X-Git-Url: https://git.whamcloud.com/?p=fs%2Flustre-release.git;a=blobdiff_plain;f=lustre%2Ftests%2Fsanityn.sh;h=a0bfb327959983ea280faef386b6e41e785fc4d8;hp=e37c56ae197a846404817649b327d3c545a757fa;hb=7fa66448c2f51775fbbd4f50d114b4ca64e59576;hpb=0209add4a5099817111c8576afe930d1e2daef03 diff --git a/lustre/tests/sanityn.sh b/lustre/tests/sanityn.sh old mode 100644 new mode 100755 index e37c56a..a0bfb32 --- a/lustre/tests/sanityn.sh +++ b/lustre/tests/sanityn.sh @@ -3,23 +3,15 @@ set -e ONLY=${ONLY:-"$*"} -# bug number for skipped test: 3192 LU-1205 15528/3811 16929 9977 15528/11549 18080 -ALWAYS_EXCEPT=" 14b 18c 19 22 28 29 35 $SANITYN_EXCEPT" +# bug number for skipped test: LU-1205 9977/LU-7105 LU-9452 +ALWAYS_EXCEPT=" 18c 28 29 $SANITYN_EXCEPT" # UPDATE THE COMMENT ABOVE WITH BUG NUMBERS WHEN CHANGING ALWAYS_EXCEPT! -# bug number for skipped test: 12652 12652 -grep -q 'Enterprise Server 10' /etc/SuSE-release 2> /dev/null && - ALWAYS_EXCEPT="$ALWAYS_EXCEPT 11 14" || true - -# It will be ported soon. -EXCEPT="$EXCEPT 22" - -SRCDIR=`dirname $0` +SRCDIR=$(dirname $0) PATH=$PWD/$SRCDIR:$SRCDIR:$SRCDIR/../utils:$PATH SIZE=${SIZE:-40960} CHECKSTAT=${CHECKSTAT:-"checkstat -v"} -MCREATE=${MCREATE:-mcreate} OPENFILE=${OPENFILE:-openfile} OPENUNLINK=${OPENUNLINK:-openunlink} export MULTIOP=${MULTIOP:-multiop} @@ -41,12 +33,12 @@ init_test_env $@ init_logging if [ $(facet_fstype $SINGLEMDS) = "zfs" ]; then -# bug number for skipped test: LU-2840 LU-2189 LU-2776 - ALWAYS_EXCEPT="$ALWAYS_EXCEPT 21 36 51a" +# bug number for skipped test: LU-2189 LU-2776 + ALWAYS_EXCEPT="$ALWAYS_EXCEPT 36 51a" # LU-2829 / LU-2887 - make allowances for ZFS slowness TEST33_NFILES=${TEST33_NFILES:-1000} fi - +# 23 (min)" [ "$SLOW" = "no" ] && EXCEPT_SLOW="33a" FAIL_ON_ERROR=false @@ -56,9 +48,6 @@ TRACE=${TRACE:-""} check_and_setup_lustre -LOVNAME=$($LCTL get_param -n llite.*.lov.common_name | tail -n 1) -OSTCOUNT=$($LCTL get_param -n lov.$LOVNAME.numobd) - assert_DIR rm -rf $DIR1/[df][0-9]* $DIR1/lnk $DIR/[df].${TESTSUITE}* @@ -72,9 +61,6 @@ check_runas_id $RUNAS_ID $RUNAS_GID $RUNAS build_test_filter -mkdir -p $MOUNT2 -mount_client $MOUNT2 - test_1a() { touch $DIR1/f1 [ -f $DIR2/f1 ] || error @@ -172,6 +158,18 @@ test_2f() { } run_test 2f "check attr/owner updates on DNE with 2 mtpt's" +test_2g() { + dd if=/dev/zero of=$DIR1/$tfile oflag=sync bs=1M count=2 + + local block1=$(stat $DIR1/$tfile | awk '/Blocks/ {print $4} ') + cancel_lru_locks osc + local block2=$(stat $DIR2/$tfile | awk '/Blocks/ {print $4} ') + echo "$DIR1/$tfile has $block1 blocks" + echo "$DIR2/$tfile has $block2 blocks" + [ $block1 -eq $block2 ] || error +} +run_test 2g "check blocks update on sync write" + test_3() { local target="this/is/good" ln -s $target $DIR1/$tfile || error "ln -s $target $DIR1/$tfile failed" @@ -293,7 +291,7 @@ test_13() { # bug 2451 - directory coherency } run_test 13 "test directory page revocation ====================" -test_14() { +test_14aa() { test_mkdir -p $DIR1/$tdir cp -p /bin/ls $DIR1/$tdir/$tfile multiop_bg_pause $DIR1/$tdir/$tfile Ow_c || return 1 @@ -303,9 +301,9 @@ test_14() { kill -USR1 $MULTIPID wait $MULTIPID || return 2 } -run_test 14 "execution of file open for write returns -ETXTBSY =" +run_test 14aa "execution of file open for write returns -ETXTBSY" -test_14a() { +test_14ab() { test_mkdir -p $DIR1/d14 cp -p `which multiop` $DIR1/d14/multiop || error "cp failed" MULTIOP_PROG=$DIR1/d14/multiop multiop_bg_pause $TMP/test14.junk O_c || return 1 @@ -315,7 +313,7 @@ test_14a() { wait $MULTIOP_PID || return 3 rm $TMP/test14.junk $DIR1/d14/multiop || error "removing multiop" } -run_test 14a "open(RDWR) of executing file returns -ETXTBSY ====" +run_test 14ab "open(RDWR) of executing file returns -ETXTBSY" test_14b() { # bug 3192, 7040 test_mkdir -p $DIR1/d14 @@ -493,7 +491,7 @@ run_test 21 " Try to remove mountpoint on another dir ====" test_23() { # Bug 5972 local at_diff=$(do_facet $SINGLEMDS \ - $LCTL get_param -n mdd.*MDT0000*.atime_diff | head -1) + $LCTL get_param -n mdd.*MDT0000*.atime_diff | head -n1) echo "atime should be updated while another read" > $DIR1/$tfile # clear the lock(mode: LCK_PW) gotten from creating operation @@ -627,11 +625,11 @@ test_27() { DD2_PID=$! usleep 50 log "dd 1 started" - + dd if=/dev/zero of=$DIR1/$tfile bs=$((16384-1024))k conv=notrunc count=1 seek=4 & DD1_PID=$! log "dd 2 started" - + sleep 1 dd if=/dev/zero of=$DIR1/$tfile bs=8k conv=notrunc count=1 seek=0 log "dd 3 finished" @@ -643,7 +641,7 @@ run_test 27 "align non-overlapping extent locks from request ===" test_28() { # bug 9977 ECHO_UUID="ECHO_osc1_UUID" - tOST=`$LCTL dl | | awk '/-osc-|OSC.*MNT/ { print $4 }' | head -1` + tOST=$($LCTL dl | awk '/-osc-|OSC.*MNT/ { print $4 }' | head -n1) $LFS setstripe $DIR1/$tfile -S 1048576 -i 0 -c 2 tOBJID=`$LFS getstripe $DIR1/$tfile | awk '$1 == 1 {print $2}'` @@ -747,42 +745,42 @@ enable_lockless_truncate() { test_32a() { # bug 11270 local p="$TMP/$TESTSUITE-$TESTNAME.parameters" save_lustre_params client "osc.*.lockless_truncate" > $p - cancel_lru_locks osc - enable_lockless_truncate 1 - rm -f $DIR1/$tfile - lfs setstripe -c -1 $DIR1/$tfile - dd if=/dev/zero of=$DIR1/$tfile count=$OSTCOUNT bs=$STRIPE_BYTES > \ - /dev/null 2>&1 - clear_osc_stats - - log "checking cached lockless truncate" - $TRUNCATE $DIR1/$tfile 8000000 - $CHECKSTAT -s 8000000 $DIR2/$tfile || error "wrong file size" - [ $(calc_osc_stats lockless_truncate) -eq 0 ] || - error "lockless truncate doesn't use cached locks" - - log "checking not cached lockless truncate" - $TRUNCATE $DIR2/$tfile 5000000 - $CHECKSTAT -s 5000000 $DIR1/$tfile || error "wrong file size" - [ $(calc_osc_stats lockless_truncate) -ne 0 ] || - error "not cached trancate isn't lockless" - - log "disabled lockless truncate" - enable_lockless_truncate 0 - clear_osc_stats - $TRUNCATE $DIR2/$tfile 3000000 - $CHECKSTAT -s 3000000 $DIR1/$tfile || error "wrong file size" - [ $(calc_osc_stats lockless_truncate) -eq 0 ] || - error "lockless truncate disabling failed" - rm $DIR1/$tfile - # restore lockless_truncate default values - restore_lustre_params < $p - rm -f $p + cancel_lru_locks osc + enable_lockless_truncate 1 + rm -f $DIR1/$tfile + lfs setstripe -c -1 $DIR1/$tfile + dd if=/dev/zero of=$DIR1/$tfile count=$OSTCOUNT bs=$STRIPE_BYTES > \ + /dev/null 2>&1 + clear_stats osc.*.osc_stats + + log "checking cached lockless truncate" + $TRUNCATE $DIR1/$tfile 8000000 + $CHECKSTAT -s 8000000 $DIR2/$tfile || error "wrong file size" + [ $(calc_stats osc.*.osc_stats lockless_truncate) -ne 0 ] || + error "cached truncate isn't lockless" + + log "checking not cached lockless truncate" + $TRUNCATE $DIR2/$tfile 5000000 + $CHECKSTAT -s 5000000 $DIR1/$tfile || error "wrong file size" + [ $(calc_stats osc.*.osc_stats lockless_truncate) -ne 0 ] || + error "not cached truncate isn't lockless" + + log "disabled lockless truncate" + enable_lockless_truncate 0 + clear_stats osc.*.osc_stats + $TRUNCATE $DIR2/$tfile 3000000 + $CHECKSTAT -s 3000000 $DIR1/$tfile || error "wrong file size" + [ $(calc_stats osc.*.osc_stats lockless_truncate) -eq 0 ] || + error "lockless truncate disabling failed" + rm $DIR1/$tfile + # restore lockless_truncate default values + restore_lustre_params < $p + rm -f $p } run_test 32a "lockless truncate" test_32b() { # bug 11270 - remote_ost_nodsh && skip "remote OST with nodsh" && return + remote_ost_nodsh && skip "remote OST with nodsh" && return local node local facets=$(get_facets OST) @@ -795,35 +793,42 @@ test_32b() { # bug 11270 "ldlm.namespaces.filter-*.contended_locks" >> $p save_lustre_params $facets \ "ldlm.namespaces.filter-*.contention_seconds" >> $p - clear_osc_stats + clear_stats osc.*.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 - for i in $(seq 5); do - dd if=/dev/zero of=$DIR1/$tfile bs=4k count=1 conv=notrunc > /dev/null 2>&1 - dd if=/dev/zero of=$DIR2/$tfile bs=4k count=1 conv=notrunc > /dev/null 2>&1 - done - [ $(calc_osc_stats lockless_write_bytes) -ne 0 ] || error "lockless i/o was not triggered" - # disable lockless i/o (it is disabled by default) - 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 - # set contention_seconds to 0 at client too, otherwise Lustre still - # remembers lock contention - lctl set_param -n osc.*.contention_seconds 0 - clear_osc_stats - for i in $(seq 1); do - dd if=/dev/zero of=$DIR1/$tfile bs=4k count=1 conv=notrunc > /dev/null 2>&1 - dd if=/dev/zero of=$DIR2/$tfile bs=4k count=1 conv=notrunc > /dev/null 2>&1 - done - [ $(calc_osc_stats lockless_write_bytes) -eq 0 ] || - error "lockless i/o works when disabled" - rm -f $DIR1/$tfile - restore_lustre_params <$p - rm -f $p + # agressive lockless i/o settings + do_nodes $(comma_list $(osts_nodes)) \ + "lctl set_param -n ldlm.namespaces.*.max_nolock_bytes=2000000 \ + ldlm.namespaces.filter-*.contended_locks=0 \ + ldlm.namespaces.filter-*.contention_seconds=60" + lctl set_param -n osc.*.contention_seconds=60 + for i in {1..5}; do + dd if=/dev/zero of=$DIR1/$tfile bs=4k count=1 conv=notrunc > \ + /dev/null 2>&1 + dd if=/dev/zero of=$DIR2/$tfile bs=4k count=1 conv=notrunc > \ + /dev/null 2>&1 + done + [ $(calc_stats osc.*.osc_stats lockless_write_bytes) -ne 0 ] || + error "lockless i/o was not triggered" + # disable lockless i/o (it is disabled by default) + do_nodes $(comma_list $(osts_nodes)) \ + "lctl set_param -n ldlm.namespaces.filter-*.max_nolock_bytes=0 \ + ldlm.namespaces.filter-*.contended_locks=32 \ + ldlm.namespaces.filter-*.contention_seconds=0" + # set contention_seconds to 0 at client too, otherwise Lustre still + # remembers lock contention + lctl set_param -n osc.