)) @@ -1799,13 +1819,13 @@ test_27A() { # b=19102 local restore_count=$($GETSTRIPE -c $MOUNT) local restore_offset=$($GETSTRIPE -i $MOUNT) $SETSTRIPE -c 0 -i -1 -S 0 $MOUNT + wait_update $HOSTNAME "$GETSTRIPE -c $MOUNT | sed 's/ *//g'" "1" 20 || + error "stripe count $($GETSTRIPE -c $MOUNT) != 1" local default_size=$($GETSTRIPE -S $MOUNT) - local default_count=$($GETSTRIPE -c $MOUNT) local default_offset=$($GETSTRIPE -i $MOUNT) local dsize=$((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 -i $restore_offset -S $restore_size $MOUNT } @@ -2106,18 +2126,20 @@ test_31m() { run_test 31m "link to file: the same, non-existing, dir===============" test_31n() { - [ -e /proc/self/fd/173 ] && echo "skipping, fd 173 is in use" && return touch $DIR/$tfile || error "cannot create '$DIR/$tfile'" nlink=$(stat --format=%h $DIR/$tfile) [ ${nlink:--1} -eq 1 ] || error "nlink is $nlink, expected 1" - exec 173<$DIR/$tfile - trap "exec 173<&-" EXIT - nlink=$(stat --dereference --format=%h /proc/self/fd/173) + local fd=$(free_fd) + local cmd="exec $fd<$DIR/$tfile" + eval $cmd + cmd="exec $fd<&-" + trap "eval $cmd" EXIT + nlink=$(stat --dereference --format=%h /proc/self/fd/$fd) [ ${nlink:--1} -eq 1 ] || error "nlink is $nlink, expected 1" rm $DIR/$tfile || error "cannot remove '$DIR/$tfile'" - nlink=$(stat --dereference --format=%h /proc/self/fd/173) + nlink=$(stat --dereference --format=%h /proc/self/fd/$fd) [ ${nlink:--1} -eq 0 ] || error "nlink is $nlink, expected 0" - exec 173<&- + eval $cmd } run_test 31n "check link count of unlinked file" @@ -2146,7 +2168,7 @@ run_test 31o "duplicate hard links with same filename" cleanup_test32_mount() { trap 0 - $UMOUNT $DIR/$tdir/ext2-mountpoint + $UMOUNT -d $DIR/$tdir/ext2-mountpoint } test_32a() { @@ -2354,7 +2376,7 @@ run_test 32p "open d32p/symlink->tmp/symlink->lustre-root/$tfile" cleanup_testdir_mount() { trap 0 - $UMOUNT $DIR/$tdir + $UMOUNT -d $DIR/$tdir } test_32q() { @@ -3090,6 +3112,81 @@ test_39m() { } run_test 39m "test atime and mtime before 1970" +test_39n() { # LU-3832 + local atime_diff=$(do_facet $SINGLEMDS \ + lctl get_param -n mdd.*MDT0000*.atime_diff) + local atime0 + local atime1 + local atime2 + + do_facet $SINGLEMDS lctl set_param -n mdd.*MDT0000*.atime_diff=1 + + rm -rf $DIR/$tfile + dd if=/dev/zero of=$DIR/$tfile bs=4096 count=1 status=noxfer + atime0=$(stat -c %X $DIR/$tfile) + + sleep 5 + $MULTIOP $DIR/$tfile oO_RDONLY:O_NOATIME:r4096c + atime1=$(stat -c %X $DIR/$tfile) + + sleep 5 + cancel_lru_locks mdc + cancel_lru_locks osc + $MULTIOP $DIR/$tfile oO_RDONLY:O_NOATIME:r4096c + atime2=$(stat -c %X $DIR/$tfile) + + do_facet $SINGLEMDS \ + lctl set_param -n mdd.*MDT0000*.atime_diff=$atime_diff + + [ "$atime0" -eq "$atime1" ] || error "atime0 $atime0 != atime1 $atime1" + [ "$atime1" -eq "$atime2" ] || error "atime0 $atime0 != atime1 $atime1" +} +run_test 39n "check that O_NOATIME is honored" + +test_39o() { + TESTDIR=$DIR/$tdir/$tfile + [ -e $TESTDIR ] && rm -rf $TESTDIR + test_mkdir -p $TESTDIR + cd $TESTDIR + links1=2 + ls + mkdir a b + ls + links2=$(stat -c %h .) + [ $(($links1 + 2)) != $links2 ] && + error "wrong links count $(($links1 + 2)) != $links2" + rmdir b + links3=$(stat -c %h .) + [ $(($links1 + 1)) != $links3 ] && + error "wrong links count $links1 != $links3" + return 0 +} +run_test 39o "directory cached attributes updated after create ========" + +test_39p() { + [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return + local MDTIDX=1 + TESTDIR=$DIR/$tdir/$tfile + [ -e $TESTDIR ] && rm -rf $TESTDIR + mkdir -p $TESTDIR + cd $TESTDIR + links1=2 + ls + $LFS mkdir -i $MDTIDX $TESTDIR/remote_dir1 + $LFS mkdir -i $MDTIDX $TESTDIR/remote_dir2 + ls + links2=$(stat -c %h .) + [ $(($links1 + 2)) != $links2 ] && + error "wrong links count $(($links1 + 2)) != $links2" + rmdir remote_dir2 + links3=$(stat -c %h .) + [ $(($links1 + 1)) != $links3 ] && + error "wrong links count $links1 != $links3" + return 0 +} +run_test 39p "remote directory cached attributes updated after create ========" + + test_40() { dd if=/dev/zero of=$DIR/$tfile bs=4096 count=1 $RUNAS $OPENFILE -f O_WRONLY:O_TRUNC $DIR/$tfile && @@ -3655,7 +3752,7 @@ run_test 51a "special situations: split htree with empty entry ==" export NUMTEST=70000 test_51b() { [ $PARALLEL == "yes" ] && skip "skip parallel run" && return - local BASE=$DIR/$tdir + local BASE=$DIR/d${base}.${TESTSUITE} # cleanup the directory rm -fr $BASE @@ -3682,7 +3779,7 @@ test_51b() { run_test 51b "exceed 64k subdirectory nlink limit" test_51ba() { # LU-993 - local BASE=$DIR/$tdir + local BASE=$DIR/d${base}.${TESTSUITE} # unlink all but 100 subdirectories, then check it still works local LEFT=100 [ -f $BASE/fnum ] && local NUMPREV=$(cat $BASE/fnum) && rm $BASE/fnum @@ -3812,37 +3909,53 @@ test_53() { local ost_last_id local ostnum local node - local found=0 + local found=false + local support_last_seq=true + + [[ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.3.60) ]] || + support_last_seq=false # only test MDT0000 - local mdtosc=$(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 $param | cut -d "." -f2 | cut -d - -f 1-2) - param_seq=$(echo ${param} | - sed -e s/prealloc_last_id/prealloc_last_seq/g) - mds_last_seq=$(do_facet $SINGLEMDS lctl get_param -n $param_seq) - mds_last=$(do_facet $SINGLEMDS lctl get_param -n $param) + local mdtosc=$(get_mdtosc_proc_path $SINGLEMDS) + local value + 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) + + if $support_last_seq; then + param_seq=$(echo $param | + sed -e s/prealloc_last_id/prealloc_last_seq/g) + mds_last_seq=$(do_facet $SINGLEMDS \ + $LCTL get_param -n $param_seq) + fi + mds_last=$(do_facet $SINGLEMDS $LCTL get_param -n $param) ostnum=$(index_from_ostuuid ${ostname}_UUID) node=$(facet_active_host ost$((ostnum+1))) param="obdfilter.