X-Git-Url: https://git.whamcloud.com/?p=fs%2Flustre-release.git;a=blobdiff_plain;f=lustre%2Ftests%2Fsanity.sh;h=ea9a8ccfe93b7860cfaa61d6cec2f6562203f107;hp=ae63acb3ad31797b339fb65ec6df39afa3251ca9;hb=9b3a45e95e231d17926ef5cffc604a27e271f66e;hpb=69fba62b74408c3978a07f2886662b9f4d1e4182 diff --git a/lustre/tests/sanity.sh b/lustre/tests/sanity.sh index ae63acb..ea9a8cc 100644 --- a/lustre/tests/sanity.sh +++ b/lustre/tests/sanity.sh @@ -1,6 +1,4 @@ #!/bin/bash -# -*- mode: Bash; tab-width: 4; indent-tabs-mode: t; -*- -# vim:autoindent:shiftwidth=4:tabstop=4: # # Run select tests by setting ONLY, or as arguments to the script. # Skip specific tests by setting EXCEPT. @@ -18,12 +16,6 @@ CPU=`awk '/model/ {print $4}' /proc/cpuinfo` # buffer i/o errs sock spc runas [ "$CPU" = "UML" ] && EXCEPT="$EXCEPT 27m 27n 27o 27p 27q 27r 31d 54a 64b 99a 99b 99c 99d 99e 99f 101a" -case `uname -r` in -2.4*) FSTYPE=${FSTYPE:-ext3} ;; -2.6*) FSTYPE=${FSTYPE:-ldiskfs} ;; -*) error "unsupported kernel" ;; -esac - SRCDIR=$(cd $(dirname $0); echo $PWD) export PATH=$PATH:/sbin @@ -38,6 +30,7 @@ LCTL=${LCTL:-lctl} MCREATE=${MCREATE:-mcreate} OPENFILE=${OPENFILE:-openfile} OPENUNLINK=${OPENUNLINK:-openunlink} +export MULTIOP=${MULTIOP:-multiop} READS=${READS:-"reads"} MUNLINK=${MUNLINK:-munlink} SOCKETSERVER=${SOCKETSERVER:-socketserver} @@ -495,6 +488,62 @@ test_17k() { #bug 22301 } run_test 17k "symlinks: rsync with xattrs enabled =========================" +# LU-1540 +test_17m() { + local short_sym="0123456789" + local WDIR=$DIR/${tdir}m + local mds_index + local devname + local cmd + local i + local rc=0 + + [ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.2.0) ] && + [ $(lustre_version_code $SINGLEMDS) -le $(version_code 2.2.93) ] && + skip "MDS 2.2.0-2.2.93 do not NUL-terminate symlinks" && return + mkdir -p $WDIR + long_sym=$short_sym + # create a long symlink file + for ((i = 0; i < 4; ++i)); do + long_sym=${long_sym}${long_sym} + done + + echo "create 512 short and long symlink files under $WDIR" + for ((i = 0; i < 256; ++i)); do + ln -sf ${long_sym}"a5a5" $WDIR/long-$i + ln -sf ${short_sym}"a5a5" $WDIR/short-$i + done + + echo "erase them" + rm -f $WDIR/* + sync + sleep 2 + + echo "recreate the 512 symlink files with a shorter string" + for ((i = 0; i < 512; ++i)); do + # rewrite the symlink file with a shorter string + ln -sf ${long_sym} $WDIR/long-$i + ln -sf ${short_sym} $WDIR/short-$i + done + + mds_index=$($LFS getstripe -M $WDIR) + mds_index=$((mds_index+1)) + devname=$(mdsdevname $mds_index) + cmd="$E2FSCK -fnvd $devname" + + echo "stop and checking mds${mds_index}: $cmd" + # e2fsck should not return error + stop mds${mds_index} -f + do_facet mds${mds_index} $cmd || rc=$? + + start mds${mds_index} $devname $MDS_MOUNT_OPTS + df $MOUNT > /dev/null 2>&1 + [ $rc -ne 0 ] && error "e2fsck should not report error upon "\ + "short/long symlink MDT: rc=$rc" + return $rc +} +run_test 17m "run e2fsck against MDT which contains short/long symlink" + test_18() { touch $DIR/f ls $DIR || error @@ -693,13 +742,13 @@ run_test 24k "touch .../R11a/f; mv .../R11a/f .../R11a/d =======" # bug 2429 - rename foo foo foo creates invalid file test_24l() { f="$DIR/f24l" - multiop $f OcNs || error + $MULTIOP $f OcNs || error } run_test 24l "Renaming a file to itself ========================" test_24m() { f="$DIR/f24m" - multiop $f OcLN ${f}2 ${f}2 || error "link ${f}2 ${f}2 failed" + $MULTIOP $f OcLN ${f}2 ${f}2 || error "link ${f}2 ${f}2 failed" # on ext3 this does not remove either the source or target files # though the "expected" operation would be to remove the source $CHECKSTAT -t file ${f} || error "${f} missing" @@ -776,7 +825,7 @@ test_24t() { run_test 24t "mkdir .../R16a/b/c; rename .../R16a/b/c .../R16a =" test_24u() { # bug12192 - multiop $DIR/$tfile C2w$((2048 * 1024))c || error + $MULTIOP $DIR/$tfile C2w$((2048 * 1024))c || error $CHECKSTAT -s $((2048 * 1024)) $DIR/$tfile || error "wrong file size" } run_test 24u "create stripe file" @@ -1358,9 +1407,6 @@ check_seq_oid() # compare lmm_object_id and lu_fid->oid [ $lmm_oid = ${fid[2]} ] || { error "OID mismatch"; return 2; } - [ "$FSTYPE" != "ldiskfs" ] && - skip "cannot check filter fid FSTYPE=$FSTYPE" && return 0 - # check the trusted.fid attribute of the OST objects of the file local have_obdidx=false local stripe_nr=0 @@ -1373,6 +1419,11 @@ check_seq_oid() local ost=$((obdidx + 1)) local dev=$(ostdevname $ost) + 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 @@ -1545,7 +1596,7 @@ run_test 31a "open-unlink file ==================================" test_31b() { touch $DIR/f31 || error ln $DIR/f31 $DIR/f31b || error - multiop $DIR/f31b Ouc || error + $MULTIOP $DIR/f31b Ouc || error $CHECKSTAT -t file $DIR/f31 || error } run_test 31b "unlink file with multiple links while open =======" @@ -1555,7 +1606,7 @@ test_31c() { ln $DIR/f31 $DIR/f31c || error multiop_bg_pause $DIR/f31 O_uc || return 1 MULTIPID=$! - multiop $DIR/f31c Ouc + $MULTIOP $DIR/f31c Ouc kill -USR1 $MULTIPID wait $MULTIPID } @@ -2066,7 +2117,7 @@ test_34h() { local sz=1000 dd if=/dev/zero of=$DIR/$tfile bs=1M count=10 || error - multiop $DIR/$tfile OG${gid}T${sz}g${gid}c & + $MULTIOP $DIR/$tfile OG${gid}T${sz}g${gid}c & MULTIPID=$! sleep 2 @@ -2427,10 +2478,14 @@ test_39i() { run_test 39i "write, rename, stat ==============================" test_39j() { + start_full_debug_logging touch $DIR1/$tfile sleep 1 - multiop_bg_pause $DIR1/$tfile oO_RDWR:w2097152_c || error "multiop failed" + #define OBD_FAIL_OSC_DELAY_SETTIME 0x412 + lctl set_param fail_loc=0x80000412 + multiop_bg_pause $DIR1/$tfile oO_RDWR:w2097152_c || + error "multiop failed" local multipid=$! local mtime1=`stat -c %Y $DIR1/$tfile` @@ -2441,12 +2496,15 @@ test_39j() { for (( i=0; i < 2; i++ )) ; do local mtime2=`stat -c %Y $DIR1/$tfile-1` - [ "$mtime1" = "$mtime2" ] || \ - error "mtime is lost on close: $mtime2, should be $mtime1" + [ "$mtime1" = "$mtime2" ] || + error "mtime is lost on close: $mtime2, " \ + "should be $mtime1" cancel_lru_locks osc if [ $i = 0 ] ; then echo "repeat after cancel_lru_locks"; fi done + lctl set_param fail_loc=0 + stop_full_debug_logging } run_test 39j "write, rename, close, stat =======================" @@ -2478,7 +2536,23 @@ run_test 39k "write, utime, close, stat ========================" # this should be set to future TEST_39_ATIME=`date -d "1 year" +%s` +is_sles11() # LU-1783 +{ + if [ -r /etc/SuSE-release ] + then + local vers=`grep VERSION /etc/SuSE-release | awk '{print $3}'` + local patchlev=`grep PATCHLEVEL /etc/SuSE-release \ + | awk '{print $3}'` + if [ $vers -eq 11 ] && [ $patchlev -eq 1 ] + then + return 0 + fi + fi + return 1 +} + test_39l() { + is_sles11 && skip "SLES 11 SP1" && return # LU-1783 remote_mds_nodsh && skip "remote MDS with nodsh" && return local atime_diff=$(do_facet $SINGLEMDS lctl get_param -n mdd.