*.contention_seconds=0 + clear_stats osc.*.osc_stats + for i in {1..1}; do + dd if=/dev/zero of=$DIR1/$tfile bs=4k count=1 conv=notrunc > \ + /dev/null 2>&1 + dd if=/dev/zero of=$DIR2/$tfile bs=4k count=1 conv=notrunc > \ + /dev/null 2>&1 + done + [ $(calc_stats osc.*.osc_stats lockless_write_bytes) -eq 0 ] || + error "lockless i/o works when disabled" + rm -f $DIR1/$tfile + restore_lustre_params <$p + rm -f $p } run_test 32b "lockless i/o" @@ -838,8 +843,8 @@ print_jbd_stat () { varsvc=${mds}_svc dev=$(basename $(do_facet $mds "lctl get_param -n osd*.${!varsvc}.mntdev|\ xargs readlink -f" )) - val=$(do_facet $mds "cat /proc/fs/jbd*/${dev}{,:*,-*}/info 2>/dev/null|\ - head -1") + val=$(do_facet $mds "cat /proc/fs/jbd*/${dev}{,:*,-*}/info 2>/dev/null | + head -n1") val=${val%% *}; stat=$(( stat + val)) done @@ -875,7 +880,7 @@ test_33a() { [ $fstype = ldiskfs ] && jbdold=$(print_jbd_stat) echo "=== START createmany old: $jbdold transaction" - local elapsed=$(do_and_time "do_nodes $CLIENT1,$CLIENT2 createmany -o $DIR1/$tdir-\\\$(hostname)-$i/f- -r $DIR2/$tdir-\\\$(hostname)-$i/f- $nfiles > /dev/null 2>&1") + local elapsed=$(do_and_time "do_nodes $CLIENT1,$CLIENT2 createmany -o $DIR1/$tdir-\\\$(hostname)-$i/f- -r$DIR2/$tdir-\\\$(hostname)-$i/f- $nfiles > /dev/null 2>&1") [ $fstype = ldiskfs ] && jbdnew=$(print_jbd_stat) [ $fstype = ldiskfs ] && jbd=$(( jbdnew - jbdold )) echo "=== END createmany new: $jbdnew transaction : $jbd transactions nfiles $nfiles time $elapsed COS=$COS" @@ -924,14 +929,14 @@ test_33b() { avgjbd=0 avgtime=0 for i in 1 2 3; do - do_node $CLIENT1 "$LFS mkdir -i $MDTIDX -p \ + do_node $CLIENT1 "$LFS mkdir -i $MDTIDX \ $DIR1/$tdir-\\\$(hostname)-$i" jbdold=$(print_jbd_stat) echo "=== START createmany old: $jbdold transaction" local elapsed=$(do_and_time "do_nodes $CLIENT1,$CLIENT2\ createmany -o $DIR1/$tdir-\\\$(hostname)-$i/f- \ - -r $DIR2/$tdir-\\\$(hostname)-$i/f- $nfiles > \ + -r$DIR2/$tdir-\\\$(hostname)-$i/f- $nfiles > \ /dev/null 2>&1") jbdnew=$(print_jbd_stat) jbd=$(( jbdnew - jbdold )) @@ -957,6 +962,118 @@ test_33b() { } run_test 33b "COS: cross create/delete, 2 clients, benchmark under remote dir" +test_33c() { + [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return + [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.7.63) ] && + skip "DNE CoS not supported" && return + + sync + + mkdir $DIR/$tdir + # remote mkdir is done on MDT2, which enqueued lock of $tdir on MDT1 + $LFS mkdir -i 1 $DIR/$tdir/d1 + do_facet mds1 "lctl set_param -n mdt.*.sync_count=0" + mkdir $DIR/$tdir/d2 + local sync_count=$(do_facet mds1 \ + "lctl get_param -n mdt.*MDT0000.sync_count") + [ $sync_count -eq 1 ] || error "Sync-Lock-Cancel not triggered" + + $LFS mkdir -i 1 $DIR/$tdir/d3 + do_facet mds1 "lctl set_param -n mdt.*.sync_count=0" + # during sleep remote mkdir should have been committed and canceled + # remote lock spontaneously, which shouldn't trigger sync + sleep 6 + mkdir $DIR/$tdir/d4 + local sync_count=$(do_facet mds1 \ + "lctl get_param -n mdt.*MDT0000.sync_count") + [ $sync_count -eq 0 ] || error "Sync-Lock-Cancel triggered" +} +run_test 33c "Cancel cross-MDT lock should trigger Sync-Lock-Cancel" + +ops_do_cos() { + local nodes=$(comma_list $(mdts_nodes)) + do_nodes $nodes "lctl set_param -n mdt.*.async_commit_count=0" + sh -c "$@" + local async_commit_count=$(do_nodes $nodes \ + "lctl get_param -n mdt.*.async_commit_count" | calc_sum) + [ $async_commit_count -gt 0 ] || error "CoS not triggerred" + + rm -rf $DIR/$tdir + sync +} + +test_33d() { + [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return + [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.7.63) ] && + skip "DNE CoS not supported" && return + + sync + # remote directory create + mkdir $DIR/$tdir + ops_do_cos "$LFS mkdir -i 1 $DIR/$tdir/subdir" + # remote directory unlink + $LFS mkdir -i 1 $DIR/$tdir + ops_do_cos "rmdir $DIR/$tdir" + # striped directory create + mkdir $DIR/$tdir + ops_do_cos "$LFS mkdir -c 2 $DIR/$tdir/subdir" + # striped directory setattr + $LFS mkdir -c 2 $DIR/$tdir + touch $DIR/$tdir + ops_do_cos "chmod 713 $DIR/$tdir" + # striped directory unlink + $LFS mkdir -c 2 $DIR/$tdir + touch $DIR/$tdir + ops_do_cos "rmdir $DIR/$tdir" + # cross-MDT link + $LFS mkdir -c 2 $DIR/$tdir + $LFS mkdir -i 0 $DIR/$tdir/d1 + $LFS mkdir -i 1 $DIR/$tdir/d2 + touch $DIR/$tdir/d1/tgt + ops_do_cos "ln $DIR/$tdir/d1/tgt $DIR/$tdir/d2/src" + # cross-MDT rename + $LFS mkdir -c 2 $DIR/$tdir + $LFS mkdir -i 0 $DIR/$tdir/d1 + $LFS mkdir -i 1 $DIR/$tdir/d2 + touch $DIR/$tdir/d1/src + ops_do_cos "mv $DIR/$tdir/d1/src $DIR/$tdir/d2/tgt" + # migrate + $LFS mkdir -i 0 $DIR/$tdir + ops_do_cos "$LFS migrate -m 1 $DIR/$tdir" + return 0 +} +run_test 33d "DNE distributed operation should trigger COS" + +test_33e() { + [ -n "$CLIENTS" ] || { skip "Need two or more clients" && return 0; } + [ $CLIENTCOUNT -ge 2 ] || + { skip "Need two or more clients, have $CLIENTCOUNT" && + return 0; } + [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return + [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.7.63) ] && + skip "DNE CoS not supported" && return + + local client2=${CLIENT2:-$(hostname)} + + sync + + local nodes=$(comma_list $(mdts_nodes)) + do_nodes $nodes "lctl set_param -n mdt.*.async_commit_count=0" + + $LFS mkdir -c 2 $DIR/$tdir + mkdir $DIR/$tdir/subdir + echo abc > $DIR/$tdir/$tfile + do_node $client2 echo dfg >> $DIR/$tdir/$tfile + do_node $client2 touch $DIR/$tdir/subdir + + local async_commit_count=$(do_nodes $nodes \ + "lctl get_param -n mdt.*.async_commit_count" | calc_sum) + [ $async_commit_count -gt 0 ] && error "CoS triggerred" + + return 0 +} +run_test 33e "DNE local operation shouldn't trigger COS" + # End commit on sharing tests get_ost_lock_timeouts() { @@ -979,6 +1096,7 @@ cleanup_34() { } test_34() { #16129 + remote_ost_nodsh && skip "remote OST with nodsh" && return local OPER local lock_in local lock_out @@ -1028,8 +1146,9 @@ run_test 34 "no lock timeout under IO" test_35() { # bug 17645 local generation=[] local count=0 - for imp in /proc/fs/lustre/mdc/$FSNAME-MDT*-mdc-*; do - g=$(awk '/generation/{print $2}' $imp/import) + gen=$(lctl get_param mdc.$FSNAME-MDT*-mdc-*.import | grep generation | + awk '/generation/{print $2}') + for g in $gen; do generation[count]=$g let count=count+1 done @@ -1071,10 +1190,20 @@ test_35() { # bug 17645 do_facet client "lctl set_param fail_loc=0x0" df -h $MOUNT1 $MOUNT2 count=0 - for imp in /proc/fs/lustre/mdc/$FSNAME-MDT*-mdc-*; do - g=$(awk '/generation/{print $2}' $imp/import) - if ! test "$g" -eq "${generation[count]}"; then - error "Eviction happened on import $(basename $imp)" + gen=$(lctl get_param mdc.$FSNAME-MDT*-mdc-*.import | grep generation | + awk '/generation/{print $2}') + for g in $gen; do + if ! test "$g" -eq "${generation[count]}"; then + list=$(lctl list_param mdc.$FSNAME-MDT*-mdc-*.import) + local c = 0 + for imp in $list; do + if [ $c = $count ]; then + break + fi + c=c+1 + done + imp=$(echo "$imp" | awk -F"." '{print $2}') + error "Eviction happened on import $imp" fi let count=count+1 done @@ -1226,6 +1355,32 @@ test_39c() { } run_test 39c "check truncate mtime update ======================" +test_39d() { # LU-7310 + touch $DIR1/$tfile + touch -m -d @$TEST_39_MTIME $DIR1/$tfile + + local mtime1=$(stat -c %Y $DIR2/$tfile) + [ "$mtime1" = $TEST_39_MTIME ] || + error "mtime: $mtime1, should be $TEST_39_MTIME" + + # force sync write + # define OBD_FAIL_OSC_NO_GRANT 0x411 + $LCTL set_param fail_loc=0x411 + + local d1=$(date +%s) + echo hello >> $DIR1/$tfile + local d2=$(date +%s) + + $LCTL set_param fail_loc=0 + + cancel_lru_locks osc + + local mtime2=$(stat -c %Y $DIR2/$tfile) + [ "$mtime2" -ge "$d1" ] && [ "$mtime2" -le "$d2" ] || + error "mtime is not updated on write: $d1 <= $mtime2 <= $d2" +} +run_test 39d "sync write should update mtime" + # check that pid exists hence second operation wasn't blocked by first one # if it is so then there is no conflict, return 0 # else second operation is conflicting with first one, return 1 @@ -1245,8 +1400,10 @@ check_pdo_conflict() { # pdirop tests # test 40: check non-blocking operations test_40a() { + remote_mds_nodsh && skip "remote MDS with nodsh" && return #define OBD_FAIL_ONCE|OBD_FAIL_MDS_PDO_LOCK 0x145 do_facet $SINGLEMDS lctl set_param fail_loc=0x80000145 + touch $DIR2 mkdir $DIR1/$tfile & PID1=$! sleep 1 @@ -1264,15 +1421,16 @@ test_40a() { rmdir $DIR2/$tfile-3 check_pdo_conflict $PID1 || error "unlink is blocked" - # all operations above shouldn't wait the first one + # all operations above shouldn't wait the first one check_pdo_conflict $PID1 || error "parallel operation is blocked" wait $PID1 - rm -r $DIR1/* + rm -rf $DIR/$tfile* return 0 } run_test 40a "pdirops: create vs others ==============" test_40b() { + remote_mds_nodsh && skip "remote MDS with nodsh" && return #define OBD_FAIL_ONCE|OBD_FAIL_MDS_PDO_LOCK 0x145 do_facet $SINGLEMDS lctl set_param fail_loc=0x80000145 touch $DIR1/$tfile & @@ -1296,12 +1454,13 @@ test_40b() { check_pdo_conflict $PID1 || error "parallel operation is blocked" wait $PID1 - rm -r $DIR1/* + rm -rf $DIR/$tfile* return 0 } run_test 40b "pdirops: open|create and others ==============" test_40c() { + remote_mds_nodsh && skip "remote MDS with nodsh" && return touch $DIR1/$tfile #define OBD_FAIL_ONCE|OBD_FAIL_MDS_PDO_LOCK 0x145 do_facet $SINGLEMDS lctl set_param