$ostname.last_id" - for ost_last in $(do_node $node lctl get_param -n $param) ; do - echo "$ostname.last_id=$ost_last ;MDS.last_id=$mds_last" - ost_last_id=$(echo $ost_last | awk -F':' '{print $2}' | - sed -e "s/^0x//g") - ost_last_seq=$(echo $ost_last | awk -F':' '{print $1}') - if [ $ost_last_seq = $mds_last_seq ]; then - if [ $ost_last_id != $mds_last ]; then - error "$ost_last != $mds_last_id" - else - found=1 - break - fi + for ost_last in $(do_node $node $LCTL get_param -n $param) ; do + echo "$ostname.last_id=$ost_last; MDS.last_id=$mds_last" + ost_last_id=$ost_last + + if $support_last_seq; then + ost_last_id=$(echo $ost_last | + awk -F':' '{print $2}' | + sed -e "s/^0x//g") + ost_last_seq=$(echo $ost_last | + awk -F':' '{print $1}') + [[ $ost_last_seq = $mds_last_seq ]] || continue + fi + + if [[ $ost_last_id != $mds_last ]]; then + error "$ost_last_id != $mds_last" + else + found=true + break fi done - done - [ $found = 0 ] && error "can not match last_seq/last_id for $mdtosc" + done + $found || error "can not match last_seq/last_id for $mdtosc" return 0 } run_test 53 "verify that MDS and OSTs agree on pre-creation ====" @@ -3895,7 +4008,7 @@ test_54c() { dd if=/dev/zero of=$tdir/tmp bs=`page_size` count=30 || error "dd write" df $tdir dd if=$tdir/tmp of=/dev/zero bs=`page_size` count=30 || error "dd read" - $UMOUNT $tdir + $UMOUNT -d $tdir losetup -d $loopdev rm $loopdev } @@ -3924,8 +4037,7 @@ run_test 54e "console/tty device works in lustre ======================" test_56a() { # was test_56 rm -rf $DIR/$tdir $SETSTRIPE -d $DIR - test_mkdir $DIR/$tdir - test_mkdir $DIR/$tdir/dir + test_mkdir -p $DIR/$tdir/dir NUMFILES=3 NUMFILESx2=$(($NUMFILES * 2)) for i in `seq 1 $NUMFILES` ; do @@ -4511,6 +4623,37 @@ test_56x() { } run_test 56x "lfs migration support" +test_56y() { + [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.4.53) ] && + skip "No HSM support on MDS of $(get_lustre_version)," \ + "need 2.4.53 at least" && return + local res="" + + local dir0=$DIR/$tdir/$testnum + mkdir -p $dir0 || error "creating dir $dir0" + local f1=$dir0/file1 + local f2=$dir0/file2 + + touch $f1 || error "creating std file $f1" + $MULTIOP $f2 H2c || error "creating released file $f2" + + # a directory can be raid0, so ask only for files + res=$($LFIND $dir0 -L raid0 -type f | wc -l) + [[ $res == 2 ]] || error "search raid0: found $res files != 2" + + res=$($LFIND $dir0 \! -L raid0 -type f | wc -l) + [[ $res == 0 ]] || error "search !raid0: found $res files != 0" + + # only files can be released, so no need to force file search + res=$($LFIND $dir0 -L released) + [[ $res == $f2 ]] || error "search released: found $res != $f2" + + res=$($LFIND $dir0 \! -L released) + [[ $res == $f1 ]] || error "search !released: found $res != $f1" + +} +run_test 56y "lfs find -L raid0|released" + test_57a() { [ $PARALLEL == "yes" ] && skip "skip parallel run" && return # note test will not do anything if MDS is not local @@ -4581,7 +4724,7 @@ test_57b() { local MDSFREE2=$(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 [ "$MDCFREE2" -lt "$((MDCFREE - 16))" ]; then if [ "$MDSFREE" != "$MDSFREE2" ]; then error "MDC before $MDCFREE != after $MDCFREE2" else @@ -4699,10 +4842,11 @@ test_62() { #run_test 62 "verify obd_match failure doesn't LBUG (should -EIO)" # bug 2319 - oig_wait() interrupted causes crash because of invalid waitq. -test_63a() { # was test_63 +# Though this test is irrelevant anymore, it helped to reveal some +# other grant bugs (LU-4482), let's keep it. +test_63a() { # was test_63 [ $PARALLEL == "yes" ] && skip "skip parallel run" && return MAX_DIRTY_MB=`lctl get_param -n osc.*.max_dirty_mb | head -n 1` - lctl set_param -n osc.*.max_dirty_mb 0 for i in `seq 10` ; do dd if=/dev/zero of=$DIR/f63 bs=8k & sleep 5 @@ -4710,7 +4854,6 @@ test_63a() { # was test_63 sleep 1 done - lctl set_param -n osc.*.max_dirty_mb $MAX_DIRTY_MB rm -f $DIR/f63 || true } run_test 63a "Verify oig_wait interruption does not crash =======" @@ -5690,6 +5833,7 @@ cleanup_101a() { test_101a() { [ $PARALLEL == "yes" ] && skip "skip parallel run" && return + [ $MDSCOUNT -ge 2 ] && skip "skip now for >= 2 MDTs" && return #LU-4322 local s local discard local nreads=10000 @@ -5804,126 +5948,118 @@ test_101c() { local STRIPE_SIZE=1048576 local FILE_LENGTH=$((STRIPE_SIZE*100)) local nreads=10000 - local osc + local osc_rpc_stats - setup_test101bc - - cancel_lru_locks osc - $LCTL set_param osc.*.rpc_stats 0 - $READS -f $DIR/$tfile -s$FILE_LENGTH -b65536 -n$nreads -t 180 - for osc in $($LCTL get_param -N osc.*); do - if [ "$osc" == "osc.num_refs" ]; then - continue - fi - - local lines=$($LCTL get_param -n ${osc}.rpc_stats | wc | awk '{print $1}') - if [ $lines -le 20 ]; then - continue - fi + setup_test101bc - local rpc4k=$($LCTL get_param -n ${osc}.rpc_stats | - awk '$1 == "1:" { print $2; exit; }') - local rpc8k=$($LCTL get_param -n ${osc}.rpc_stats | - awk '$1 == "2:" { print $2; exit; }') - local rpc16k=$($LCTL get_param -n ${osc}.rpc_stats | - awk '$1 == "4:" { print $2; exit; }') - local rpc32k=$($LCTL get_param -n ${osc}.rpc_stats | - awk '$1 == "8:" { print $2; exit; }') - - [ $rpc4k != 0 ] && error "Small 4k read IO ${rpc4k}!" - [ $rpc8k != 0 ] && error "Small 8k read IO ${rpc8k}!" - [ $rpc16k != 0 ] && error "Small 16k read IO ${rpc16k}!" - [ $rpc32k != 0 ] && error "Small 32k read IO ${rpc32k}!" - echo "${osc} rpc check passed!" - done - cleanup_test101bc - true + cancel_lru_locks osc + $LCTL set_param osc.*.rpc_stats 0 + $READS -f $DIR/$tfile -s$FILE_LENGTH -b65536 -n$nreads -t 180 + for osc_rpc_stats in $($LCTL get_param -N osc.