*.atime_diff) @@ -2781,7 +2855,7 @@ run_test 42e "verify sub-RPC writes are not done synchronously" test_43() { mkdir -p $DIR/$tdir cp -p /bin/ls $DIR/$tdir/$tfile - multiop $DIR/$tdir/$tfile Ow_c & + $MULTIOP $DIR/$tdir/$tfile Ow_c & pid=$! # give multiop a chance to open sleep 1 @@ -2793,10 +2867,10 @@ run_test 43 "execution of file opened for write should return -ETXTBSY" test_43a() { mkdir -p $DIR/d43 - cp -p `which multiop` $DIR/d43/multiop || cp -p multiop $DIR/d43/multiop + cp -p `which $MULTIOP` $DIR/d43/multiop || cp -p multiop $DIR/d43/multiop MULTIOP_PROG=$DIR/d43/multiop multiop_bg_pause $TMP/test43.junk O_c || return 1 MULTIOP_PID=$! - multiop $DIR/d43/multiop Oc && error "expected error, got success" + $MULTIOP $DIR/d43/multiop Oc && error "expected error, got success" kill -USR1 $MULTIOP_PID || return 2 wait $MULTIOP_PID || return 3 rm $TMP/test43.junk @@ -2805,7 +2879,7 @@ run_test 43a "open(RDWR) of file being executed should return -ETXTBSY" test_43b() { mkdir -p $DIR/d43 - cp -p `which multiop` $DIR/d43/multiop || cp -p multiop $DIR/d43/multiop + cp -p `which $MULTIOP` $DIR/d43/multiop || cp -p multiop $DIR/d43/multiop MULTIOP_PROG=$DIR/d43/multiop multiop_bg_pause $TMP/test43.junk O_c || return 1 MULTIOP_PID=$! $TRUNCATE $DIR/d43/multiop 0 && error "expected error, got success" @@ -3030,6 +3104,30 @@ test_48e() { # bug 4134 } run_test 48e "Access to recreated parent subdir (should return errors)" +test_49() { # LU-1030 + # get ost1 size - lustre-OST0000 + ost1_size=$(do_facet ost1 lfs df |grep ${ost1_svc} |awk '{print $4}') + # write 800M at maximum + [ $ost1_size -gt 819200 ] && ost1_size=819200 + + lfs setstripe -c 1 -i 0 $DIR/$tfile + dd if=/dev/zero of=$DIR/$tfile bs=4k count=$((ost1_size >> 2)) & + local dd_pid=$! + + # change max_pages_per_rpc while writing the file + local osc1_mppc=osc.$(get_osc_import_name client ost1).max_pages_per_rpc + local orig_mppc=`$LCTL get_param -n $osc1_mppc` + # loop until dd process exits + while ps ax -opid | grep -wq $dd_pid; do + $LCTL set_param $osc1_mppc=$((RANDOM % 256 + 1)) + sleep $((RANDOM % 5 + 1)) + done + # restore original max_pages_per_rpc + $LCTL set_param $osc1_mppc=$orig_mppc + rm $DIR/$tfile || error "rm $DIR/$tfile failed" +} +run_test 49 "Change max_pages_per_rpc won't break osc extent" + test_50() { # bug 1485 mkdir $DIR/d50 @@ -3071,28 +3169,30 @@ test_51b() { run_test 51b "mkdir .../t-0 --- .../t-$NUMTEST ====================" test_51ba() { # LU-993 - local BASE=$DIR/d51b - # unlink all but 100 subdirectories, then check it still works - local LEFT=100 - local DELETE=$((NUMTEST - LEFT)) - - ! [ -d "${BASE}/t-$DELETE" ] && skip "test_51b() not run" && return 0 - - # for ldiskfs the nlink count should be 1, but this is OSD specific - # and so this is listed for informational purposes only - log "nlink before: $(stat -c %h $BASE)" - unlinkmany -d $BASE/t- $DELETE || - error "unlink of first $DELETE subdirs failed" - - log "nlink between: $(stat -c %h $BASE)" - local FOUND=$(ls -l ${BASE} | wc -l) - FOUND=$((FOUND - 1)) # trim the first line of ls output - [ $FOUND -ne $LEFT ] && - error "can't find subdirs: found only $FOUND/$LEFT" - - unlinkmany -d $BASE/t- $DELETE $LEFT || - error "unlink of second $LEFT subdirs failed" - log "nlink after: $(stat -c %h $BASE)" + local BASE=$DIR/d51b + # unlink all but 100 subdirectories, then check it still works + local LEFT=100 + local DELETE=$((NUMTEST - LEFT)) + + # continue on to run this test even if 51b didn't finish, + # just to delete the many subdirectories created. + ! [ -d "${BASE}/t-1" ] && skip "test_51b() not run" && return 0 + + # for ldiskfs the nlink count should be 1, but this is OSD specific + # and so this is listed for informational purposes only + log "nlink before: $(stat -c %h $BASE)" + unlinkmany -d $BASE/t- $DELETE || + error "unlink of first $DELETE subdirs failed" + + log "nlink between: $(stat -c %h $BASE)" + local FOUND=$(ls -l ${BASE} | wc -l) + FOUND=$((FOUND - 1)) # trim the first line of ls output + [ $FOUND -ne $LEFT ] && + error "can't find subdirs: found only $FOUND/$LEFT" + + unlinkmany -d $BASE/t- $DELETE $LEFT || + error "unlink of second $LEFT subdirs failed" + log "nlink after: $(stat -c %h $BASE)" } run_test 51ba "rmdir .../t-0 --- .../t-$NUMTEST" @@ -4020,7 +4120,7 @@ test_61() { f="$DIR/f61" dd if=/dev/zero of=$f bs=`page_size` count=1 cancel_lru_locks osc - multiop $f OSMWUc || error + $MULTIOP $f OSMWUc || error sync } run_test 61 "mmap() writes don't make sync hang ================" @@ -4066,7 +4166,7 @@ test_63b() { #define OBD_FAIL_OSC_BRW_PREP_REQ 0x406 lctl set_param fail_loc=0x80000406 - multiop $DIR/$tfile Owy && \ + $MULTIOP $DIR/$tfile Owy && \ error "sync didn't return ENOMEM" sync; sleep 2; sync # do a real sync this time to flush page lctl get_param -n llite.*.dump_page_cache | grep locked && \ @@ -4180,7 +4280,7 @@ test_65ib() { # bug12836 run_test 65ib "getstripe -v on -1 default directory striping" test_65ic() { # bug12836 - $LFS find -mtime -1 $MOUNT || error "find $MOUNT failed" + $LFS find -mtime -1 $MOUNT > /dev/null || error "find $MOUNT failed" } run_test 65ic "new find on -1 default directory striping" @@ -4241,7 +4341,8 @@ run_test 65l "lfs find on -1 stripe dir ========================" test_66() { COUNT=${COUNT:-8} dd if=/dev/zero of=$DIR/f66 bs=1k count=$COUNT - sync; sleep 1; sync + sync; sync_all_data; sync; sync_all_data + cancel_lru_locks osc BLOCKS=`ls -s $DIR/f66 | awk '{ print $1 }'` [ $BLOCKS -ge $COUNT ] || error "$DIR/f66 blocks $BLOCKS < $COUNT" } @@ -4436,11 +4537,11 @@ test_73() { pid1=$! lctl set_param fail_loc=0x80000129 - multiop $DIR/d73-1/f73-2 Oc & + $MULTIOP $DIR/d73-1/f73-2 Oc & sleep 1 lctl set_param fail_loc=0 - multiop $DIR/d73-2/f73-3 Oc & + $MULTIOP $DIR/d73-2/f73-3 Oc & pid3=$! kill -USR1 $pid1 @@ -4695,7 +4796,7 @@ test_77i() { # bug 13805 for VALUE in `lctl get_param osc.*osc-[^mM]*.checksum_type`; do PARAM=`echo ${VALUE[0]} | cut -d "=" -f1` algo=`lctl get_param -n $PARAM | sed 's/.*\[\(.*\)\].*/\1/g'` - [ "$algo" = "crc32" ] || error "algo set to $algo instead of crc32" + [ "$algo" = "adler" ] || error "algo set to $algo instead of adler" done remount_client $MOUNT } @@ -4816,7 +4917,7 @@ test_81a() { # LU-456 # write should trigger a retry and success $SETSTRIPE -i 0 -c 1 $DIR/$tfile - multiop $DIR/$tfile oO_CREAT:O_RDWR:O_SYNC:w4096c + $MULTIOP $DIR/$tfile oO_CREAT:O_RDWR:O_SYNC:w4096c RC=$? if [ $RC -ne 0 ] ; then error "write should success, but failed for $RC" @@ -4832,7 +4933,7 @@ test_81b() { # LU-456 # write should retry several times and return -ENOSPC finally $SETSTRIPE -i 0 -c 1 $DIR/$tfile - multiop $DIR/$tfile oO_CREAT:O_RDWR:O_SYNC:w4096c + $MULTIOP $DIR/$tfile oO_CREAT:O_RDWR:O_SYNC:w4096c RC=$? ENOSPC=28 if [ $RC -ne $ENOSPC ] ; then @@ -4841,6 +4942,27 @@ test_81b() { # LU-456 } run_test 81b "OST should return -ENOSPC when retry still fails =======" +test_82() { # LU-1031 + dd if=/dev/zero of=$DIR/$tfile bs=1M count=10 + local gid1=14091995 + local gid2=16022000 + + multiop_bg_pause $DIR/$tfile OG${gid1}_g${gid1}c || return 1 + local MULTIPID1=$! + multiop_bg_pause $DIR/$tfile O_G${gid2}r10g${gid2}c || return 2 + local MULTIPID2=$! + kill -USR1 $MULTIPID2 + sleep 2 + if [[ `ps h -o comm -p $MULTIPID2` == "" ]]; then + error "First grouplock does not block second one" + else + echo "Second grouplock blocks first one" + fi + kill -USR1 $MULTIPID1 + wait $MULTIPID1 + wait $MULTIPID2 +} +run_test 82 "Basic grouplock test ===============================" test_99a() { [ -z "$(which cvs 2>/dev/null)" ] && skip_env "could not find cvs" && \ @@ -5752,9 +5874,9 @@ test_115() { cut -c11-20) # don't return an error - [ $OSTIO_post -eq $OSTIO_pre ] && echo \ - "WARNING: No new ll_ost_io threads were created ($OSTIO_pre)" &&\ - echo "This may be fine, depending on what ran before this test" &&\ + [ $OSTIO_post == $OSTIO_pre ] && echo \ + "WARNING: No new ll_ost_io threads were created ($OSTIO_pre)" && + echo "This may be fine, depending on what ran before this test" && echo "and how fast this system is." && return echo "Started with $OSTIO_pre threads, ended with $OSTIO_post" @@ -5902,7 +6024,7 @@ test_118a() #bug 11710 { reset_async - multiop $DIR/$tfile oO_CREAT:O_RDWR:O_SYNC:w4096c + $MULTIOP $DIR/$tfile oO_CREAT:O_RDWR:O_SYNC:w4096c DIRTY=$(lctl get_param -n llite.*.dump_page_cache | grep -c dirty) WRITEBACK=$(lctl get_param -n llite.*.dump_page_cache | grep -c writeback) @@ -5922,7 +6044,7 @@ test_118b() #define OBD_FAIL_OST_ENOENT 0x217 set_nodes_failloc "$(osts_nodes)" 0x217 - multiop $DIR/$tfile oO_CREAT:O_RDWR:O_SYNC:w4096c + $MULTIOP $DIR/$tfile oO_CREAT:O_RDWR:O_SYNC:w4096c RC=$? set_nodes_failloc "$(osts_nodes)" 0 DIRTY=$(lctl get_param -n llite.*.dump_page_cache | grep -c dirty) @@ -5943,7 +6065,7 @@ test_118b() # Due to the above error the OSC will issue all RPCs syncronously # until a subsequent RPC completes successfully without error. - multiop $DIR/$tfile Ow4096yc + $MULTIOP $DIR/$tfile Ow4096yc rm -f $DIR/$tfile return 0 @@ -5960,7 +6082,7 @@ test_118c() set_nodes_failloc "$(osts_nodes)" 0x216 # multiop should block due to fsync until pages are written - multiop $DIR/$tfile oO_CREAT:O_RDWR:O_SYNC:w4096c & + $MULTIOP $DIR/$tfile oO_CREAT:O_RDWR:O_SYNC:w4096c & MULTIPID=$! sleep 1 @@ -6003,7 +6125,7 @@ test_118d() #define OBD_FAIL_OST_BRW_PAUSE_BULK set_nodes_failloc "$(osts_nodes)" 0x214 # multiop should block due to fsync until pages are written - multiop $DIR/$tfile oO_CREAT:O_RDWR:O_SYNC:w4096c & + $MULTIOP $DIR/$tfile oO_CREAT:O_RDWR:O_SYNC:w4096c & MULTIPID=$! sleep 1 @@ -6040,7 +6162,7 @@ test_118f() { lctl set_param fail_loc=0x8000040a # Should simulate EINVAL error which is fatal - multiop $DIR/$tfile oO_CREAT:O_RDWR:O_SYNC:w4096c + $MULTIOP $DIR/$tfile oO_CREAT:O_RDWR:O_SYNC:w4096c RC=$? if [[ $RC -eq 0 ]]; then error "Must return error due to dropped pages, rc=$RC" @@ -6075,7 +6197,7 @@ test_118g() { lctl set_param fail_loc=0x406 # simulate local -ENOMEM - multiop $DIR/$tfile oO_CREAT:O_RDWR:O_SYNC:w4096c + $MULTIOP $DIR/$tfile oO_CREAT:O_RDWR:O_SYNC:w4096c RC=$? lctl set_param fail_loc=0 @@ -6111,7 +6233,7 @@ test_118h() { #define OBD_FAIL_OST_BRW_WRITE_BULK 0x20e set_nodes_failloc "$(osts_nodes)" 0x20e # Should simulate ENOMEM error which is recoverable and should be handled by timeout - multiop $DIR/$tfile oO_CREAT:O_RDWR:O_SYNC:w4096c + $MULTIOP $DIR/$tfile oO_CREAT:O_RDWR:O_SYNC:w4096c RC=$? set_nodes_failloc "$(osts_nodes)" 0 @@ -6147,7 +6269,7 @@ test_118i() { set_nodes_failloc "$(osts_nodes)" 0x20e # Should simulate ENOMEM error which is recoverable and should be handled by timeout - multiop $DIR/$tfile oO_CREAT:O_RDWR:O_SYNC:w4096c & + $MULTIOP $DIR/$tfile oO_CREAT:O_RDWR:O_SYNC:w4096c & PID=$! sleep 5 set_nodes_failloc "$(osts_nodes)" 0 @@ -6185,7 +6307,7 @@ test_118j() { set_nodes_failloc "$(osts_nodes)" 0x220 # return -EIO from OST - multiop $DIR/$tfile oO_CREAT:O_RDWR:O_SYNC:w4096c + $MULTIOP $DIR/$tfile oO_CREAT:O_RDWR:O_SYNC:w4096c RC=$? set_nodes_failloc "$(osts_nodes)" 0x0 if [[ $RC -eq 0 ]]; then @@ -6237,7 +6359,7 @@ test_118l() { # LU-646 mkdir -p $DIR/$tdir - multiop $DIR/$tdir Dy || error "fsync dir failed" + $MULTIOP $DIR/$tdir Dy || error "fsync dir failed" rm -rf $DIR/$tdir } run_test 118l "fsync dir =========" @@ -6267,7 +6389,7 @@ test_119b() # bug 11737 $SETSTRIPE -c 2 $DIR/$tfile || error "setstripe failed" dd if=/dev/zero of=$DIR/$tfile bs=1M count=1 seek=1 || error "dd failed" sync - multiop $DIR/$tfile oO_RDONLY:O_DIRECT:r$((2048 * 1024)) || \ + $MULTIOP $DIR/$tfile oO_RDONLY:O_DIRECT:r$((2048 * 1024)) || \ error "direct read failed" rm -f $DIR/$tfile } @@ -6892,7 +7014,10 @@ set_dir_limits () { done } test_129() { - [ "$FSTYPE" != "ldiskfs" ] && skip "not needed for FSTYPE=$FSTYPE" && return 0 + if [ "$(facet_type_fstype MDS)" != ldiskfs ]; then + skip "Only applicable to ldiskfs-based MDTs" + return + fi remote_mds_nodsh && skip "remote MDS with nodsh" && return EFBIG=27 @@ -6905,7 +7030,7 @@ test_129() { I=0 J=0 while [ ! $I -gt $((MAX * MDSCOUNT)) ]; do - multiop $DIR/$tdir/$J Oc + $MULTIOP $DIR/$tdir/$J Oc rc=$? if [ $rc -eq $EFBIG ]; then set_dir_limits 0 @@ -6931,18 +7056,26 @@ cleanup_130() { } test_130a() { - filefrag_op=$(filefrag -e 2>&1 | grep "invalid option") - [ -n "$filefrag_op" ] && skip "filefrag does not support FIEMAP" && return + local filefrag_op=$(filefrag -e 2>&1 | grep "invalid option") + [ -n "$filefrag_op" ] && skip_env "filefrag does not support FIEMAP" && + return trap cleanup_130 EXIT RETURN local fm_file=$DIR/$tfile $SETSTRIPE -S 65536 -c 1 $fm_file || error "setstripe on $fm_file" - dd if=/dev/zero of=$fm_file bs=65536 count=1 || error "dd failed for $fm_file" + dd if=/dev/zero of=$fm_file bs=65536 count=1 || + error "dd failed for $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"` + # LU-1795: test filefrag/FIEMAP once, even if unsupported + filefrag -ves $fm_file + RC=$? + [ "$(facet_fstype ost$(($($GETSTRIPE -i $fm_file) + 1)))" = "zfs" ] && + skip "ORI-366/LU-1941: FIEMAP unimplemented on ZFS" && return + [ $RC != 0 ] && error "filefrag $fm_file failed" + filefrag_op=$(filefrag -ve $fm_file | grep -A 100 "ext:" | + grep -v "ext:" | grep -v "found") lun=$($GETSTRIPE -i $fm_file) start_blk=`echo $filefrag_op | cut -d: -f2 | cut -d. -f1` @@ -6973,21 +7106,28 @@ test_130a() { run_test 130a "FIEMAP (1-stripe file)" test_130b() { - [ "$OSTCOUNT" -lt "2" ] && skip_env "skipping FIEMAP on 2-stripe file test" && return + [ "$OSTCOUNT" -lt "2" ] && + skip_env "skipping FIEMAP on 2-stripe file test" && return - filefrag_op=$(filefrag -e 2>&1 | grep "invalid option") - [ -n "$filefrag_op" ] && skip "filefrag does not support FIEMAP" && return + local filefrag_op=$(filefrag -e 2>&1 | grep "invalid option") + [ -n "$filefrag_op" ] && skip_env "filefrag does not support FIEMAP" && + return trap cleanup_130 EXIT RETURN local fm_file=$DIR/$tfile $SETSTRIPE -S 65536 -c 2 $fm_file || error "setstripe on $fm_file" - dd if=/dev/zero of=$fm_file bs=1M count=2 || error "dd failed 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=2 || + 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 $filefrag_op | cut -d: -f5` + last_lun=$(echo $filefrag_op | cut -d: -f5) IFS=$'\n' tot_len=0 @@ -7022,15 +7162,20 @@ test_130b() { run_test 130b "FIEMAP (2-stripe file)" test_130c() { - [ "$OSTCOUNT" -lt "2" ] && skip_env "skipping FIEMAP on 2-stripe file with hole test" && return + [ "$OSTCOUNT" -lt "2" ] && + skip_env "skipping FIEMAP on 2-stripe file" && return filefrag_op=$(filefrag -e 2>&1 | grep "invalid option") - [ -n "$filefrag_op" ] && skip "filefrag does not support FIEMAP" && return + [ -n "$filefrag_op" ] && skip "filefrag does not support FIEMAP" && + return trap cleanup_130 EXIT RETURN local fm_file=$DIR/$tfile $SETSTRIPE -S 65536 -c 2 $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 seek=1 bs=1M count=1 || error "dd failed on $fm_file" filefrag -ves $fm_file || error "filefrag $fm_file failed" @@ -7086,6 +7231,8 @@ test_130d() { local fm_file=$DIR/$tfile $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" filefrag -ves $fm_file || error "filefrag $fm_file failed" @@ -7135,6 +7282,9 @@ test_130e() { local fm_file=$DIR/$tfile $SETSTRIPE -S 131072 -c 2 $fm_file || error "setstripe on $fm_file" + [ "$(facet_fstype ost$(($($GETSTRIPE -i $fm_file) + 1)))" = "zfs" ] && + skip "ORI-366/LU-1941: FIEMAP unimplemented on ZFS" && return + NUM_BLKS=512 EXPECTED_LEN=$(( (NUM_BLKS / 2) * 64 )) for ((i = 0; i < $NUM_BLKS; i++)) @@ -7374,8 +7524,11 @@ test_133b() { chmod 444 ${testdir}/${tfile} || error "chmod failed" check_stats $SINGLEMDS "setattr" 1 do_facet $SINGLEMDS $LCTL set_param mdt.*.md_stats=clear - ls -l ${testdir}/${tfile} > /dev/null|| error "ls failed" - check_stats $SINGLEMDS "getattr" 1 + if [ $(lustre_version_code $SINGLEMDS) -ne $(version_code 2.2.0) ] + then # LU-1740 + ls -l ${testdir}/${tfile} > /dev/null|| error "ls failed" + check_stats $SINGLEMDS "getattr" 1 + fi $LFS df || error "lfs failed" check_stats $SINGLEMDS "statfs" 1 @@ -7478,22 +7631,24 @@ test_133d() { mkdir -p ${testdir2} || error "mkdir failed" createmany -o $testdir1/test 512 || error "createmany failed" - local testdir1_size=$(ls -l $DIR/${tdir} | - awk '/stats_testdir1/ {print $5}') - local testdir2_size=$(ls -l $DIR/${tdir} | - awk '/stats_testdir2/ {print $5}') - testdir1_size=$(order_2 $testdir1_size) - testdir2_size=$(order_2 $testdir2_size) + # check samedir rename size + mv ${testdir1}/test0 ${testdir1}/test_0 + + local testdir1_size=$(ls -l $DIR/${tdir} | + awk '/stats_testdir1/ {print $5}') + local testdir2_size=$(ls -l $DIR/${tdir} | + awk '/stats_testdir2/ {print $5}') - testdir1_size=$(size_in_KMGT $testdir1_size) - testdir2_size=$(size_in_KMGT $testdir2_size) + testdir1_size=$(order_2 $testdir1_size) + testdir2_size=$(order_2 $testdir2_size) - echo "source rename dir size: ${testdir1_size}" - echo "target rename dir size: ${testdir2_size}" + testdir1_size=$(size_in_KMGT $testdir1_size) + testdir2_size=$(size_in_KMGT $testdir2_size) + + echo "source rename dir size: ${testdir1_size}" + echo "target rename dir size: ${testdir2_size}" - # check samedir rename size - mv ${testdir1}/test0 ${testdir1}/test_0 local cmd="do_facet $SINGLEMDS $LCTL get_param mdt.*.rename_stats" eval $cmd || error "$cmd failed" local samedir=$($cmd | grep 'same_dir') @@ -7506,6 +7661,21 @@ test_133d() { # check crossdir rename size mv ${testdir1}/test_0 ${testdir2}/test_0 + + testdir1_size=$(ls -l $DIR/${tdir} | + awk '/stats_testdir1/ {print $5}') + testdir2_size=$(ls -l $DIR/${tdir} | + awk '/stats_testdir2/ {print $5}') + + testdir1_size=$(order_2 $testdir1_size) + testdir2_size=$(order_2 $testdir2_size) + + testdir1_size=$(size_in_KMGT $testdir1_size) + testdir2_size=$(size_in_KMGT $testdir2_size) + + echo "source rename dir size: ${testdir1_size}" + echo "target rename dir size: ${testdir2_size}" + eval $cmd || error "$cmd failed" local crossdir=$($cmd | grep 'crossdir') local src_sample=$(get_rename_size $testdir1_size) @@ -7580,44 +7750,43 @@ test_150() { run_test 150 "truncate/append tests" function roc_hit() { - local list=$(comma_list $(osts_nodes)) + local list=$(comma_list $(osts_nodes)) - ACCNUM=$(do_nodes $list $LCTL get_param -n obdfilter.*.stats | \ - awk '/'cache_hit'/ {sum+=$2} END {print sum}') - echo $ACCNUM + echo $(get_obdfilter_param $list '' stats | + awk '/'cache_hit'/ {sum+=$2} END {print sum}') } function set_cache() { - local on=1 + local on=1 - if [ "$2" == "off" ]; then - on=0; - fi - local list=$(comma_list $(osts_nodes)) - do_nodes $list lctl set_param obdfilter.*.