fail_loc=0x80000145 @@ -1326,12 +1485,13 @@ test_40c() { # all operations above shouldn't wait the first one check_pdo_conflict $PID1 || error "parallel operation is blocked" wait $PID1 - rm -r $DIR1/* + rm -rf $DIR/$tfile* return 0 } run_test 40c "pdirops: link and others ==============" test_40d() { + remote_mds_nodsh && skip "remote MDS with nodsh" && return touch $DIR1/$tfile #define OBD_FAIL_ONCE|OBD_FAIL_MDS_PDO_LOCK 0x145 do_facet $SINGLEMDS lctl set_param fail_loc=0x80000145 @@ -1361,6 +1521,7 @@ test_40d() { run_test 40d "pdirops: unlink and others ==============" test_40e() { + remote_mds_nodsh && skip "remote MDS with nodsh" && return touch $DIR1/$tfile #define OBD_FAIL_ONCE|OBD_FAIL_MDS_PDO_LOCK 0x145 do_facet $SINGLEMDS lctl set_param fail_loc=0x80000145 @@ -1383,7 +1544,7 @@ test_40e() { # all operations above shouldn't wait the first one check_pdo_conflict $PID1 || error "parallel operation is blocked" wait $PID1 - rm -r $DIR1/* + rm -rf $DIR/$tfile* return 0 } run_test 40e "pdirops: rename and others ==============" @@ -1397,7 +1558,7 @@ test_41a() { sleep 1 mkdir $DIR2/$tfile && error "mkdir must fail" check_pdo_conflict $PID1 && { wait $PID1; echo "mkdir isn't blocked"; } - rm -r $DIR1/* + rm -rf $DIR/$tfile* return 0 } run_test 41a "pdirops: create vs mkdir ==============" @@ -1410,7 +1571,7 @@ test_41b() { sleep 1 $MULTIOP $DIR2/$tfile oO_CREAT:O_EXCL:c && error "create must fail" check_pdo_conflict $PID1 && { wait $PID1; error "create isn't blocked"; } - rm -r $DIR1/* + rm -rf $DIR/$tfile* return 0 } run_test 41b "pdirops: create vs create ==============" @@ -1424,7 +1585,7 @@ test_41c() { sleep 1 link $DIR2/$tfile-2 $DIR2/$tfile && error "link must fail" check_pdo_conflict $PID1 && { wait $PID1; error "link isn't blocked"; } - rm -r $DIR1/* + rm -rf $DIR/$tfile* return 0 } run_test 41c "pdirops: create vs link ==============" @@ -1437,7 +1598,7 @@ test_41d() { sleep 1 rm $DIR2/$tfile || error "unlink must succeed" check_pdo_conflict $PID1 && { wait $PID1; error "unlink isn't blocked"; } - rm -r $DIR1/* + rm -rf $DIR/$tfile* return 0 } run_test 41d "pdirops: create vs unlink ==============" @@ -1451,7 +1612,7 @@ test_41e() { sleep 1 mv $DIR2/$tfile-2 $DIR2/$tfile || error "rename must succeed" check_pdo_conflict $PID1 && { wait $PID1; error "rename isn't blocked"; } - rm -r $DIR1/* + rm -rf $DIR/$tfile* return 0 } run_test 41e "pdirops: create and rename (tgt) ==============" @@ -1464,7 +1625,7 @@ test_41f() { sleep 1 mv $DIR2/$tfile $DIR2/$tfile-2 || error "rename must succeed" check_pdo_conflict $PID1 && { wait $PID1; error "rename isn't blocked"; } - rm -r $DIR1/* + rm -rf $DIR/$tfile* return 0 } run_test 41f "pdirops: create and rename (src) ==============" @@ -1477,7 +1638,7 @@ test_41g() { sleep 1 stat $DIR2/$tfile > /dev/null || error "stat must succeed" check_pdo_conflict $PID1 && { wait $PID1; error "getattr isn't blocked"; } - rm -r $DIR1/* + rm -rf $DIR/$tfile* return 0 } run_test 41g "pdirops: create vs getattr ==============" @@ -1490,7 +1651,7 @@ test_41h() { sleep 1 ls -lia $DIR2/ > /dev/null check_pdo_conflict $PID1 && { wait $PID1; error "readdir isn't blocked"; } - rm -r $DIR1/* + rm -rf $DIR/$tfile* return 0 } run_test 41h "pdirops: create vs readdir ==============" @@ -1504,7 +1665,7 @@ test_42a() { sleep 1 mkdir $DIR2/$tfile && error "mkdir must fail" check_pdo_conflict $PID1 && { wait $PID1; error "mkdir isn't blocked"; } - rm -r $DIR1/* + rm -rf $DIR/$tfile* return 0 } run_test 42a "pdirops: mkdir vs mkdir ==============" @@ -1517,7 +1678,7 @@ test_42b() { sleep 1 $MULTIOP $DIR2/$tfile oO_CREAT:O_EXCL:c && error "create must fail" check_pdo_conflict $PID1 && { wait $PID1; error "create isn't blocked"; } - rm -r $DIR1/* + rm -rf $DIR/$tfile* return 0 } run_test 42b "pdirops: mkdir vs create ==============" @@ -1531,7 +1692,7 @@ test_42c() { sleep 1 link $DIR2/$tfile-2 $DIR2/$tfile && error "link must fail" check_pdo_conflict $PID1 && { wait $PID1; error "link isn't blocked"; } - rm -r $DIR1/* + rm -rf $DIR/$tfile* return 0 } run_test 42c "pdirops: mkdir vs link ==============" @@ -1544,7 +1705,7 @@ test_42d() { sleep 1 rmdir $DIR2/$tfile || error "unlink must succeed" check_pdo_conflict $PID1 && { wait $PID1; error "unlink isn't blocked"; } - rm -r $DIR1/* + rm -rf $DIR/$tfile* return 0 } run_test 42d "pdirops: mkdir vs unlink ==============" @@ -1558,7 +1719,7 @@ test_42e() { sleep 1 mv -T $DIR2/$tfile-2 $DIR2/$tfile && error "rename must fail" check_pdo_conflict $PID1 && { wait $PID1; error "rename isn't blocked"; } - rm -r $DIR1/* + rm -rf $DIR/$tfile* return 0 } run_test 42e "pdirops: mkdir and rename (tgt) ==============" @@ -1571,7 +1732,7 @@ test_42f() { sleep 1 mv $DIR2/$tfile $DIR2/$tfile-2 || error "rename must succeed" check_pdo_conflict $PID1 && { wait $PID1; error "rename isn't blocked"; } - rm -r $DIR1/* + rm -rf $DIR/$tfile* return 0 } run_test 42f "pdirops: mkdir and rename (src) ==============" @@ -1584,7 +1745,7 @@ test_42g() { sleep 1 stat $DIR2/$tfile > /dev/null || error "stat must succeed" check_pdo_conflict $PID1 && { wait $PID1; error "getattr isn't blocked"; } - rm -r $DIR1/* + rm -rf $DIR/$tfile* return 0 } run_test 42g "pdirops: mkdir vs getattr ==============" @@ -1597,7 +1758,7 @@ test_42h() { sleep 1 ls -lia $DIR2/ > /dev/null check_pdo_conflict $PID1 && { wait $PID1; error "readdir isn't blocked"; } - rm -r $DIR1/* + rm -rf $DIR/$tfile* return 0 } run_test 42h "pdirops: mkdir vs readdir ==============" @@ -1612,7 +1773,7 @@ test_43a() { sleep 1 mkdir $DIR2/$tfile || error "mkdir must succeed" check_pdo_conflict $PID1 && { wait $PID1; error "mkdir isn't blocked"; } - rm -r $DIR1/* + rm -rf $DIR/$tfile* return 0 } run_test 43a "pdirops: unlink vs mkdir ==============" @@ -1626,7 +1787,7 @@ test_43b() { sleep 1 $MULTIOP $DIR2/$tfile oO_CREAT:O_EXCL:c || error "create must succeed" check_pdo_conflict $PID1 && { wait $PID1; error "create isn't blocked"; } - rm -r $DIR1/* + rm -rf $DIR/$tfile* return 0 } run_test 43b "pdirops: unlink vs create ==============" @@ -1641,7 +1802,7 @@ test_43c() { sleep 1 link $DIR2/$tfile-2 $DIR2/$tfile || error "link must succeed" check_pdo_conflict $PID1 && { wait $PID1; error "link isn't blocked"; } - rm -r $DIR1/* + rm -rf $DIR/$tfile* return 0 } run_test 43c "pdirops: unlink vs link ==============" @@ -1655,7 +1816,7 @@ test_43d() { sleep 1 rm $DIR2/$tfile && error "unlink must fail" check_pdo_conflict $PID1 && { wait $PID1; error "unlink isn't blocked"; } - rm -r $DIR1/* + rm -rf $DIR/$tfile* return 0 } run_test 43d "pdirops: unlink vs unlink ==============" @@ -1670,7 +1831,7 @@ test_43e() { sleep 1 mv -u $DIR2/$tfile-2 $DIR2/$tfile || error "rename must succeed" check_pdo_conflict $PID1 && { wait $PID1; error "rename isn't blocked"; } - rm -r $DIR1/* + rm -rf $DIR/$tfile* return 0 } run_test 43e "pdirops: unlink and rename (tgt) ==============" @@ -1684,7 +1845,7 @@ test_43f() { sleep 1 mv $DIR2/$tfile $DIR2/$tfile-2 && error "rename must fail" check_pdo_conflict $PID1 && { wait $PID1; error "rename isn't blocked"; } - rm -r $DIR1/* + rm -rf $DIR/$tfile* return 0 } run_test 43f "pdirops: unlink and rename (src) ==============" @@ -1698,7 +1859,7 @@ test_43g() { sleep 1 stat $DIR2/$tfile > /dev/null && error "stat must fail" check_pdo_conflict $PID1 && { wait $PID1; error "getattr isn't blocked"; } - rm -r $DIR1/* + rm -rf $DIR/$tfile* return 0 } run_test 43g "pdirops: unlink vs getattr ==============" @@ -1712,7 +1873,7 @@ test_43h() { sleep 1 ls -lia $DIR2/ > /dev/null check_pdo_conflict $PID1 && { wait $PID1; error "readdir isn't blocked"; } - rm -r $DIR1/* + rm -rf $DIR/$tfile* return 0 } run_test 43h "pdirops: unlink vs readdir ==============" @@ -1728,7 +1889,7 @@ test_43i() { $LFS mkdir -i 1 $DIR2/$tfile || error "remote mkdir must succeed" check_pdo_conflict $PID1 && { wait $PID1; error "remote mkdir isn't blocked"; } - rm -r $DIR1/* + rm -rf $DIR/$tfile* return 0 } run_test 43i "pdirops: unlink vs remote mkdir" @@ -1743,7 +1904,7 @@ test_44a() { sleep 1 mkdir $DIR2/$tfile && error "mkdir must fail" check_pdo_conflict $PID1 && { wait $PID1; error "mkdir isn't blocked"; } - rm -r $DIR1/* + rm -rf $DIR/$tfile* return 0 } run_test 44a "pdirops: rename tgt vs mkdir ==============" @@ -1757,7 +1918,7 @@ test_44b() { sleep 1 $MULTIOP $DIR2/$tfile oO_CREAT:O_EXCL:c && error "create must fail" check_pdo_conflict $PID1 && { wait $PID1; error "create isn't blocked"; } - rm -r $DIR1/* + rm -rf $DIR/$tfile* return 0 } run_test 44b "pdirops: rename tgt vs create ==============" @@ -1772,7 +1933,7 @@ test_44c() { sleep 1 link $DIR2/$tfile-3 $DIR2/$tfile && error "link must fail" check_pdo_conflict $PID1 && { wait $PID1; error "link isn't blocked"; } - rm -r $DIR1/* + rm -rf $DIR/$tfile* return 0 } run_test 44c "pdirops: rename tgt vs link ==============" @@ -1786,7 +1947,7 @@ test_44d() { sleep 1 rm $DIR2/$tfile || error "unlink must succeed" check_pdo_conflict $PID1 && { wait $PID1; error "unlink isn't blocked"; } - rm -r $DIR1/* + rm -rf $DIR/$tfile* return 0 } run_test 44d "pdirops: rename tgt vs unlink ==============" @@ -1802,7 +1963,7 @@ test_44e() { sleep 1 mv $DIR2/$tfile-3 $DIR2/$tfile || error "rename must succeed" check_pdo_conflict $PID1 && { wait $PID1; error "rename isn't blocked"; } - rm -r $DIR1/* + rm -rf $DIR/$tfile* return 0 } run_test 44e "pdirops: rename tgt and rename (tgt) ==============" @@ -1817,7 +1978,7 @@ test_44f() { sleep 1 mv $DIR2/$tfile $DIR2/$tfile-3 || error "rename must