*.rpc_stats); do + local stats=$($LCTL get_param -n $osc_rpc_stats) + local lines=$(echo "$stats" | awk 'END {print NR;}') + local size + + if [ $lines -le 20 ]; then + continue + fi + for size in 1 2 4 8; do + local rpc=$(echo "$stats" | + awk '($1 == "'$size':") {print $2; exit; }') + [ $rpc != 0 ] && + error "Small $((size*4))k read IO $rpc !" + done + echo "$osc_rpc_stats check passed!" + done + cleanup_test101bc + true } run_test 101c "check stripe_size aligned 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 + $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() { [ $PARALLEL == "yes" ] && skip "skip parallel run" && return local file=$DIR/$tfile - local size=${FILESIZE_101c:-500} + local sz_MB=${FILESIZE_101d:-500} local ra_MB=${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 $space" && return; } + local free_MB=$(($(df -P $DIR | tail -n 1 | awk '{ print $4 }') / 1024)) + [ $free_MB -lt $sz_MB ] && + skip "Need free space ${sz_MB}M, have ${free_MB}M" && return - echo Creating ${size}M test file $file - dd if=/dev/zero of=$file bs=1M count=$size || error "dd failed" - echo Cancel LRU locks on lustre client to flush the client cache - cancel_lru_locks osc + echo "Create test file $file size ${sz_MB}M, ${free_MB}M free" + $SETSTRIPE -c -1 $file || error "setstripe failed" - echo Disable read-ahead - local old_READAHEAD=$(set_read_ahead 0) + dd if=/dev/zero of=$file bs=1M count=$sz_MB || error "dd failed" + echo Cancel LRU locks on lustre client to flush the client cache + cancel_lru_locks osc - 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 Disable read-ahead + local old_READAHEAD=$(set_read_ahead 0) - 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 $file with read-ahead disabled + local raOFF=$(do_and_time "dd if=$file of=/dev/null bs=1M count=$sz_MB") + + 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 Reading the test file $file with read-ahead enabled + local raON=$(do_and_time "dd if=$file of=/dev/null bs=1M count=$sz_MB") - echo read-ahead disabled time read $time_ra_OFF - echo read-ahead enabled time read $time_ra_ON + echo "read-ahead disabled time read $raOFF" + echo "read-ahead enabled time read $raON" set_read_ahead $old_READAHEAD rm -f $file wait_delete_completed - [ $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" + [ $raOFF -le 1 -o $raON -lt $raOFF ] || + error "readahead ${raON}s > no-readahead ${raOFF}s ${sz_MB}M" } -run_test 101d "file read with and without read-ahead enabled =================" +run_test 101d "file read with and without read-ahead enabled" test_101e() { [ $PARALLEL == "yes" ] && skip "skip parallel run" && return - local file=$DIR/$tfile - local size=500 #KB - local count=100 - local blksize=1024 - - local space=$(df -P $DIR | tail -n 1 | awk '{ print $4 }') - local need_space=$((count * size)) - [ $space -gt $need_space ] || - { skip_env "Need free space $need_space, have $space" && return; } - - echo Creating $count ${size}K test files - for ((i = 0; i < $count; i++)); do - dd if=/dev/zero of=${file}_${i} bs=$blksize count=$size 2>/dev/null - done + local file=$DIR/$tfile + local size_KB=500 #KB + local count=100 + local bsize=1024 + + local free_KB=$(df -P $DIR | tail -n 1 | awk '{ print $4 }') + local need_KB=$((count * size_KB)) + [ $free_KB -le $need_KB ] && + skip_env "Need free space $need_KB, have $free_KB" && return + + echo "Creating $count ${size_KB}K test files" + for ((i = 0; i < $count; i++)); do + dd if=/dev/zero of=$file.$i bs=$bsize count=$size_KB 2>/dev/null + done - echo Cancel LRU locks on lustre client to flush the client cache - cancel_lru_locks osc + echo "Cancel LRU locks on lustre client to flush the client cache" + cancel_lru_locks osc - echo Reset readahead stats - $LCTL set_param -n llite.*.read_ahead_stats 0 + echo "Reset readahead stats" + $LCTL set_param -n llite.*.read_ahead_stats 0 - for ((i = 0; i < $count; i++)); do - dd if=${file}_${i} of=/dev/null bs=$blksize count=$size 2>/dev/null - done + for ((i = 0; i < $count; i++)); do + dd if=$file.$i of=/dev/null bs=$bsize count=$size_KB 2>/dev/null + done - local miss=$($LCTL get_param -n llite.*.read_ahead_stats | \ - get_named_value 'misses' | cut -d" " -f1 | calc_total) + local miss=$($LCTL get_param -n llite.*.read_ahead_stats | + get_named_value 'misses' | cut -d" " -f1 | calc_total) - for ((i = 0; i < $count; i++)); do - rm -rf ${file}_${i} 2>/dev/null - done + for ((i = 0; i < $count; i++)); do + rm -rf $file.$i 2>/dev/null + done - #10000 means 20% reads are missing in readahead - [ $miss -lt 10000 ] || error "misses too much for small reads" + #10000 means 20% reads are missing in readahead + [ $miss -lt 10000 ] || error "misses too much for small reads" } run_test 101e "check read-ahead for small read(1k) for small files(500k)" @@ -6285,10 +6421,87 @@ test_102l() { return 0; } -run_test 102l "listxattr filter test ==================================" +run_test 102l "listxattr size test ==================================" + +test_102m() { # LU-3403 llite: error of listxattr when buffer is small + local path=$DIR/$tfile + touch $path + + listxattr_size_check $path || error "listattr_size_check $path failed" +} +run_test 102m "Ensure listxattr fails on small bufffer ========" cleanup_test102 +getxattr() { # getxattr path name + # Return the base64 encoding of the value of xattr name on path. + local path=$1 + local name=$2 + + # # getfattr --absolute-names --encoding=base64 --name=trusted.lov $path + # file: $path + # trusted.lov=0s0AvRCwEAAAAGAAAAAAAAAAAEAAACAAAAAAAQAAEAA...AAAAAAAAA= + # + # We print just 0s0AvRCwEAAAAGAAAAAAAAAAAEAAACAAAAAAAQAAEAA...AAAAAAAAA= + + getfattr --absolute-names --encoding=base64 --name=$name $path | + awk -F= -v name=$name '$1 == name { + print substr($0, index($0, "=") + 1); + }' +} + +test_102n() { # LU-4101 mdt: protect internal xattrs + local file0=$DIR/$tfile.0 + local file1=$DIR/$tfile.1 + local xattr0=$TMP/$tfile.0 + local xattr1=$TMP/$tfile.1 + local name + local value + + if [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.5.50) ] + then + skip "MDT < 2.5.50 allows setxattr on internal trusted xattrs" + return + fi + + rm -rf $file0 $file1 $xattr0 $xattr1 + touch $file0 $file1 + + # Get 'before' xattrs of $file1. + getfattr --absolute-names --dump --match=- $file1 > $xattr0 + + for name in lov lma lmv link fid version som hsm lfsck_namespace; do + # Try to copy xattr from $file0 to $file1. + value=$(getxattr $file0 trusted.