succeed" check_pdo_conflict $PID1 && { wait $PID1; error "rename isn't blocked"; } - rm -r $DIR1/* + rm -rf $DIR/$tfile* return 0 } run_test 44f "pdirops: rename tgt and rename (src) ==============" @@ -1831,7 +1992,7 @@ test_44g() { sleep 1 stat $DIR2/$tfile > /dev/null || error "stat must succeed" check_pdo_conflict $PID1 && { wait $PID1; error "getattr isn't blocked"; } - rm -r $DIR1/* + rm -rf $DIR/$tfile* return 0 } run_test 44g "pdirops: rename tgt vs getattr ==============" @@ -1845,7 +2006,7 @@ test_44h() { sleep 1 ls -lia $DIR2/ > /dev/null check_pdo_conflict $PID1 && { wait $PID1; error "readdir isn't blocked"; } - rm -r $DIR1/* + rm -rf $DIR/$tfile* return 0 } run_test 44h "pdirops: rename tgt vs readdir ==============" @@ -1862,7 +2023,7 @@ test_44i() { $LFS mkdir -i 1 $DIR2/$tfile && error "remote mkdir must fail" check_pdo_conflict $PID1 && { wait $PID1; error "remote mkdir isn't blocked"; } - rm -r $DIR1/* + rm -rf $DIR/$tfile* return 0 } run_test 44i "pdirops: rename tgt vs remote mkdir" @@ -1877,7 +2038,7 @@ test_45a() { sleep 1 mkdir $DIR2/$tfile || error "mkdir must succeed" check_pdo_conflict $PID1 && { wait $PID1; error "mkdir isn't blocked"; } - rm -r $DIR1/* + rm -rf $DIR/$tfile* return 0 } run_test 45a "pdirops: rename src vs mkdir ==============" @@ -1891,7 +2052,7 @@ test_45b() { sleep 1 $MULTIOP $DIR2/$tfile oO_CREAT:O_EXCL:c || error "create must succeed" check_pdo_conflict $PID1 && { wait $PID1; error "create isn't blocked"; } - rm -r $DIR1/* + rm -rf $DIR/$tfile* return 0 } run_test 45b "pdirops: rename src vs create ==============" @@ -1906,7 +2067,7 @@ test_45c() { sleep 1 link $DIR2/$tfile-3 $DIR2/$tfile || error "link must succeed" check_pdo_conflict $PID1 && { wait $PID1; error "link isn't blocked"; } - rm -r $DIR1/* + rm -rf $DIR/$tfile* return 0 } run_test 45c "pdirops: rename src vs link ==============" @@ -1920,7 +2081,7 @@ test_45d() { sleep 1 rm $DIR2/$tfile && error "unlink must fail" check_pdo_conflict $PID1 && { wait $PID1; error "unlink isn't blocked"; } - rm -r $DIR1/* + rm -rf $DIR/$tfile* return 0 } run_test 45d "pdirops: rename src vs unlink ==============" @@ -1935,7 +2096,7 @@ test_45e() { sleep 1 mv $DIR2/$tfile-3 $DIR2/$tfile || error "rename must succeed" check_pdo_conflict $PID1 && { wait $PID1; error "rename isn't blocked"; } - rm -r $DIR1/* + rm -rf $DIR/$tfile* return 0 } run_test 45e "pdirops: rename src and rename (tgt) ==============" @@ -1949,7 +2110,7 @@ test_45f() { sleep 1 mv $DIR2/$tfile $DIR2/$tfile-3 && error "rename must fail" check_pdo_conflict $PID1 && { wait $PID1; error "rename isn't blocked"; } - rm -r $DIR1/* + rm -rf $DIR/$tfile* return 0 } run_test 45f "pdirops: rename src and rename (src) ==============" @@ -1963,7 +2124,7 @@ test_45g() { sleep 1 stat $DIR2/$tfile > /dev/null && error "stat must fail" check_pdo_conflict $PID1 && { wait $PID1; error "getattr isn't blocked"; } - rm -r $DIR1/* + rm -rf $DIR/$tfile* return 0 } run_test 45g "pdirops: rename src vs getattr ==============" @@ -1977,7 +2138,7 @@ test_45h() { sleep 1 ls -lia $DIR2/ > /dev/null check_pdo_conflict $PID1 && { wait $PID1; error "readdir isn't blocked"; } - rm -r $DIR1/* + rm -rf $DIR/$tfile* return 0 } run_test 45h "pdirops: unlink vs readdir ==============" @@ -1993,7 +2154,7 @@ test_45i() { $LFS mkdir -i 1 $DIR2/$tfile || error "create remote dir must succeed" check_pdo_conflict $PID1 && { wait $PID1; error "create remote dir isn't blocked"; } - rm -r $DIR1/* + rm -rf $DIR/$tfile* return 0 } run_test 45i "pdirops: rename src vs remote mkdir" @@ -2008,7 +2169,7 @@ test_46a() { sleep 1 mkdir $DIR2/$tfile && error "mkdir must fail" check_pdo_conflict $PID1 && { wait $PID1; error "mkdir isn't blocked"; } - rm -r $DIR1/* + rm -rf $DIR/$tfile* return 0 } run_test 46a "pdirops: link vs mkdir ==============" @@ -2022,7 +2183,7 @@ test_46b() { sleep 1 $MULTIOP $DIR2/$tfile oO_CREAT:O_EXCL:c && error "create must fail" check_pdo_conflict $PID1 && { wait $PID1; error "create isn't blocked"; } - rm -r $DIR1/* + rm -rf $DIR/$tfile* return 0 } run_test 46b "pdirops: link vs create ==============" @@ -2036,7 +2197,7 @@ test_46c() { sleep 1 link $DIR2/$tfile $DIR2/$tfile && error "link must fail" check_pdo_conflict $PID1 && { wait $PID1; error "link isn't blocked"; } - rm -r $DIR1/* + rm -rf $DIR/$tfile* return 0 } run_test 46c "pdirops: link vs link ==============" @@ -2050,7 +2211,7 @@ test_46d() { sleep 1 rm $DIR2/$tfile || error "unlink must succeed" check_pdo_conflict $PID1 && { wait $PID1; error "unlink isn't blocked"; } - rm -r $DIR1/* + rm -rf $DIR/$tfile* return 0 } run_test 46d "pdirops: link vs unlink ==============" @@ -2065,7 +2226,7 @@ test_46e() { sleep 1 mv $DIR2/$tfile-3 $DIR2/$tfile || error "rename must succeed" check_pdo_conflict $PID1 && { wait $PID1; error "rename isn't blocked"; } - rm -r $DIR1/* + rm -rf $DIR/$tfile* return 0 } run_test 46e "pdirops: link and rename (tgt) ==============" @@ -2080,7 +2241,7 @@ test_46f() { sleep 1 mv $DIR2/$tfile $DIR2/$tfile-3 || error "rename must succeed" check_pdo_conflict $PID1 && { wait $PID1; error "rename isn't blocked"; } - rm -r $DIR1/* + rm -rf $DIR/$tfile* return 0 } run_test 46f "pdirops: link and rename (src) ==============" @@ -2094,7 +2255,7 @@ test_46g() { sleep 1 stat $DIR2/$tfile > /dev/null || error "stat must succeed" check_pdo_conflict $PID1 && { wait $PID1; error "getattr isn't blocked"; } - rm -r $DIR1/* + rm -rf $DIR/$tfile* return 0 } run_test 46g "pdirops: link vs getattr ==============" @@ -2109,7 +2270,7 @@ test_46h() { ls -lia $DIR2/ > /dev/null check_pdo_conflict $PID1 && { wait $PID1; error "readdir isn't blocked"; } - rm -r $DIR1/* + rm -rf $DIR/$tfile* return 0 } run_test 46h "pdirops: link vs readdir ==============" @@ -2125,7 +2286,7 @@ test_46i() { $LFS mkdir -i 1 $DIR2/$tfile && error "remote mkdir must fail" check_pdo_conflict $PID1 && { wait $PID1; error "remote mkdir isn't blocked"; } - rm -r $DIR1/* + rm -rf $DIR/$tfile* return 0 } run_test 46i "pdirops: link vs remote mkdir" @@ -2140,7 +2301,7 @@ test_47a() { sleep 1 mkdir $DIR2/$tfile && error "mkdir must fail" check_pdo_conflict $PID1 && { wait $PID1; error "mkdir isn't blocked"; } - rm -r $DIR1/* + rm -rf $DIR/$tfile* return 0 } run_test 47a "pdirops: remote mkdir vs mkdir" @@ -2155,7 +2316,7 @@ test_47b() { multiop $DIR2/$tfile oO_CREAT:O_EXCL:c && error "create must fail" check_pdo_conflict $PID1 && { wait $PID1; error "create isn't blocked"; } - rm -r $DIR1/* + rm -rf $DIR/$tfile* return 0 } run_test 47b "pdirops: remote mkdir vs create" @@ -2170,7 +2331,7 @@ test_47c() { sleep 1 link $DIR2/$tfile-2 $DIR2/$tfile && error "link must fail" check_pdo_conflict $PID1 && { wait $PID1; error "link isn't blocked"; } - rm -r $DIR1/* + rm -rf $DIR/$tfile* return 0 } run_test 47c "pdirops: remote mkdir vs link" @@ -2185,7 +2346,7 @@ test_47d() { rmdir $DIR2/$tfile || error "unlink must succeed" check_pdo_conflict $PID1 && { wait $PID1; error "unlink isn't blocked"; } - rm -r $DIR1/* + rm -rf $DIR/$tfile* return 0 } run_test 47d "pdirops: remote mkdir vs unlink" @@ -2201,7 +2362,7 @@ test_47e() { mv -T $DIR2/$tfile-2 $DIR2/$tfile && error "rename must fail" check_pdo_conflict $PID1 && { wait $PID1; error "rename isn't blocked"; } - rm -r $DIR1/* + rm -rf $DIR/$tfile* return 0 } run_test 47e "pdirops: remote mkdir and rename (tgt)" @@ -2216,7 +2377,7 @@ test_47f() { mv $DIR2/$tfile $DIR2/$tfile-2 || error "rename must succeed" check_pdo_conflict $PID1 && { wait $PID1; error "rename isn't blocked"; } - rm -r $DIR1/* + rm -rf $DIR/$tfile* return 0 } run_test 47f "pdirops: remote mkdir and rename (src)" @@ -2231,7 +2392,7 @@ test_47g() { stat $DIR2/$tfile > /dev/null || error "stat must succeed" check_pdo_conflict $PID1 && { wait $PID1; error "getattr isn't blocked"; } - rm -r $DIR1/* + rm -rf $DIR/$tfile* return 0 } run_test 47g "pdirops: remote mkdir vs getattr" @@ -2253,13 +2414,13 @@ test_51a() { local filesize local origfile=/etc/hosts - filesize=`stat -c %s $origfile` + filesize=$(stat -c %s $origfile) # create an empty file - $MCREATE $DIR1/$tfile + $MCREATE $DIR1/$tfile || error "can't create $DIR1/$tfile" # cache layout lock on both mount point - stat $DIR1/$tfile > /dev/null - stat $DIR2/$tfile > /dev/null + stat $DIR1/$tfile > /dev/null || error "stat $DIR1/$tfile failed" + stat $DIR2/$tfile > /dev/null || error "stat $DIR2/$tfile failed" # open and sleep 2 seconds then read $MULTIOP $DIR2/$tfile o_2r${filesize}c & @@ -2267,12 +2428,13 @@ test_51a() { sleep 1 # create the layout of testing file - dd if=$origfile of=$DIR1/$tfile conv=notrunc > /dev/null + dd if=$origfile of=$DIR1/$tfile conv=notrunc > /dev/null || + error "dd $DIR1/$tfile failed" # MULTIOP proc should be able to read enough bytes and exit sleep 2 - kill -0 $pid && error "multiop is still there" - cmp $origfile $DIR2/$tfile || error "$MCREATE and $DIR2/$tfile differs" + kill -0 $pid 2> /dev/null && error "multiop is still there" + cmp $origfile $DIR2/$tfile || error "$origfile and $DIR2/$tfile differs" rm -f $DIR1/$tfile } @@ -2285,7 +2447,7 @@ test_51b() { local tmpfile=`mktemp` # create an empty file - $MCREATE $DIR1/$tfile + $MCREATE $DIR1/$tfile || error "mcreate $DIR1/$tfile failed" # delay glimpse so that layout has changed when glimpse finish #define OBD_FAIL_GLIMPSE_DELAY 0x1404 @@ -2295,10 +2457,11 @@ test_51b() { sleep 1 # create layout of testing file - dd if=/dev/zero of=$DIR1/$tfile bs=1k count=1 conv=notrunc > /dev/null + dd if=/dev/zero of=$DIR1/$tfile bs=1k count=1 conv=notrunc >/dev/null || + error "dd $DIR1/$tfile failed" wait $pid - local fsize=`cat $tmpfile` + local fsize=$(cat $tmpfile) [ x$fsize = x1024 ] || error "file size is $fsize, should be 1024" @@ -2307,29 +2470,223 @@ test_51b() { run_test 51b "layout lock: glimpse should be able to restart if layout changed" test_51c() { - # create an empty file - $MCREATE $DIR1/$tfile + [ $OSTCOUNT -ge 2 ] || { skip "need at least 2 osts"; return; } + + # set default layout to have 1 stripe + mkdir $DIR1/$tdir + $LFS setstripe -c 1 $DIR1/$tdir + + # create a file with empty layout + $MCREATE $DIR1/$tdir/$tfile || + error "$MCREATE $DIR1/$tdir/$tfile failed" #define OBD_FAIL_MDS_LL_BLOCK 0x172 - $LCTL set_param fail_loc=0x172 + do_facet $SINGLEMDS $LCTL set_param fail_loc=0x172 # change the layout of testing file - echo "Setting layout ..." - $LFS setstripe -c $OSTCOUNT $DIR1/$tfile & + echo "Setting layout to have $OSTCOUNT stripes ..." + $LFS setstripe -c $OSTCOUNT $DIR1/$tdir/$tfile & pid=$! sleep 1 - # get layout of this file should wait until dd is finished - local stripecnt=`$LFS getstripe -c $DIR2/$tfile` - [ $stripecnt -eq $OSTCOUNT ] || error "layout wrong" + # write something to the file, it should be blocked on fetching layout + dd if=/dev/zero of=$DIR2/$tdir/$tfile bs=1k count=1 conv=notrunc + local stripecnt=$($LFS getstripe -c $DIR2/$tdir/$tfile) + wait $pid - rm -f $DIR1/$tfile + # lod_qos.c::min_stripe_count() allows setstripe with a default stripe + # count to succeed with only 3/4 of the number of stripes (rounded up), + # so creating striped files does not fail if an OST is offline or full + [ $stripecnt -ge $((OSTCOUNT - $OSTCOUNT / 4)) ] || + error "layout wrong: getstripe -c $stripecnt < $OSTCOUNT * 3/4" + + rm -fr $DIR1/$tdir } run_test 51c "layout lock: IT_LAYOUT blocked and correct layout can be returned" +test_51d() { + dd if=/dev/zero of=/$DIR1/$tfile bs=1M count=1 + cancel_lru_locks mdc + + # open should grant LAYOUT lock, mmap and read will install pages + $MULTIOP $DIR1/$tfile oO_RDWR:SMR_Uc & + local PID=$! + sleep 1 + + # rss before revoking + local br=$(grep -A 10 $tfile /proc/$PID/smaps | awk '/^Rss/{print $2}') + echo "Before revoking layout lock: $br KB mapped" + + # delete the file will revoke layout lock + rm -f $DIR2/$tfile + + # rss after revoking + local ar=$(grep -A 10 $tfile /proc/$PID/smaps | awk '/^Rss/{print $2}') + + kill -USR1 $PID + wait $PID || error + + [ $ar -eq 0 ] || error "rss before: $br, after $ar, some pages remained" +} +run_test 51d "layout lock: losing layout lock should clean up memory map region" + +test_54_part1() +{ + echo "==> rename vs getattr vs setxattr should not deadlock" + mkdir -p $DIR/d1/d2/d3 || error "(1) mkdir failed" + + do_facet mds1 $LCTL set_param fail_loc=$1 + + mv -T $DIR/d1/d2/d3 $DIR/d1/d3 & + PID1=$! + sleep 1 + + stat $DIR/d1/d2 & + PID2=$! + sleep 1 + + setfattr -n user.attr1 -v value1 $DIR2/d1 || error "(2) setfattr failed" + wait $PID1 || error "(3) mv failed" + wait $PID2 || error "(4) stat failed" + echo + + rm -rf $DIR/d1 +} + +test_54_part2() { + echo "==> rename vs getattr vs open vs getattr should not deadlock" + mkdir -p $DIR/d1/d2/d3 || error "(1) mkdir failed" + + do_facet mds1 $LCTL set_param fail_loc=$1 + + mv -T $DIR/d1/d2/d3 $DIR/d1/d3 & + PID1=$! + sleep 1 + + stat $DIR/d1/d2 & + PID2=$! + sleep 1 + + $MULTIOP $DIR2/d1/d2 Oc & + PID3=$! + sleep 1 + + stat $DIR/d1 || error "(2) stat failed" + + wait $PID1 || error "(3) mv failed" + wait $PID2 || error "(4) stat failed" + wait $PID3 && error "(5) multiop failed" + echo + rm -rf $DIR/d1 +} + +test_54() { + local p="$TMP/$TESTSUITE-$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; } + +#define OBD_FAIL_MDS_RENAME 0x153 +#define OBD_FAIL_MDS_RENAME2 0x154 + test_54_part1 0x80000153 || error 10 + test_54_part1 0x80000154 || error 11 + test_54_part2 0x80000153 || error 12 + test_54_part2 0x80000154 || error 13 + + restore_lustre_params < $p + rm -f $p +} +run_test 54 "rename locking" + +test_55a() { + mkdir -p $DIR/d1/d2 $DIR/d3 || error "(1) mkdir failed" + +#define OBD_FAIL_MDS_RENAME4 0x156 + do_facet mds1 $LCTL set_param fail_loc=0x80000156 + + mv -T $DIR/d1/d2 $DIR/d3/d2 & + PID1=$! + sleep 1 + + rm -r $DIR2/d3 + wait $PID1 && error "(2) mv succeeded" + + rm -rf $DIR/d1 +} +run_test 55a "rename vs unlink target dir" + +test_55b() +{ + mkdir -p $DIR/d1/d2 $DIR/d3 || error "(1) mkdir failed" + +#define OBD_FAIL_MDS_RENAME4 0x156 + do_facet mds1 $LCTL set_param fail_loc=0x80000156 + + mv -T $DIR/d1/d2 $DIR/d3/d2 & + PID1=$! + sleep 1 + + rm -r $DIR2/d1 + wait $PID1 && error "(2) mv succeeded" + + rm -rf $DIR/d3 +} +run_test 55b "rename vs unlink source dir" + +test_55c() +{ + mkdir -p $DIR/d1/d2 $DIR/d3 || error "(1) mkdir failed" + +#define OBD_FAIL_MDS_RENAME4 0x156 + do_facet mds1 $LCTL set_param fail_loc=0x156 + + mv -T $DIR/d1/d2 $DIR/d3/d2 & + PID1=$! + sleep 1 + + # while rename is sleeping, open and remove d3 + $MULTIOP $DIR2/d3 D_c & + PID2=$! + sleep 1 + rm -rf $DIR2/d3 + sleep 5 + + # while rename is sleeping 2nd time, close d3 + kill -USR1 $PID2 + wait $PID2 || error "(3) multiop failed" + + wait $PID1 && error "(2) mv succeeded" + + rm -rf $DIR/d1 +} +run_test 55c "rename vs unlink orphan target dir" + +test_55d() +{ + touch $DIR/f1 + +#define OBD_FAIL_MDS_RENAME3 0x155 + do_facet mds1 $LCTL set_param fail_loc=0x155 + mv $DIR/f1 $DIR/$tdir & + PID1=$! + sleep 2 + + # while rename is sleeping, create $tdir, but as a directory + mkdir -p $DIR2/$tdir || error "(1) mkdir failed" + + # link in reverse locking order + ln $DIR2/f1 $DIR2/$tdir/ + + wait $PID1 && error "(2) mv succeeded" + rm -rf $DIR/f1 +} +run_test 55d "rename file vs link" + test_60() { - [[ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.3.0) ]] || - { skip "Need MDS version at least 2.3.0"; return; } + local MDSVER=$(lustre_build_version $SINGLEMDS) + [ $(version_code $MDSVER) -lt $(version_code 2.3.0) ] && + skip "MDS version $MDSVER must be >= 2.3.0" && return 0 + # Create a file test_mkdir -p $DIR1/$tdir file1=$DIR1/$tdir/file @@ -2413,7 +2770,7 @@ test_70b() { # LU-2781 } run_test 70b "remove files after calling rm_entry" -test_71() { +test_71a() { local server_version=$(lustre_version_code $SINGLEMDS) [[ $server_version -lt $(version_code 2.1.6) ]] && @@ -2422,7 +2779,8 @@ test_71() { # Patch not applied to 2.2 and 2.3 branches [[ $server_version -ge $(version_code 2.2.0) ]] && [[ $server_version -lt $(version_code 2.4.0) ]] && - skip "Need MDS version at least 2.4.0" && return + skip "Need MDS version earlier than 2.2.0 or at least 2.4.0" && + return checkfiemap --test || { skip "checkfiemap not runnable: $?" && return; } @@ -2454,7 +2812,34 @@ test_71() { echo $can2 [ $can3 -eq $can4 ] || error $((can2-can1)) "cancel RPC occured." } -run_test 71 "correct file map just after write operation is finished" +run_test 71a "correct file map just after write operation is finished" + +test_71b() { + local server_version=$(lustre_version_code $SINGLEMDS) + + [[ $server_version -lt $(version_code 2.1.6) ]] && + skip "Need MDS version at least 2.1.6" && return + + # Patch not applied to 2.2 and 2.3 branches + [[ $server_version -ge $(version_code 2.2.0) ]] && + [[ $server_version -lt $(version_code 2.4.0) ]] && + skip "Need MDS version earlier than 2.2.0 or at least 2.4.0" && + return + [[ $OSTCOUNT -ge 2 ]] || { skip "need at least 2 osts"; return; } + + checkfiemap --test || + { skip "error $?: checkfiemap failed" && return; } + + mkdir -p $DIR1/$tdir + + $LFS setstripe -c -1 $DIR1/$tdir || error "setstripe failed" + dd if=/dev/urandom of=$DIR1/$tdir/$tfile bs=40K count=1 + [ "$(facet_fstype ost$(($($GETSTRIPE -i $DIR1/$tdir/$tfile) + 1)))" = \ + "zfs" ] && + skip "ORI-366/LU-1941: FIEMAP unimplemented on ZFS" && return 0 + checkfiemap $DIR1/$tdir/$tfile 40960 || error "checkfiemap failed" +} +run_test 71b "check fiemap support for stripecount > 1" test_72() { local p="$TMP/sanityN-$TESTNAME.parameters" @@ -2499,14 +2884,16 @@ test_73() { error "setfattr1 failed" getfattr -n user.attr1 $DIR2/$tfile || error "getfattr1 failed" getfattr -n user.attr1 $DIR1/$tfile || error "getfattr2 failed" - clear_llite_stats + clear_stats llite.*.stats # PR lock should be cached by now on both clients getfattr -n user.attr1 $DIR1/$tfile || error "getfattr3 failed" # 2 hits for getfattr(0)+getfattr(size) - [ $(calc_llite_stats getxattr_hits) -eq 2 ] || error "not cached in $DIR1" + [ $(calc_stats llite.*.stats getxattr_hits) -eq 2 ] || + error "not cached in $DIR1" getfattr -n user.attr1 $DIR2/$tfile || error "getfattr4 failed" # 4 hits for more getfattr(0)+getfattr(size) - [ $(calc_llite_stats getxattr_hits) -eq 4 ] || error "not cached in $DIR2" + [ $(calc_stats llite.*.stats getxattr_hits) -eq 4 ] || + error "not cached in $DIR2" rm -f $DIR2/$tfile restore_lustre_params < $p @@ -2549,14 +2936,9 @@ run_test 75 "osc: upcall after unuse lock===================" test_76() { #LU-946 [[ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.5.53) ]] && skip "Need MDS version at least 2.5.53" && return - [ $MDSCOUNT -ge 2 ] && skip "skip now for LU-4573" && return #LU-4573 remote_mds_nodsh && skip "remote MDS with nodsh" && return - local fcount=2048 - local fd - local cmd - local mdt_idx - local mds_idx + local fcount=$((MDSCOUNT * 256)) declare -a fd_list declare -a fid_list @@ -2568,29 +2950,32 @@ test_76() { #LU-946 rm -rf $DIR/$tdir test_mkdir -p $DIR/$tdir - if [ $MDSCOUNT -gt 1 ]; then - mdt_idx=$($LFS getdirstripe -i $DIR/$tdir) - else - mdt_idx=0 - fi - mds_idx=$((mdt_idx + 1)) - proc_ofile="mdt.