$name 2> /dev/null) + + setfattr --name=trusted.$name --value="$value" $file1 || + error "setxattr 'trusted.$name' failed" + + # Try to set a garbage xattr. + value=0sVGhlIHF1aWNrIGJyb3duIGZveCBqdW1wcyBvdmVyIGl0c2VsZi4= + + setfattr --name=trusted.$name --value="$value" $file1 || + error "setxattr 'trusted.$name' failed" + + # Try to remove the xattr from $file1. We don't care if this + # appears to succeed or fail, we just don't want there to be + # any changes or crashes. + setfattr --remove=$trusted.$name $file1 2> /dev/null + done + + # Get 'after' xattrs of file1. + getfattr --absolute-names --dump --match=- $file1 > $xattr1 + + if ! diff $xattr0 $xattr1; then + error "before and after xattrs of '$file1' differ" + fi + + rm -rf $file0 $file1 $xattr0 $xattr1 + + return 0 +} +run_test 102n "silently ignore setxattr on internal trusted xattrs" + run_acl_subtest() { $LUSTRE/tests/acl/run $LUSTRE/tests/acl/$1.test @@ -6550,53 +6763,79 @@ test_116a() { # was previously test_116() [ "$OSTCOUNT" -lt "2" ] && skip_env "$OSTCOUNT < 2 OSTs" && return echo -n "Free space priority " - lctl get_param -n lov.*-clilov-*.qos_prio_free + do_facet $SINGLEMDS lctl get_param -n lo*.*-mdtlov.qos_prio_free | + head -1 declare -a AVAIL free_min_max - [ $MINV -gt 960000 ] && skip "too much free space in OST$MINI, skip" &&\ - return - # generate uneven OSTs + [ $MINV -eq 0 ] && skip "no free space in OST$MINI, skip" && return + trap simple_cleanup_common EXIT + + # Check if we need to generate uneven OSTs test_mkdir -p $DIR/$tdir/OST${MINI} - declare -i FILL - FILL=$(($MINV / 4)) - echo "Filling 25% remaining space in OST${MINI} with ${FILL}Kb" - $SETSTRIPE -i $MINI -c 1 $DIR/$tdir/OST${MINI}||error "setstripe failed" - i=0 - while [ $FILL -gt 0 ]; do - i=$(($i + 1)) - dd if=/dev/zero of=$DIR/$tdir/OST${MINI}/$tfile-$i bs=2M count=1 2>/dev/null - FILL=$(($FILL - 2048)) - echo -n . - done - FILL=$(($MINV / 4)) - sync - sleep_maxage + local FILL=$(($MINV / 4)) + local DIFF=$(($MAXV - $MINV)) + local DIFF2=$(($DIFF * 100 / $MINV)) + + local threshold=$(do_facet $SINGLEMDS \ + lctl get_param -n *.*MDT0000-mdtlov.qos_threshold_rr | head -1) + threshold=${threshold%%%} + echo -n "Check for uneven OSTs: " + echo -n "diff=${DIFF}KB (${DIFF2}%) must be > ${threshold}% ..." + + if [ $DIFF2 -gt $threshold ]; then + echo "ok" + echo "Don't need to fill OST$MINI" + else + # generate uneven OSTs. Write 2% over the QOS threshold value + echo "no" + DIFF=$(($threshold - $DIFF2 + 2)) + DIFF2=$(( ($MINV * $DIFF)/100 )) + echo "Fill ${DIFF}% remaining space in OST${MINI} with ${DIFF2}KB" + $SETSTRIPE -i $MINI -c 1 $DIR/$tdir/OST${MINI} || + error "setstripe failed" + DIFF=$(($DIFF2 / 2048)) + i=0 + while [ $i -lt $DIFF ]; do + i=$(($i + 1)) + dd if=/dev/zero of=$DIR/$tdir/OST${MINI}/$tfile-$i \ + bs=2M count=1 2>/dev/null + echo -n . + done + echo . + sync + sleep_maxage + free_min_max + fi - free_min_max DIFF=$(($MAXV - $MINV)) DIFF2=$(($DIFF * 100 / $MINV)) - echo -n "diff=${DIFF}=${DIFF2}% must be > 20% for QOS mode..." - if [ $DIFF2 -gt 20 ]; then - echo "ok" + echo -n "diff=${DIFF}=${DIFF2}% must be > ${threshold}% for QOS mode..." + if [ $DIFF2 -gt $threshold ]; then + echo "ok" else - echo "failed - QOS mode won't be used" - error_ignore "QOS imbalance criteria not met" - return + echo "failed - QOS mode won't be used" + skip "QOS imbalance criteria not met" + simple_cleanup_common + return fi MINI1=$MINI; MINV1=$MINV MAXI1=$MAXI; MAXV1=$MAXV # now fill using QOS - echo writing a bunch of files to QOS-assigned OSTs $SETSTRIPE -c 1 $DIR/$tdir + FILL=$(($FILL / 200)) + if [ $FILL -gt 600 ]; then + FILL=600 + fi + echo "writing $FILL files to QOS-assigned OSTs" i=0 - while [ $FILL -gt 0 ]; do - i=$(($i + 1)) - dd if=/dev/zero of=$DIR/$tdir/$tfile-$i bs=1024 count=200 2>/dev/null - FILL=$(($FILL - 200)) - echo -n . + while [ $i -lt $FILL ]; do + i=$(($i + 1)) + dd if=/dev/zero of=$DIR/$tdir/$tfile-$i bs=200k \ + count=1 2>/dev/null + echo -n . done echo "wrote $i 200k files" sync @@ -6608,26 +6847,30 @@ test_116a() { # was previously test_116() echo "free space delta: orig $DIFF final $DIFF2" [ $DIFF2 -gt $DIFF ] && echo "delta got worse!" DIFF=$(($MINV1 - ${AVAIL[$MINI1]})) - echo "Wrote $DIFF to smaller OST $MINI1" + echo "Wrote ${DIFF}KB to smaller OST $MINI1" DIFF2=$(($MAXV1 - ${AVAIL[$MAXI1]})) - echo "Wrote $DIFF2 to larger OST $MAXI1" - [ $DIFF -gt 0 ] && echo "Wrote $(($DIFF2 * 100 / $DIFF - 100))% more data to larger OST $MAXI1" + echo "Wrote ${DIFF2}KB to larger OST $MAXI1" + FILL=$(($DIFF2 * 100 / $DIFF - 100)) + [ $DIFF -gt 0 ] && + echo "Wrote ${FILL}% more data to larger OST $MAXI1" # Figure out which files were written where UUID=$(lctl get_param -n lov.${FSNAME}-clilov-*.target_obd | - awk '/'$MINI1': / {print $2; exit}') + awk '/'$MINI1': / {print $2; exit}') echo $UUID - MINC=$($GETSTRIPE --obd $UUID $DIR/$tdir | wc -l) + MINC=$($GETSTRIPE --ost $UUID $DIR/$tdir | grep $DIR | wc -l) echo "$MINC files created on smaller OST $MINI1" UUID=$(lctl get_param -n lov.