*$mdt_idx.exports.'$nid'.open_files" + # drop all open locks and close any cached "open" files on the client cancel_lru_locks mdc - echo -n "open files " - ulimit -n 8096 - for (( i = 0; i < $fcount; i++ )) ; do + local open_fids_cmd="$LCTL get_param -n mdt.*.exports.'$nid'.open_files" + local fid_list=($(do_nodes $(comma_list $(mdts_nodes)) $open_fids_cmd)) + local already=${#fid_list[@]} + for (( i = 0; i < $already; i++ )) ; do + log "already open[$i]: $($LFS fid2path $DIR2 ${fid_list[i]})" + done + + echo -n "opening files: " + ulimit -n $((fcount + 50)) + for ((i = 0; i < $fcount; i++)); do touch $DIR/$tdir/f_$i - fd=$(free_fd) - cmd="exec $fd<$DIR/$tdir/f_$i" - eval $cmd + local fd=$(free_fd ${fd_list[i]}) + local open_cmd="exec $fd<$DIR/$tdir/f_$i" + eval $open_cmd + fd_list[i]=$fd - echo -n "." + + (( $i % 32 == 0 )) && echo -n "." done echo - fid_list=($(do_facet mds$mds_idx $LCTL get_param -n $proc_ofile)) + fid_list=($(do_nodes $(comma_list $(mdts_nodes)) $open_fids_cmd)) # Possible errors in openfiles FID list. # 1. Missing FIDs. Check 1 @@ -2600,19 +2985,28 @@ test_76() { #LU-946 # 5. Valid FID, points to some other file. Check 3 # Check 1 - [ ${#fid_list[@]} -ne $fcount ] && - error "${#fid_list[@]} != $fcount open files" - - for (( i = 0; i < $fcount; i++ )) ; do - cmd="exec ${fd_list[i]} /dev/null || + error "dd at ${i}MB on client failed (3)" & + local pids_r[$i]=$! + done + cancel_lru_locks osc + + for ((i = 0; i < $n; i++)); do + wait ${pids_w[$i]} + wait ${pids_r[$i]} + done + rm -rf $dir || error "rm -rf $dir failed" +} + +test_77a() { #LU-3266 + oss=$(comma_list $(osts_nodes)) + do_nodes $oss lctl set_param ost.OSS.*.nrs_policies="fifo" + nrs_write_read + + return 0 +} +run_test 77a "check FIFO NRS policy" + +test_77b() { #LU-3266 + oss=$(comma_list $(osts_nodes)) + + do_nodes $oss lctl set_param ost.OSS.*.nrs_policies="crrn" \ + ost.OSS.*.nrs_crrn_quantum=1 + + echo "policy: crr-n, crrn_quantum 1" + nrs_write_read + + do_nodes $oss lctl set_param ost.OSS.*.nrs_crrn_quantum=64 + + echo "policy: crr-n, crrn_quantum 64" + nrs_write_read + + # cleanup + do_nodes $oss lctl set_param ost.OSS.ost_io.nrs_policies="fifo" + return 0 +} +run_test 77b "check CRR-N NRS policy" + +orr_trr() { + local policy=$1 + + oss=$(comma_list $(osts_nodes)) + + do_nodes $oss lctl set_param ost.OSS.ost_io.nrs_policies=$policy \ + ost.OSS.*.nrs_"$policy"_quantum=1 \ + ost.OSS.*.nrs_"$policy"_offset_type="physical" \ + ost.OSS.*.nrs_"$policy"_supported="reads" + + echo "policy: $policy, ${policy}_quantum 1, ${policy}_offset_type physical, ${policy}_supported reads" + nrs_write_read + + do_nodes $oss lctl set_param ost.OSS.*.nrs_${policy}_supported="writes" \ + ost.OSS.*.nrs_${policy}_quantum=64 + + echo "policy: $policy, ${policy}_quantum 64, ${policy}_offset_type physical, ${policy}_supported writes" + nrs_write_read + + do_nodes $oss lctl set_param ost.OSS.*.nrs_${policy}_supported="reads_and_writes" \ + ost.OSS.*.nrs_${policy}_offset_type="logical" + echo "policy: $policy, ${policy}_quantum 64, ${policy}_offset_type logical, ${policy}_supported reads_and_writes" + nrs_write_read + + # cleanup + do_nodes $oss lctl set_param ost.OSS.ost_io.nrs_policies="fifo" + return 0 +} + +test_77c() { #LU-3266 + orr_trr "orr" + return 0 +} +run_test 77c "check ORR NRS policy" + +test_77d() { #LU-3266 + orr_trr "trr" + return 0 +} +run_test 77d "check TRR nrs policy" + +tbf_rule_operate() +{ + local facet=$1 + shift 1 + + do_facet $facet lctl set_param \ + ost.OSS.ost_io.nrs_tbf_rule="$*" + [ $? -ne 0 ] && + error "failed to run operate '$*' on TBF rules" +} + +tbf_verify() { + local dir=$DIR/$tdir + local client1=${CLIENT1:-$(hostname)} + local myRUNAS="$3" + + mkdir $dir || error "mkdir $dir failed" + $LFS setstripe -c 1 $dir || error "setstripe to $dir failed" + chmod 777 $dir + + echo "Limited write rate: $1, read rate: $2" + echo "Verify the write rate is under TBF control" + local rate=$(do_node $client1 $myRUNAS dd if=/dev/zero of=$dir/tbf \ + bs=1M count=100 oflag=direct 2>&1 | awk '/bytes/ {print $8}') + echo "Write speed is $rate" + + # verify the write rate does not exceed 110% of TBF limited rate + [ $(bc <<< "$rate < 1.1 * $1") -eq 1 ] || + error_ignore LU-9140 "The write rate ($rate) exceeds 110% of preset rate ($1)" + + cancel_lru_locks osc + + echo "Verify the read rate is under TBF control" + rate=$(do_node $client1 $myRUNAS dd if=$dir/tbf of=/dev/null \ + bs=1M count=100 iflag=direct 2>&1 | awk '/bytes/ {print $8}') + echo "Read speed is $rate" + + # verify the read rate does not exceed 110% of TBF limited rate + [ $(bc <<< "$rate < 1.1 * $2") -eq 1 ] || + error_ignore LU-9140 "The read rate ($rate) exceeds 110% of preset rate ($2)" + + cancel_lru_locks osc + rm -rf $dir || error "rm -rf $dir failed" +} + +test_77e() { + local server_version=$(lustre_version_code ost1) + [[ $server_version -ge $(version_code 2.7.58) ]] || + { skip "Need server version newer than 2.7.57"; return 0; } + + oss=$(comma_list $(osts_nodes)) + + do_nodes $oss lctl set_param ost.OSS.ost_io.nrs_policies="tbf\ nid" + [ $? -ne 0 ] && error "failed to set TBF policy" + + local idis + local rateis + if [ $(lustre_version_code ost1) -ge $(version_code 2.8.54) ]; then + idis="nid=" + rateis="rate=" + fi + + # Only operate rules on ost1 since OSTs might run on the same OSS + # Add some rules + tbf_rule_operate ost1 "start\ localhost\ ${idis}{0@lo}\ ${rateis}1000" + local address=$(comma_list "$(host_nids_address $CLIENTS $NETTYPE)") + local client_nids=$(nids_list $address "\\") + tbf_rule_operate ost1 "start\ clients\ ${idis}{$client_nids}\ ${rateis}100" + tbf_rule_operate ost1 "start\ others\ ${idis}{*.*.*.*@$NETTYPE}\ ${rateis}50" + nrs_write_read + + # Change the rules + tbf_rule_operate ost1 "change\ localhost\ ${rateis}1001" + tbf_rule_operate ost1 "change\ clients\ ${rateis}101" + tbf_rule_operate ost1 "change\ others\ ${rateis}51" + nrs_write_read + + # Stop the rules + tbf_rule_operate ost1 "stop\ localhost" + tbf_rule_operate ost1 "stop\ clients" + tbf_rule_operate ost1 "stop\ others" + nrs_write_read + + # Cleanup the TBF policy + do_nodes $oss lctl set_param ost.OSS.ost_io.nrs_policies="fifo" + [ $? -ne 0 ] && error "failed to set policy back to fifo" + nrs_write_read + return 0 +} +run_test 77e "check TBF NID nrs policy" + +test_77f() { + local server_version=$(lustre_version_code ost1) + [[ $server_version -ge $(version_code 2.7.58) ]] || + { skip "Need server version newer than 2.7.57"; return 0; } + + oss=$(comma_list $(osts_nodes)) + + # Configure jobid_var + local saved_jobid_var=$($LCTL get_param -n jobid_var) + if [ $saved_jobid_var != procname_uid ]; then + set_conf_param_and_check client \ + "$LCTL get_param -n jobid_var" \ + "$FSNAME.sys.jobid_var" procname_uid + fi + + do_nodes $oss lctl set_param ost.OSS.ost_io.nrs_policies="tbf\ jobid" + [ $? -ne 0 ] && error "failed to set TBF policy" + + local idis + local rateis + if [ $(lustre_version_code ost1) -ge $(version_code 2.8.54) ]; then + idis="jobid=" + rateis="rate=" + fi + + # Only operate rules on ost1 since OSTs might run on the same OSS + # Add some rules + tbf_rule_operate ost1 "start\ runas\ ${idis}{iozone.$RUNAS_ID\ dd.$RUNAS_ID\ tiotest.$RUNAS_ID}\ ${rateis}1000" + tbf_rule_operate ost1 "start\ iozone_runas\ ${idis}{iozone.$RUNAS_ID}\ ${rateis}100" + tbf_rule_operate ost1 "start\ dd_runas\ ${idis}{dd.$RUNAS_ID}\ ${rateis}50" + nrs_write_read "$RUNAS" + + # Change the rules + tbf_rule_operate ost1 "change\ runas\ ${rateis}1001" + tbf_rule_operate ost1 "change\ iozone_runas\ ${rateis}101" + tbf_rule_operate ost1 "change\ dd_runas\ ${rateis}51" + nrs_write_read "$RUNAS" + + # Stop the rules + tbf_rule_operate ost1 "stop\ runas" + tbf_rule_operate ost1 "stop\ iozone_runas" + tbf_rule_operate ost1 "stop\ dd_runas" + nrs_write_read "$RUNAS" + + # Cleanup the TBF policy + do_nodes $oss lctl set_param ost.OSS.ost_io.nrs_policies="fifo" + [ $? -ne 0 ] && error "failed to set policy back to fifo" + nrs_write_read "$RUNAS" + + local current_jobid_var=$($LCTL get_param -n jobid_var) + if [ $saved_jobid_var != $current_jobid_var ]; then + set_conf_param_and_check client \ + "$LCTL get_param -n jobid_var" \ + "$FSNAME.sys.jobid_var" $saved_jobid_var + fi + return 0 +} +run_test 77f "check TBF JobID nrs policy" + +test_77g() { + local server_version=$(lustre_version_code ost1) + [[ $server_version -ge $(version_code 2.7.58) ]] || + { skip "Need server version newer than 2.7.57"; return 0; } + + oss=$(comma_list $(osts_nodes)) + + do_nodes $oss lctl set_param ost.OSS.ost_io.nrs_policies="tbf\ nid" + [ $? -ne 0 ] && error "failed to set TBF policy" + + do_nodes $oss lctl set_param ost.OSS.ost_io.nrs_policies="tbf\ jobid" + [ $? -ne 0 ] && error "failed to set TBF policy" + + local idis + local rateis + if [ $(lustre_version_code ost1) -ge $(version_code 2.8.54) ]; then + idis="jobid=" + rateis="rate=" + fi + + # Add a rule that only valid for Jobid TBF. If direct change between + # TBF types is not supported, this operation will fail. + tbf_rule_operate ost1 "start\ dd_runas\ ${idis}{dd.