${FSNAME}-clilov-*.target_obd | - awk '/'$MAXI1': / {print $2; exit}') + awk '/'$MAXI1': / {print $2; exit}') echo $UUID - MAXC=$($GETSTRIPE --obd $UUID $DIR/$tdir | wc -l) + MAXC=$($GETSTRIPE --ost $UUID $DIR/$tdir | grep $DIR | wc -l) echo "$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 + FILL=$(($MAXC * 100 / $MINC - 100)) + [ $MINC -gt 0 ] && + echo "Wrote ${FILL}% more files to larger OST $MAXI1" + [ $MAXC -gt $MINC ] || + error_ignore LU-9 "stripe QOS didn't balance free space" + simple_cleanup_common } run_test 116a "stripe QOS: free space balance ===================" @@ -6657,7 +6900,7 @@ test_117() # bug 10891 echo "Truncate succeeded." rm -f $DIR/$tfile } -run_test 117 "verify fsfilt_extend ==========" +run_test 117 "verify osd extend ==========" NO_SLOW_RESENDCOUNT=4 export OLD_RESENDCOUNT="" @@ -7742,11 +7985,15 @@ test_129() { ENOSPC=28 EFBIG=27 + rm -rf $DIR/$tdir test_mkdir -p $DIR/$tdir - MAX=$(stat -c%s "$DIR/$tdir") + # block size of mds1 + local MDT_DEV=$(mdsdevname ${SINGLEMDS//mds/}) + local MDSBLOCKSIZE=$($LCTL get_param -n mdc.*MDT0000*.blocksize) + local MAX=$((MDSBLOCKSIZE * 3)) set_dir_limits $MAX - local I=0 + local I=$(stat -c%s "$DIR/$tdir") local J=0 while [ ! $I -gt $MAX ]; do $MULTIOP $DIR/$tdir/$J Oc @@ -7754,18 +8001,22 @@ test_129() { #check two errors ENOSPC for new version of ext4 max_dir_size patch #mainline kernel commit df981d03eeff7971ac7e6ff37000bfa702327ef1 #and EFBIG for previous versions - if [ $rc -eq $EFBIG -o $rc -eq $ENOSPC ] && [ $I -gt 0 ]; then + if [ $rc -eq $EFBIG -o $rc -eq $ENOSPC ]; then set_dir_limits 0 echo "return code $rc received as expected" - multiop $DIR/$tdir/$J Oc - rc=$? + multiop $DIR/$tdir/$J Oc || + error_exit "multiop failed w/o dir size limit" + I=$(stat -c%s "$DIR/$tdir") - if [ $I -gt $MAX ] && [ $rc -eq 0 ]; then - return 0 + + if [ $(lustre_version_code $SINGLEMDS) -lt \ + $(version_code 2.4.51) ] + then + [ $I -eq $MAX ] && return 0 else - error_exit "return code $rc current dir size $I " \ - "previous limit $MAX" + [ $I -gt $MAX ] && return 0 fi + error_exit "current dir size $I, previous limit $MAX" elif [ $rc -ne 0 ]; then set_dir_limits 0 error_exit "return code $rc received instead of expected " \ @@ -7973,10 +8224,14 @@ test_130d() { $SETSTRIPE -S 65536 -c $OSTCOUNT $fm_file||error "setstripe on $fm_file" [ "$(facet_fstype ost$(($($GETSTRIPE -i $fm_file) + 1)))" = "zfs" ] && skip "ORI-366/LU-1941: FIEMAP unimplemented on ZFS" && return - dd if=/dev/zero of=$fm_file bs=1M count=$OSTCOUNT || error "dd failed on $fm_file" + + local actual_stripecnt=$($GETSTRIPE -c $fm_file) + dd if=/dev/zero of=$fm_file bs=1M count=$actual_stripecnt || + error "dd failed on $fm_file" filefrag -ves $fm_file || error "filefrag $fm_file failed" - filefrag_op=`filefrag -ve $fm_file | grep -A 100 "ext:" | grep -v "ext:" | grep -v "found"` + filefrag_op=`filefrag -ve $fm_file | grep -A 100 "ext:" | + grep -v "ext:" | grep -v "found"` last_lun=`echo $filefrag_op | cut -d: -f5` @@ -8000,7 +8255,7 @@ test_130d() { (( tot_len += ext_len )) last_lun=$frag_lun done - if (( num_luns != OSTCOUNT || tot_len != 1024 )); then + if (( num_luns != actual_stripecnt || tot_len != 1024 )); then cleanup_130 error "FIEMAP on $fm_file failed; returned wrong number of luns or wrong len for OST $last_lun" return @@ -8587,12 +8842,48 @@ test_150() { } run_test 150 "truncate/append tests" +#LU-2902 roc_hit was not able to read all values from lproc +function roc_hit_init() { + local list=$(comma_list $(osts_nodes)) + local dir=$DIR/$tdir-check + local file=$dir/file + local BEFORE + local AFTER + local idx + + test_mkdir -p $dir + #use setstripe to do a write to every ost + for i in $(seq 0 $((OSTCOUNT-1))); do + $SETSTRIPE -c 1 -i $i $dir || error "$SETSTRIPE $file failed" + dd if=/dev/urandom of=$file bs=4k count=4 2>&1 > /dev/null + idx=$(printf %04x $i) + BEFORE=$(get_osd_param $list *OST*$idx stats | + awk '$1 == "cache_access" {sum += $2} END { print sum }') + if [ -z "$BEFORE" ]; then + BEFORE=0 + fi + + cancel_lru_locks osc + cat $file >/dev/null + + AFTER=$(get_osd_param $list *OST*$idx stats | + awk '$1 == "cache_access" {sum += $2} END { print sum }') + + echo BEFORE:$BEFORE AFTER:$AFTER + if ! let "AFTER - BEFORE == 4"; then + rm -rf $dir + error "roc_hit is not safe to use" + fi + rm $file + done + + rm -rf $dir +} + function roc_hit() { local list=$(comma_list $(osts_nodes)) - #debug temp debug for LU-2902: lets see what values we get back - echo $(get_osd_param $list '' stats) 1>&2 echo $(get_osd_param $list '' stats | - awk '/'cache_hit'/ {sum+=$2} END {print sum}') + awk '$1 == "cache_hit" {sum += $2} END { print sum }') } function set_cache() { @@ -8634,6 +8925,8 @@ test_151() { return 0 fi + roc_hit_init + #define OBD_FAIL_OBD_NO_LRU 0x609 do_nodes $list $LCTL set_param fail_loc=0x609 @@ -8918,6 +9211,39 @@ test_154c() { } run_test 154c "lfs path2fid and fid2path multiple arguments" +test_154d() { + [[ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.5.53) ]] && + skip "Need MDS version at least 2.5.53" && return + + if remote_mds; then + nid=$($LCTL list_nids | sed "s/\./\\\./g") + else + nid="0@lo" + fi + local proc_ofile="mdt.*.exports.'$nid'.open_files" + local fd + local cmd + + rm -f $DIR/$tfile + touch $DIR/$tfile + + fid=$($LFS path2fid $DIR/$tfile) + # Open the file + fd=$(free_fd) + cmd="exec $fd<$DIR/$tfile" + eval $cmd + fid_list=$(do_facet $SINGLEMDS $LCTL get_param $proc_ofile) + echo $fid_list | grep $fid + rc=$? + + cmd="exec $fd>/dev/null" + eval $cmd + if [ $rc -ne 0 ]; then + error "FID $fid not found in open files list $fid_list" + fi +} +run_test 154d "Verify open file fid" + test_155_small_load() { local temp=$TMP/$tfile local file=$DIR/$tfile @@ -9049,6 +9375,8 @@ test_156() { skip "LU-1956/LU-2261: stats unimplemented on OSD ZFS" && return + roc_hit_init + log "Turn on read and write cache" set_cache read on set_cache writethrough on @@ -9141,12 +9469,11 @@ test_156() { BEFORE=`roc_hit` cat $file >/dev/null AFTER=`roc_hit` - if ! let "AFTER - BEFORE == 0"; then - error_ignore 20762 "IN CACHE: before: $BEFORE, after: $AFTER" - else - log "cache hits:: before: $BEFORE, after: $AFTER" - fi - + if ! let "AFTER - BEFORE == 0"; then + error_ignore bz20762 "IN CACHE: before: $BEFORE, after: $AFTER" + else + log "cache hits:: before: $BEFORE, after: $AFTER" + fi log "Turn on the read cache and turn off the write cache" set_cache read on @@ -9160,11 +9487,11 @@ test_156() { cancel_lru_locks osc cat $file >/dev/null AFTER=`roc_hit` - if ! let "AFTER - BEFORE == 0"; then - error_ignore 20762 "IN CACHE: before: $BEFORE, after: $AFTER" - else - log "cache hits:: before: $BEFORE, after: $AFTER" - fi + if ! let "AFTER - BEFORE == 0"; then + error_ignore bz20762 "IN CACHE: before: $BEFORE, after: $AFTER" + else + log "cache hits:: before: $BEFORE, after: $AFTER" + fi log "Read again; it should be satisfied from the cache." BEFORE=`roc_hit` @@ -9184,7 +9511,7 @@ run_test 156 "Verification of tunables ============================" #Changelogs err17935 () { if [ $MDSCOUNT -gt 1 ]; then - error_ignore 17935 $* + error_ignore bz17935 $* else error $* fi @@ -9203,7 +9530,18 @@ changelog_chmask() fi } -test_160() { +changelog_extract_field() { + local mdt=$1 + local cltype=$2 + local file=$3 + local identifier=$4 + + $LFS changelog $mdt | gawk "/$cltype.*$file$/ { + print gensub(/^.* "$identifier'(\[[^\]]*\]).*$/,"\\1",1)}' | + tail -1 +} + +test_160a() { [ $PARALLEL == "yes" ] && skip "skip parallel run" && return remote_mds_nodsh && skip "remote MDS with nodsh" && return [ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.2.0) ] || @@ -9248,14 +9586,14 @@ test_160() { # verify contents echo "verifying target fid" - fidc=$($LFS changelog $MDT0|awk '/CREAT.*timestamp$/{print $6}'|tail -1) + fidc=$(changelog_extract_field $MDT0 "CREAT" "timestamp" "t=") fidf=$($LFS path2fid $DIR/$tdir/pics/zach/timestamp) - [ "$fidc" == "t=$fidf" ] || + [ "$fidc" == "$fidf" ] || err17935 "fid in changelog $fidc != file fid $fidf" echo "verifying parent fid" - fidc=$($LFS changelog $MDT0|awk '/CREAT.*timestamp$/{print $7}'|tail -1) + fidc=$(changelog_extract_field $MDT0 "CREAT" "timestamp" "p=") fidf=$($LFS path2fid $DIR/$tdir/pics/zach) - [ "$fidc" == "p=$fidf" ] || + [ "$fidc" == "$fidf" ] || err17935 "pfid in changelog $fidc != dir fid $fidf" USER_REC1=$($GET_CL_USERS | awk "\$1 == \"$USER\" {print \$2}") @@ -9299,7 +9637,38 @@ test_160() { echo "$USERS other changelog users; can't verify off" fi } -run_test 160 "changelog sanity" +run_test 160a "changelog sanity" + +test_160b() { # LU-3587 + [ $PARALLEL == "yes" ] && skip "skip parallel run" && return + remote_mds_nodsh && skip "remote MDS with nodsh" && return + [ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.2.0) ] || + { skip "Need MDS version at least 2.2.0"; return; } + + local CL_USERS="mdd.$MDT0.changelog_users" + local GET_CL_USERS="do_facet $SINGLEMDS $LCTL get_param -n $CL_USERS" + USER=$(do_facet $SINGLEMDS $LCTL --device $MDT0 changelog_register -n) + echo "Registered as changelog user $USER" + $GET_CL_USERS | grep -q $USER || + error "User $USER not found in changelog_users" + + local LONGNAME1=$(str_repeat a 255) + local LONGNAME2=$(str_repeat b 255) + + cd $DIR + echo "creating very long named file" + touch $LONGNAME1 || error "create of $LONGNAME1 failed" + echo "moving very long named file" + mv $LONGNAME1 $LONGNAME2 + + $LFS changelog $MDT0 | grep RENME + + echo "deregistering $USER" + do_facet $SINGLEMDS $LCTL --device $MDT0 changelog_deregister $USER + + rm -f $LONGNAME2 +} +run_test 160b "Verify that very long rename doesn't crash in changelog" test_161a() { [ $PARALLEL == "yes" ] && skip "skip parallel run" && return @@ -9395,6 +9764,93 @@ test_161b() { } run_test 161b "link ea sanity under remote directory" +test_161c() { + [ $PARALLEL == "yes" ] && skip "skip parallel run" && return + [[ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.1.5) ]] && + skip "Need MDS version at least 2.1.5" && return + + # define CLF_RENAME_LAST 0x0001 + # rename overwrite a target having nlink = 1 (changelog flag 0x1) + local USER=$(do_facet $SINGLEMDS $LCTL --device $MDT0 \ + changelog_register -n) + rm -rf $DIR/$tdir + mkdir -p $DIR/$tdir + touch $DIR/$tdir/foo_161c + touch $DIR/$tdir/bar_161c + mv -f $DIR/$tdir/foo_161c $DIR/$tdir/bar_161c + $LFS changelog $MDT0 | grep RENME + local flags=$($LFS changelog $MDT0 | grep RENME | tail -1 | \ + cut -f5 -d' ') + $LFS changelog_clear $MDT0 $USER 0 + if [ x$flags != "x0x1" ]; then + do_facet $SINGLEMDS $LCTL --device $MDT0 changelog_deregister \ + $USER + error "flag $flags is not 0x1" + fi + echo "rename overwrite a target having nlink = 1," \ + "changelog record has flags of $flags" + + # rename overwrite a target having nlink > 1 (changelog flag 0x0) + touch $DIR/$tdir/foo_161c + touch $DIR/$tdir/bar_161c + ln $DIR/$tdir/bar_161c $DIR/$tdir/foobar_161c + mv -f $DIR/$tdir/foo_161c $DIR/$tdir/bar_161c + $LFS changelog $MDT0 | grep RENME + flags=$($LFS changelog $MDT0 | grep RENME | tail -1 | cut -f5 -d' ') + $LFS changelog_clear $MDT0 $USER 0 + if [ x$flags != "x0x0" ]; then + do_facet $SINGLEMDS $LCTL --device $MDT0 changelog_deregister \ + $USER + error "flag $flags is not 0x0" + fi + echo "rename overwrite a target having nlink > 1," \ + "changelog record has flags of $flags" + + # rename doesn't overwrite a target (changelog flag 0x0) + touch $DIR/$tdir/foo_161c + mv -f $DIR/$tdir/foo_161c $DIR/$tdir/foo2_161c + $LFS changelog $MDT0 | grep RENME + flags=$($LFS changelog $MDT0 | grep RENME | tail -1 | cut -f5 -d' ') + $LFS changelog_clear $MDT0 $USER 0 + if [ x$flags != "x0x0" ]; then + do_facet $SINGLEMDS $LCTL --device $MDT0 changelog_deregister \ + $USER + error "flag $flags is not 0x0" + fi + echo "rename doesn't overwrite a target," \ + "changelog record has flags of $flags" + + # define CLF_UNLINK_LAST 0x0001 + # unlink a file having nlink = 1 (changelog flag 0x1) + rm -f $DIR/$tdir/foo2_161c + $LFS changelog $MDT0 | grep UNLNK + flags=$($LFS changelog $MDT0 | grep UNLNK | tail -1 | cut -f5 -d' ') + $LFS changelog_clear $MDT0 $USER 0 + if [ x$flags != "x0x1" ]; then + do_facet $SINGLEMDS $LCTL --device $MDT0 changelog_deregister \ + $USER + error "flag $flags is not 0x1" + fi + echo "unlink a file having nlink = 1," \ + "changelog record has flags of $flags" + + # unlink a file having nlink > 1 (changelog flag 0x0) + ln -f $DIR/$tdir/bar_161c $DIR/$tdir/foobar_161c + rm -f $DIR/$tdir/foobar_161c + $LFS changelog $MDT0 | grep UNLNK + flags=$($LFS changelog $MDT0 | grep UNLNK | tail -1 | cut -f5 -d' ') + $LFS changelog_clear $MDT0 $USER 0 + if [ x$flags != "x0x0" ]; then + do_facet $SINGLEMDS $LCTL --device $MDT0 changelog_deregister \ + $USER + error "flag $flags is not 0x0" + fi + echo "unlink a file having nlink > 1," \ + "changelog record has flags of $flags" + do_facet $SINGLEMDS $LCTL --device $MDT0 changelog_deregister $USER +} +run_test 161c "check CL_RENME[UNLINK] changelog record flags" + check_path() { local expected=$1 shift @@ -9457,20 +9913,6 @@ test_162() { } run_test 162 "path lookup sanity" -test_163() { - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return - remote_mds_nodsh && skip "remote MDS with nodsh" && return - copytool --test $FSNAME || { skip "copytool not runnable: $?" && return; } - copytool $FSNAME & - sleep 1 - local uuid=$($LCTL get_param -n mdc.${FSNAME}-MDT0000-mdc-*.uuid) - # this proc file is temporary and linux-only - do_facet $SINGLEMDS lctl set_param mdt.${FSNAME}-MDT0000.mdccomm=$uuid ||\ - error "kernel->userspace send failed" - kill -INT $! -} -run_test 163 "kernel <-> userspace comms" - test_169() { # do directio so as not to populate the page cache log "creating a 10 Mb file" @@ -9590,6 +10032,15 @@ obdecho_test() { local pages=${3:-64} local rc=0 local id + + local count=10 + local obd_size=$(get_obd_size $node $OBD) + local page_size=$(get_page_size $node) + if [[ -n "$obd_size" ]]; then + local new_count=$((obd_size / (pages * page_size / 1024))) + [[ $new_count -ge $count ]] || count=$new_count + fi + 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; } @@ -9601,7 +10052,7 @@ obdecho_test() { [ $rc -eq 0 ] && { do_facet $node "$LCTL --device ec getattr $id" || rc=4; } [ $rc -eq 0 ] && { do_facet $node "$LCTL --device ec " \ - "test_brw 10 w v $pages $id" || rc=4; } + "test_brw $count w v $pages $id" || rc=4; } [ $rc -eq 0 ] && { do_facet $node "$LCTL --device ec destroy $id 1" || rc=4; } [ $rc -eq 0 -o $rc -gt 2 ] && { do_facet $node "$LCTL --device ec " \ @@ -10565,6 +11016,31 @@ test_208() { } run_test 208 "Exclusive open" +test_209() { + [ -z "$(lctl get_param -n mdc.*.connect_flags | grep disp_stripe)" ] && + skip_env "must have disp_stripe" && return + + touch $DIR/$tfile + sync; sleep 5; sync; + + echo 3 > /proc/sys/vm/drop_caches + req_before=$(awk '/ptlrpc_cache / { print $2 }' /proc/slabinfo) + + # open/close 500 times + for i in $(seq 500); do + cat $DIR/$tfile + done + + echo 3 > /proc/sys/vm/drop_caches + req_after=$(awk '/ptlrpc_cache / { print $2 }' /proc/slabinfo) + + echo "before: $req_before, after: $req_after" + [ $((req_after - req_before)) -ge 300 ] && + error "open/close requests are not freed" + return 0 +} +run_test 209 "read-only open/close requests should be freed promptly" + test_212() { size=`date +%s` size=$((size % 8192 + 1)) @@ -10586,14 +11062,15 @@ test_213() { run_test 213 "OSC lock completion and cancel race don't crash - bug 18829" test_214() { # for bug 20133 - test_mkdir -p $DIR/d214p/d214c + mkdir -p $DIR/$tdir/d214c || error "mkdir $DIR/$tdir/d214c failed" for (( i=0; i < 340; i++ )) ; do - touch $DIR/d214p/d214c/a$i + touch $DIR/$tdir/d214c/a$i done - ls -l $DIR/d214p || error "ls -l $DIR/d214p failed" - mv $DIR/d214p/d214c $DIR/ || error "mv $DIR/d214p/d214c $DIR/ failed" + ls -l $DIR/$tdir || error "ls -l $DIR/d214p failed" + mv $DIR/$tdir/d214c $DIR/ || error "mv $DIR/d214p/d214c $DIR/ failed" ls $DIR/d214c || error "ls $DIR/d214c failed" + rm -rf $DIR/$tdir || error "rm -rf $DIR/d214* failed" rm -rf $DIR/d214* || error "rm -rf $DIR/d214* failed" } run_test 214 "hash-indexed directory test - bug 20133" @@ -10665,12 +11142,13 @@ test_215() { # for bugs 18102, 21079, 21517 # /proc/sys/lnet/routes should look like this: # Routing disabled/enabled - # net hops state router - # where net is a string like tcp0, hops >= 0, state is up/down, + # net hops priority state router + # where net is a string like tcp0, hops > 0, priority >= 0, + # state is up/down, # router is a string like 192.168.1.1@tcp2 L1="^Routing (disabled|enabled)$" - L2="^net +hops +state +router$" - BR="^$NET +$N +(up|down) +$NID$" + L2="^net +hops +priority +state +router$" + BR="^$NET +$N +(0|1) +(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" @@ -10817,6 +11295,10 @@ test_219() { dd if=/dev/zero of=$DIR/$tfile bs=4096 count=1 seek=3 $LCTL set_param fail_loc=0x411 dd if=/dev/zero of=$DIR/$tfile bs=1024 count=1 seek=2 conv=notrunc + + # LU-4201 + dd if=/dev/zero of=$DIR/$tfile-2 bs=1024 count=1 + $CHECKSTAT -s 1024 $DIR/$tfile-2 || error "checkstat wrong size" } run_test 219 "LU-394: Write partial won't cause uncontiguous pages vec at LND" @@ -11234,6 +11716,9 @@ test_228c() { run_test 228c "NOT shrink the last entry in OI index node to recycle idle leaf" test_229() { # LU-2482, LU-3448 + [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.4.53) ] && + skip "No HSM support on MDS of $(get_lustre_version)," \ + "need 2.4.53 at least" && return [ $PARALLEL == "yes" ] && skip "skip parallel run" && return [ $OSTCOUNT -lt 2 ] && skip "needs >= 2 OSTs" && return @@ -11245,34 +11730,20 @@ test_229() { # LU-2482, LU-3448 $GETSTRIPE -v $DIR/$tfile - local pattern=$($GETSTRIPE -v $DIR/$tfile | - grep lmm_stripe_pattern | awk '{print $2}') + local pattern=$($GETSTRIPE -L $DIR/$tfile) [ X"$pattern" = X"80000001" ] || error "pattern error ($pattern)" local stripe_count=$($GETSTRIPE -c $DIR/$tfile) || error "getstripe" [ $stripe_count -eq 2 ] || error "stripe count not 2 ($stripe_count)" stat $DIR/$tfile || error "failed to stat released file" - $TRUNCATE $DIR/$tfile 200000 && - error "truncate of released file should fail" - - # Ensure that nothing happened anyway. - $CHECKSTAT -s 0 $DIR/$tfile || - error "released file size should not change" - - # Stripe count should be no change after truncate - stripe_count=$($GETSTRIPE -c $DIR/$tfile) || error "getstripe failed" - [ $stripe_count -eq 2 ] || error "after trunc: ($stripe_count)" - chown $RUNAS_ID $DIR/$tfile || error "chown $RUNAS_ID $DIR/$tfile failed" chgrp $RUNAS_ID $DIR/$tfile || error "chgrp $RUNAS_ID $DIR/$tfile failed" - touch $DIR/$tfile || - error "touch $DIR/$tfile failed" - + touch $DIR/$tfile || error "touch $DIR/$tfile failed" rm $DIR/$tfile || error "failed to remove released file" } run_test 229 "getstripe/stat/rm/attr changes work on released files" @@ -11301,44 +11772,6 @@ test_230a() { } run_test 230a "Create remote directory and files under the remote directory" -test_230b() { - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return - [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return - local MDTIDX=1 - local remote_dir=$DIR/$tdir/remote_dir - local rc=0 - - mkdir -p $DIR/$tdir - $LFS mkdir -i $MDTIDX $remote_dir || - error "create remote directory failed" - - $LFS mkdir -i 0 $remote_dir/new_dir && - error "nested remote directory create succeed!" - - do_facet mds$((MDTIDX + 1)) lctl set_param mdt.*.enable_remote_dir=1 - $LFS mkdir -i 0 $remote_dir/new_dir || rc=$? - do_facet mds$((MDTIDX + 1)) lctl set_param mdt.*.enable_remote_dir=0 - - [ $rc -ne 0 ] && - error "create remote directory failed after set enable_remote_dir" - - rm -rf $remote_dir || error "first unlink remote directory failed" - - $RUNAS -G$RUNAS_GID $LFS mkdir -i $MDTIDX $DIR/$tfile && - error "chown worked" - - do_facet mds$MDTIDX lctl set_param \ - mdt.*.enable_remote_dir_gid=$RUNAS_GID - $LFS mkdir -i $MDTIDX $remote_dir || rc=$? - do_facet mds$MDTIDX lctl set_param mdt.*.enable_remote_dir_gid=0 - - [ $rc -ne 0 ] && - error "create remote dir failed after set enable_remote_dir_gid" - - rm -r $DIR/$tdir || error "second unlink remote directory failed" -} -run_test 230b "nested remote directory should be failed" - test_231a() { # For simplicity this test assumes that max_pages_per_rpc @@ -11403,12 +11836,109 @@ test_232() { } run_test 232 "failed lock should not block umount" -test_233() { +test_233a() { + [ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.3.64) ] || + { skip "Need MDS version at least 2.3.64"; return; } + local fid=$($LFS path2fid $MOUNT) stat $MOUNT/.lustre/fid/$fid > /dev/null || error "cannot access $MOUNT using its FID '$fid'" } -run_test 233 "checking that OBF of the FS root succeeds" +run_test 233a "checking that OBF of the FS root succeeds" + +test_233b() { + [ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.5.90) ] || + { skip "Need MDS version at least 2.5.90"; return; } + + local fid=$($LFS path2fid $MOUNT/.lustre) + stat $MOUNT/.lustre/fid/$fid > /dev/null || + error "cannot access $MOUNT/.lustre using its FID '$fid'" + + fid=$($LFS path2fid $MOUNT/.lustre/fid) + stat $MOUNT/.lustre/fid/$fid > /dev/null || + error "cannot access $MOUNT/.lustre/fid using its FID '$fid'" +} +run_test 233b "checking that OBF of the FS .lustre succeeds" + +test_234() { + local p="$TMP/sanityN-$TESTNAME.parameters" + save_lustre_params client "llite.*.xattr_cache" > $p + lctl set_param llite.*.xattr_cache 1 || + { skip "xattr cache is not supported"; return 0; } + + mkdir -p $DIR/$tdir || error "mkdir failed" + touch $DIR/$tdir/$tfile || error "touch failed" + # OBD_FAIL_LLITE_XATTR_ENOMEM + $LCTL set_param fail_loc=0x1405 + if [ ! -f /etc/SuSE-release ]; then + # attr pre-2.4.44-7 had a bug with rc + # LU-3703 - SLES clients have older attr + getfattr -n user.attr $DIR/$tdir/$tfile && + error "getfattr should have failed with ENOMEM" + fi + $LCTL set_param fail_loc=0x0 + rm -rf $DIR/$tdir + + restore_lustre_params < $p + rm -f $p +} +run_test 234 "xattr cache should not crash on ENOMEM" + +test_235() { + [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.4.52) ] && + skip "Need MDS version at least 2.4.52" && return + flock_deadlock $DIR/$tfile + local RC=$? + case $RC in + 0) + ;; + 124) error "process hangs on a deadlock" + ;; + *) error "error executing flock_deadlock $DIR/$tfile" + ;; + esac +} +run_test 235 "LU-1715: flock deadlock detection does not work properly" + +#LU-2935 +test_236() { + check_swap_layouts_support && return 0 + test_mkdir -p $DIR/$tdir || error "mkdir $tdir failed" + + local ref1=/etc/passwd + local ref2=/etc/group + local file1=$DIR/$tdir/f1 + local file2=$DIR/$tdir/f2 + + $SETSTRIPE -c 1 $file1 || error "cannot setstripe on '$file1': rc = $?" + cp $ref1 $file1 || error "cp $ref1 $file1 failed: rc = $?" + $SETSTRIPE -c 2 $file2 || error "cannot setstripe on '$file2': rc = $?" + cp $ref2 $file2 || error "cp $ref2 $file2 failed: rc = $?" + local fd=$(free_fd) + local cmd="exec $fd<>$file2" + eval $cmd + rm $file2 + $LFS swap_layouts $file1 /proc/self/fd/${fd} || + error "cannot swap layouts of '$file1' and /proc/self/fd/${fd}" + cmd="exec $fd>&-" + eval $cmd + cmp $ref2 $file1 || error "content compare failed ($ref2 != $file1)" + + #cleanup + rm -rf $DIR/$tdir +} +run_test 236 "Layout swap on open unlinked file" + +# test to verify file handle related system calls +# (name_to_handle_at/open_by_handle_at) +# The new system calls are supported in glibc >= 2.14. + +test_237() { + echo "Test file_handle syscalls" > $DIR/$tfile + check_fhandle_syscalls $DIR/$tfile || + error "check_fhandle_syscalls failed" +} +run_test 237 "Verify name_to_handle_at/open_by_handle_at syscalls" # # tests that do cleanup/setup should be run at the end