$RUNAS_ID}\ ${rateis}50" + + # Cleanup the TBF policy + do_nodes $oss lctl set_param ost.OSS.ost_io.nrs_policies="fifo" + [ $? -ne 0 ] && error "failed to set policy back to fifo" + return 0 +} +run_test 77g "Change TBF type directly" + +test_77h() { + [ $(lustre_version_code ost1) -ge $(version_code 2.8.55) ] || + { skip "Need OST version at least 2.8.55"; return 0; } + + local old_policy=$(do_facet ost1 \ + lctl get_param ost.OSS.ost_io.nrs_policies) + local new_policy + + do_facet ost1 lctl set_param \ + ost.OSS.ost_io.nrs_policies="abc" + [ $? -eq 0 ] && error "should return error" + + do_facet ost1 lctl set_param \ + ost.OSS.ost_io.nrs_policies="tbf\ abc" + [ $? -eq 0 ] && error "should return error" + + do_facet ost1 lctl set_param \ + ost.OSS.ost_io.nrs_policies="tbf\ reg\ abc" + [ $? -eq 0 ] && error "should return error" + + do_facet ost1 lctl set_param \ + ost.OSS.ost_io.nrs_policies="tbf\ abc\ efg" + [ $? -eq 0 ] && error "should return error" + + new_policy=$(do_facet ost1 lctl get_param ost.OSS.ost_io.nrs_policies) + [ $? -eq 0 ] || error "shouldn't LBUG" + + [ "$old_policy" = "$new_policy" ] || error "NRS policy should be same" + + return 0 +} +run_test 77h "Wrong policy name should report error, not LBUG" + +tbf_rule_check() +{ + local facet=$1 + local expected=$2 + local error_message=$3 + local rule_number=0 + for rule in $expected; do + rule_number=$((rule_number + 1)) + done + local stop_line=$(($rule_number + 3)) + local awk_command="awk 'NR >= 4 && NR <= $stop_line {print \$1}'" + + local output=$(do_facet $facet lctl get_param \ + ost.OSS.ost_io.nrs_tbf_rule | + eval $awk_command | + tr "\n" " " | + sed 's/[ ]*$//') + if [ "$output" != "$expected" ]; then + error "$error_message, expected '$expected', got '$output'" + fi +} + +test_77i() { + [ $(lustre_version_code ost1) -ge $(version_code 2.8.55) ] || + { skip "Need OST version at least 2.8.55"; return 0; } + + for i in $(seq 1 $OSTCOUNT) + do + do_facet ost"$i" lctl set_param \ + ost.OSS.ost_io.nrs_policies="tbf\ jobid" + [ $? -ne 0 ] && + error "failed to set TBF policy" + done + + tbf_rule_check ost1 "default" "error before inserting any rule" + + tbf_rule_operate ost1 "start\ before\ jobid={jobid}\ rate=1000" + tbf_rule_check ost1 "before default" \ + "error when inserting rule 'before'" + + tbf_rule_operate ost1 "start\ after\ jobid={jobid}\ rate=1000\ rank=default" + tbf_rule_check ost1 "before after default" \ + "error when inserting rule 'after'" + + tbf_rule_operate ost1 "start\ target\ jobid={jobid}\ rate=1000\ rank=after" + tbf_rule_check ost1 "before target after default" \ + "error when inserting rule 'target'" + + echo "Move before itself" + tbf_rule_operate ost1 "change\ target\ rank=target" + tbf_rule_check ost1 "before target after default" \ + "error when moving before itself" + + echo "Move to higher rank" + tbf_rule_operate ost1 "change\ target\ rank=before" + tbf_rule_check ost1 "target before after default" \ + "error when moving to higher rank" + + echo "Move to lower rank" + tbf_rule_operate ost1 "change\ target\ rank=after" + tbf_rule_check ost1 "before target after default" \ + "error when moving to lower rank" + + echo "Move before default" + tbf_rule_operate ost1 "change\ target\ rank=default" + tbf_rule_check ost1 "before after target default" \ + error "error when moving before default" + + # Cleanup the TBF policy + do_nodes $(comma_list $(osts_nodes)) \ + $LCTL set_param ost.OSS.ost_io.nrs_policies=fifo + return 0 +} +run_test 77i "Change rank of TBF rule" + +test_77j() { + local idis + local rateis + if [ $(lustre_version_code ost1) -ge $(version_code 2.8.60) ]; then + idis="opcode=" + rateis="rate=" + fi + + do_nodes $(comma_list $(osts_nodes)) \ + lctl set_param jobid_var=procname_uid \ + ost.OSS.ost_io.nrs_policies="tbf\ opcode" \ + ost.OSS.ost_io.nrs_tbf_rule="start\ ost_r\ ${idis}{ost_read}\ ${rateis}5" \ + ost.OSS.ost_io.nrs_tbf_rule="start\ ost_w\ ${idis}{ost_write}\ ${rateis}20" + [ $? -ne 0 ] && error "failed to set TBF OPCode policy" + + nrs_write_read + tbf_verify 20 5 + + do_nodes $(comma_list $(osts_nodes)) \ + lctl set_param ost.OSS.ost_io.nrs_tbf_rule="stop\ ost_r" \ + ost.OSS.ost_io.nrs_tbf_rule="stop\ ost_w" \ + ost.OSS.ost_io.nrs_policies="fifo" + + # sleep 3 seconds to wait the tbf policy stop completely, + # or the next test case is possible get -EAGAIN when + # setting the tbf policy + sleep 3 +} +run_test 77j "check TBF-OPCode NRS policy" + +test_77k() { + [[ $(lustre_version_code ost1) -ge $(version_code 2.9.53) ]] || + { skip "Need OST version at least 2.9.53"; return 0; } + + do_nodes $(comma_list $(osts_nodes)) \ + lctl set_param ost.OSS.ost_io.nrs_policies="tbf" \ + ost.OSS.ost_io.nrs_tbf_rule="start\ ext_w\ jobid={dd.$RUNAS_ID}\&opcode={ost_write}\ rate=20" \ + ost.OSS.ost_io.nrs_tbf_rule="start\ ext_r\ jobid={dd.$RUNAS_ID}\&opcode={ost_read}\ rate=10" + + nrs_write_read "$RUNAS" + tbf_verify 20 10 "$RUNAS" + + local address=$(comma_list "$(host_nids_address $CLIENTS $NETTYPE)") + local client_nids=$(nids_list $address "\\") + do_nodes $(comma_list $(osts_nodes)) \ + lctl set_param ost.OSS.ost_io.nrs_tbf_rule="stop\ ext_w" \ + ost.OSS.ost_io.nrs_tbf_rule="stop\ ext_r" \ + ost.OSS.ost_io.nrs_tbf_rule="start\ ext_w\ nid={0@lo\ $client_nids}\&opcode={ost_write}\ rate=20" \ + ost.OSS.ost_io.nrs_tbf_rule="start\ ext_r\ nid={0@lo\ $client_nids}\&opcode={ost_read}\ rate=10" + + nrs_write_read + tbf_verify 20 10 + + do_nodes $(comma_list $(osts_nodes)) \ + lctl set_param ost.OSS.ost_io.nrs_tbf_rule="stop\ ext_w" \ + ost.OSS.ost_io.nrs_tbf_rule="stop\ ext_r" \ + ost.OSS.ost_io.nrs_tbf_rule="start\ ext\ nid={0@lo\ $client_nids}\&jobid={dd.$RUNAS_ID}\ rate=20" + + nrs_write_read "$RUNAS" + tbf_verify 20 20 "$RUNAS" + + do_nodes $(comma_list $(osts_nodes)) \ + lctl set_param ost.OSS.ost_io.nrs_tbf_rule="stop\ ext" \ + ost.OSS.ost_io.nrs_tbf_rule="start\ ext_a\ jobid={dd.$RUNAS_ID},opcode={ost_write}\ rate=20" \ + ost.OSS.ost_io.nrs_tbf_rule="start\ ext_b\ jobid={dd.$RUNAS_ID},opcode={ost_read}\ rate=10" + + nrs_write_read "$RUNAS" + # with parameter "RUNAS", it will match the latest rule + # "ext_b" first, so the limited write rate is 10. + tbf_verify 10 10 "$RUNAS" + tbf_verify 20 10 + + do_nodes $(comma_list $(osts_nodes)) \ + lctl set_param ost.OSS.ost_io.nrs_tbf_rule="stop\ ext_a" \ + ost.OSS.ost_io.nrs_tbf_rule="stop\ ext_b" \ + ost.OSS.ost_io.nrs_policies="fifo" + + sleep 3 +} +run_test 77k "check the extended TBF policy with NID/JobID/OPCode expression" + +test_77l() { + if [ $(lustre_version_code ost1) -lt $(version_code 2.9.54) ]; then + skip "Need OST version at least 2.9.54" + return 0 + fi + + local dir=$DIR/$tdir + + mkdir $dir || error "mkdir $dir failed" + $LFS setstripe -c $OSTCOUNT $dir || error "setstripe to $dir failed" + chmod 777 $dir + + local nodes=$(comma_list $(osts_nodes)) + do_nodes $nodes lctl set_param ost.OSS.ost_io.nrs_policies=delay \ + ost.OSS.ost_io.nrs_delay_min=4 \ + ost.OSS.ost_io.nrs_delay_max=4 \ + ost.OSS.ost_io.nrs_delay_pct=100 + [ $? -ne 0 ] && error "Failed to set delay policy" + + local start=$SECONDS + do_nodes "${SINGLECLIENT:-$HOSTNAME}" "$RUNAS" \ + dd if=/dev/zero of="$dir/nrs_delay_$HOSTNAME" bs=1M count=1 \ + oflag=direct conv=fdatasync || + { do_nodes $nodes lctl set_param ost.OSS.ost_io.nrs_policies="fifo"; + error "dd on client failed (1)"; } + local elapsed=$((SECONDS - start)) + + # NRS delay doesn't do sub-second timing, so a request enqueued at + # 0.9 seconds can be dequeued at 4.0 + [ $elapsed -lt 3 ] && + { do_nodes $nodes lctl set_param ost.OSS.ost_io.nrs_policies="fifo"; + error "Single 1M write should take at least 3 seconds"; } -test_80() { + start=$SECONDS + do_nodes "${SINGLECLIENT:-$HOSTNAME}" "$RUNAS" \ + dd if=/dev/zero of="$dir/nrs_delay_$HOSTNAME" bs=1M count=10 \ + oflag=direct conv=fdatasync || + { do_nodes $nodes lctl set_param ost.OSS.ost_io.nrs_policies="fifo"; + error "dd on client failed (2)"; } + elapsed=$((SECONDS - start)) + + [ $elapsed -lt 30 ] && + { do_nodes $nodes lctl set_param ost.OSS.ost_io.nrs_policies="fifo"; + error "Ten 1M writes should take at least 30 seconds"; } + + do_nodes $nodes lctl set_param ost.OSS.ost_io.nrs_policies="fifo" + [ $? -ne 0 ] && error "failed to set policy back to fifo" + + return 0 +} +run_test 77l "check NRS Delay slows write RPC processing" + +test_78() { #LU-6673 + local server_version=$(lustre_version_code ost1) + [[ $server_version -ge $(version_code 2.7.58) ]] || + { skip "Need server version newer than 2.7.57"; return 0; } + + local rc + + oss=$(comma_list $(osts_nodes)) + do_nodes $oss lctl set_param ost.OSS.ost_io.nrs_policies="orr" & + do_nodes $oss lctl set_param ost.OSS.*.nrs_orr_quantum=1 + rc=$? + # Valid return codes are: + # 0: Tuning succeeded + # ENODEV: Policy is still stopped + # EAGAIN: Policy is being initialized + [ $rc -eq 0 -o $rc -eq 19 -o $rc -eq 11 ] || + error "Expected set_param to return 0|ENODEV|EAGAIN" + + # Cleanup the ORR policy + do_nodes $oss lctl set_param ost.OSS.ost_io.nrs_policies="fifo" + [ $? -ne 0 ] && error "failed to set policy back to fifo" + return 0 +} +run_test 78 "Enable policy and specify tunings right away" + +test_79() { + remote_mds_nodsh && skip "remote MDS with nodsh" && return + test_mkdir -p $DIR/$tdir + + # Prevent interference from layout intent RPCs due to + # asynchronous writeback. These will be tested in 130c below. + do_nodes ${CLIENTS:-$HOSTNAME} sync + + setfattr -n trusted.name1 -v value1 $DIR/$tdir || + error "setfattr -n trusted.name1=value1 $DIR/$tdir failed" + +#define OBD_FAIL_MDS_INTENT_DELAY 0x160 + local mdtidx=$($LFS getstripe -M $DIR/$tdir) + local facet=mds$((mdtidx + 1)) + stat $DIR/$tdir + set_nodes_failloc $(facet_active_host $facet) 0x80000160 + getfattr -n trusted.name1 $DIR/$tdir 2> /dev/null & + local pid=$! + sleep 2 + +#define OBD_FAIL_MDS_GETXATTR_PACK 0x131 + set_nodes_failloc $(facet_active_host $facet) 0x80000131 + + wait $pid + return 0 +} +run_test 79 "xattr: intent error" + +test_80a() { [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return local MDTIDX=1 local mdt_index @@ -2643,12 +3619,12 @@ test_80() { # give multiop a chance to open sleep 1 - $LFS mv -M $MDTIDX $DIR1/$tdir && + $LFS migrate -m $MDTIDX $DIR1/$tdir && error "migrate open files should failed with open files" kill -USR1 $pid - $LFS mv -M $MDTIDX $DIR1/$tdir || + $LFS migrate -m $MDTIDX $DIR1/$tdir || error "migrate remote dir error" echo "Finish migration, then checking.." @@ -2663,11 +3639,357 @@ test_80() { rm -rf $DIR1/$tdir || error "rm dir failed after migration" } -run_test 80 "migrate directory when some children is being opened" +run_test 80a "migrate directory when some children is being opened" + +cleanup_80b() { + trap 0 + kill -9 $migrate_pid +} + +test_80b() { + [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return + local migrate_dir1=$DIR1/$tdir/migrate_dir + local migrate_dir2=$DIR2/$tdir/migrate_dir + local migrate_run=$LUSTRE/tests/migrate.sh + local start_time + local end_time + local show_time=1 + local mdt_idx + local rc=0 + local rc1=0 + + trap cleanup_80b EXIT + #prepare migrate directory + mkdir -p $migrate_dir1 + for F in {1,2,3,4,5}; do + echo "$F$F$F$F$F" > $migrate_dir1/file$F + echo "$F$F$F$F$F" > $DIR/$tdir/file$F + done + + #migrate the directories among MDTs + ( + while true; do + mdt_idx=$((RANDOM % MDSCOUNT)) + $LFS migrate -m $mdt_idx $migrate_dir1 2&>/dev/null || + rc=$? + [ $rc -ne 0 -o $rc -ne 16 ] || break + done + ) & + migrate_pid=$! + + echo "start migration thread $migrate_pid" + #Access the files at the same time + start_time=$(date +%s) + echo "accessing the migrating directory for 5 minutes..." + while true; do + ls $migrate_dir2 > /dev/null || { + echo "read dir fails" + break + } + diff -u $DIR2/$tdir/file1 $migrate_dir2/file1 || { + echo "access file1 fails" + break + } + + cat $migrate_dir2/file2 > $migrate_dir2/file3 || { + echo "access file2/3 fails" + break + } + + echo "aaaaa" > $migrate_dir2/file4 > /dev/null || { + echo "access file4 fails" + break + } + + stat $migrate_dir2/file5 > /dev/null || { + echo "stat file5 fails" + break + } + + touch $migrate_dir2/source_file > /dev/null || rc1=$? + [ $rc1 -ne 0 -o $rc1 -ne 1 ] || { + echo "touch file failed with $rc1" + break; + } + + if [ -e $migrate_dir2/source_file ]; then + ln $migrate_dir2/source_file $migrate_dir2/link_file \ + 2&>/dev/null || rc1=$? + if [ -e $migrate_dir2/link_file ]; then + rm -rf $migrate_dir2/link_file + fi + + mrename $migrate_dir2/source_file \ + $migrate_dir2/target_file 2&>/dev/null || rc1=$? + [ $rc1 -ne 0 -o $rc1 -ne 1 ] || { + echo "rename failed with $rc1" + break + } + + if [ -e $migrate_dir2/target_file ]; then + rm -rf $migrate_dir2/target_file 2&>/dev/null || + rc1=$? + else + rm -rf $migrate_dir2/source_file 2&>/dev/null || + rc1=$? + fi + [ $rc1 -ne 0 -o $rc1 -ne 1 ] || { + echo "unlink failed with $rc1" + break + } + fi + + end_time=$(date +%s) + duration=$((end_time - start_time)) + if [ $((duration % 10)) -eq 0 ]; then + if [ $show_time -eq 1 ]; then + echo "...$duration seconds" + show_time=0 + fi + else + show_time=1 + fi + + kill -0 $migrate_pid || { + echo "migration stopped 1" + break + } + + [ $duration -ge 300 ] && break + done + + #check migration are still there + kill -0 $migrate_pid || error "migration stopped 2" + cleanup_80b +} +run_test 80b "Accessing directory during migration" + +test_81() { + [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return + + rm -rf $DIR1/$tdir + + mkdir -p $DIR1/$tdir + + $LFS setdirstripe -i0 -c$MDSCOUNT $DIR1/$tdir/d0 + $LFS setdirstripe -i0 -c$MDSCOUNT $DIR1/$tdir/d1 + + cd $DIR1/$tdir + touch d0/0 || error "create 0 failed" + mv d0/0 d1/0 || error "rename d0/0 d1/0 failed" + stat d0/0 && error "stat mv filed succeed" + mv $DIR2/$tdir/d1/0 $DIR2/$tdir/d0/0 || error "rename d1/0 d0/0 failed" + stat d0/0 || error "stat failed" + + local t=$(ls -ai $DIR1/$tdir/d0 | sort -u | wc -l) + + if [ $t -ne 3 ]; then + ls -ai $DIR1/$tdir/d0 + error "expect 3 get $t" + fi + + return 0 +} +run_test 81 "rename and stat under striped directory" + +test_82() { + [[ $(lustre_version_code $SINGLEMDS) -gt $(version_code 2.6.91) ]] || + { skip "Need MDS version at least 2.6.92"; return 0; } + + # Client 1 creates a file. + multiop_bg_pause $DIR1/$tfile O_ac || error "multiop_bg_pause 1" + pid1=$! + # Client 2 opens the file. + multiop_bg_pause $DIR2/$tfile o_Ac || error "multiop_bg_pause 2" + pid2=$! + # Client 1 makes the file an orphan. + rm $DIR1/$tfile || error "rm" + # Client 2 sets EA "user.multiop". + kill -s USR1 $pid2 + wait $pid2 || error "multiop 2" + # Client 1 gets EA "user.multiop". This used to fail because the EA + # cache refill would get "trusted.link" from mdd_xattr_list() but + # -ENOENT when trying to get "trusted.link"'s value. See also sanity + # 102q. + kill -s USR1 $pid1 + wait $pid1 || error "multiop 1" +} +run_test 82 "fsetxattr and fgetxattr on orphan files" + +test_83() { + [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return + local pid1 + local pid2 + + ( + cd $DIR1 + while true; do + $LFS mkdir -i1 -c2 $tdir + rmdir $tdir + done + ) & + pid1=$! + echo "start pid $pid1 to create/unlink striped directory" + + # Access the directory at the same time + ( + cd $DIR2 + while true; do + stat $tdir > /dev/null 2>&1 + done + ) & + pid2=$! + echo "start pid $pid2 to stat striped directory" + + sleep 120 + kill $pid1 $pid2 + wait $pid1 $pid2 + + return 0 +} +run_test 83 "access striped directory while it is being created/unlinked" + +test_90() { + [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return + local pid1 + local pid2 + local duration=180 + + [ "$SLOW" = "yes" ] && duration=600 + # Open/Create under striped directory + ( + cd $DIR1 + while true; do + $LFS mkdir -c$MDSCOUNT $tdir > /dev/null 2>&1 + touch $tdir/f{0..3} > /dev/null 2>&1 + done + ) & + pid1=$! + echo "start pid $pid1 to open/create under striped directory" + + # unlink the striped directory at the same time + ( + cd $DIR2 + while true; do + rm -rf $tdir > /dev/null 2>&1 + done + ) & + pid2=$! + echo "start pid $pid2 to unlink striped directory" + + sleep $duration + + kill $pid1 $pid2 + wait $pid1 $pid2 + + return 0 +} +run_test 90 "open/create and unlink striped directory" + +test_91() { + [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return + local pid1 + local pid2 + local duration=180 + + [ "$SLOW" = "yes" ] && duration=600 + # chmod striped directory + ( + cd $DIR1 + while true; do + $LFS mkdir -c$MDSCOUNT $tdir > /dev/null 2>&1 + chmod go+w $tdir > /dev/null 2>&1 + done + ) & + pid1=$! + echo "start pid $pid1 to chmod striped directory" + + # unlink the striped directory at the same time + ( + cd $DIR2 + while true; do + rm -rf $tdir > /dev/null 2>&1 + done + ) & + pid2=$! + echo "start pid $pid2 to unlink striped directory" + + sleep $duration + + kill $pid1 $pid2 + wait $pid1 $pid2 + + return 0 +} +run_test 91 "chmod and unlink striped directory" + +test_92() { + [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return + + local fd=$(free_fd) + local cmd="exec $fd<$DIR1/$tdir" + $LFS setdirstripe -c$MDSCOUNT $DIR1/$tdir || error "mkdir $tdir fails" + eval $cmd + cmd="exec $fd<&-" + trap "eval $cmd" EXIT + cd $DIR1/$tdir || error "cd $DIR1/$tdir fails" + rmdir ../$tdir || error "rmdir ../$tdir fails" + + #define OBD_FAIL_LLITE_NO_CHECK_DEAD 0x1408 + $LCTL set_param fail_loc=0x1408 + mkdir $DIR2/$tdir/dir && error "create dir succeeds" + $LFS setdirstripe -i1 $DIR2/$tdir/remote_dir && + error "create remote dir succeeds" + $LCTL set_param fail_loc=0 + eval $cmd + return 0 +} +run_test 92 "create remote directory under orphan directory" + +test_93() { + local rc1=0 + local rc2=0 + local old_rr + + mkdir -p $DIR1/$tfile-1/ + mkdir -p $DIR2/$tfile-2/ + local old_rr=$(do_facet $SINGLEMDS lctl get_param -n \ + 'lod.lustre-MDT*/qos_threshold_rr' | sed -e 's/%//') + do_facet $SINGLEMDS lctl set_param -n \ + 'lod.lustre-MDT*/qos_threshold_rr' 100 + #define OBD_FAIL_MDS_LOV_CREATE_RACE 0x163 + do_facet $SINGLEMDS "lctl set_param fail_loc=0x00000163" + + $SETSTRIPE -c -1 $DIR1/$tfile-1/file1 & + local PID1=$! + sleep 1 + $SETSTRIPE -c -1 $DIR2/$tfile-2/file2 & + local PID2=$! + wait $PID2 + wait $PID1 + do_facet $SINGLEMDS "lctl set_param fail_loc=0x0" + do_facet $SINGLEMDS "lctl set_param -n \ + 'lod.lustre-MDT*/qos_threshold_rr' $old_rr" + + $GETSTRIPE $DIR1/$tfile-1/file1 + rc1=$($GETSTRIPE -q $DIR1/$tfile-1/file1 | + awk '{if (/[0-9]/) print $1 }' | sort | uniq -d | wc -l) + $GETSTRIPE $DIR2/$tfile-2/file2 + rc2=$($GETSTRIPE -q $DIR2/$tfile-2/file2 | + awk '{if (/[0-9]/) print $1 }' | sort | uniq -d | wc -l) + echo "rc1=$rc1 and rc2=$rc2 " + [ $rc1 -eq 0 ] && [ $rc2 -eq 0 ] || + error "object allocate on same ost detected" +} +run_test 93 "alloc_rr should not allocate on same ost" log "cleanup: ======================================================" -[ "$(mount | grep $MOUNT2)" ] && umount $MOUNT2 +# kill and wait in each test only guarentee script finish, but command in script +# like 'rm' 'chmod' may still be running, wait for all commands to finish +# otherwise umount below will fail +[ "$(mount | grep $MOUNT2)" ] && wait_update $HOSTNAME "fuser -m $MOUNT2" "" || + true complete $SECONDS rm -f $SAMPLE_FILE