${1}_cache_enable $on + if [ "$2" == "off" ]; then + on=0; + fi + local list=$(comma_list $(osts_nodes)) + set_obdfilter_param $list '' $1_cache_enable $on - cancel_lru_locks osc + cancel_lru_locks osc } test_151() { - remote_ost_nodsh && skip "remote OST with nodsh" && return + remote_ost_nodsh && skip "remote OST with nodsh" && return - local CPAGES=3 - local list=$(comma_list $(osts_nodes)) + local CPAGES=3 + local list=$(comma_list $(osts_nodes)) - # check whether obdfilter is cache capable at all - if ! do_nodes $list $LCTL get_param -n obdfilter.*.read_cache_enable > /dev/null; then - echo "not cache-capable obdfilter" - return 0 - fi + # check whether obdfilter is cache capable at all + if ! get_obdfilter_param $list '' read_cache_enable >/dev/null; then + echo "not cache-capable obdfilter" + return 0 + fi - # check cache is enabled on all obdfilters - if do_nodes $list $LCTL get_param -n obdfilter.*.read_cache_enable | grep 0 >&/dev/null; then - echo "oss cache is disabled" - return 0 - fi + # check cache is enabled on all obdfilters + if get_obdfilter_param $list '' read_cache_enable | grep 0; then + echo "oss cache is disabled" + return 0 + fi - do_nodes $list $LCTL set_param -n obdfilter.*.writethrough_cache_enable 1 + set_obdfilter_param $list '' writethrough_cache_enable 1 # pages should be in the case right after write dd if=/dev/urandom of=$DIR/$tfile bs=4k count=$CPAGES || error "dd failed" @@ -7631,7 +7800,7 @@ test_151() { # the following read invalidates the cache cancel_lru_locks osc - do_nodes $list $LCTL set_param -n obdfilter.*.read_cache_enable 0 + set_obdfilter_param $list '' read_cache_enable 0 cat $DIR/$tfile >/dev/null # now data shouldn't be found in the cache @@ -7643,7 +7812,7 @@ test_151() { error "IN CACHE: before: $BEFORE, after: $AFTER" fi - do_nodes $list $LCTL set_param -n obdfilter.*.read_cache_enable 1 + set_obdfilter_param $list '' read_cache_enable 1 rm -f $DIR/$tfile } run_test 151 "test cache on oss and controls ===============================" @@ -7672,11 +7841,14 @@ test_152() { run_test 152 "test read/write with enomem ============================" test_153() { - multiop $DIR/$tfile Ow4096Ycu || error "multiop failed" + $MULTIOP $DIR/$tfile Ow4096Ycu || error "multiop failed" } run_test 153 "test if fdatasync does not crash =======================" test_154() { + [[ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.2.51) ]] || + { skip "Need MDS version at least 2.2.51"; return 0; } + cp /etc/hosts $DIR/$tfile fid=$($LFS path2fid $DIR/$tfile) @@ -7742,6 +7914,46 @@ test_154() { touch $DIR/.lustre/fid/$tfile && \ error "touch $DIR/.lustre/fid/$tfile should fail." + echo "setxattr to $DIR/.lustre/fid" + setfattr -n trusted.name1 -v value1 $DIR/.lustre/fid && + error "setxattr should fail." + + echo "listxattr for $DIR/.lustre/fid" + getfattr -d -m "^trusted" $DIR/.lustre/fid && + error "listxattr should fail." + + echo "delxattr from $DIR/.lustre/fid" + setfattr -x trusted.name1 $DIR/.lustre/fid && + error "delxattr should fail." + + echo "touch invalid fid: $DIR/.lustre/fid/[0x200000400:0x2:0x3]" + touch $DIR/.lustre/fid/[0x200000400:0x2:0x3] && + error "touch invalid fid should fail." + + echo "touch non-normal fid: $DIR/.lustre/fid/[0x1:0x2:0x0]" + touch $DIR/.lustre/fid/[0x1:0x2:0x0] && + error "touch non-normal fid should fail." + + echo "rename $tdir to $DIR/.lustre/fid" + mrename $DIR/$tdir $DIR/.lustre/fid && + error "rename to $DIR/.lustre/fid should fail." + + echo "rename .lustre to itself" + fid=$($LFS path2fid $DIR) + mrename $DIR/.lustre $DIR/.lustre/fid/$fid/.lustre && + error "rename .lustre to itself should fail." + + $OPENFILE -f O_LOV_DELAY_CREATE:O_CREAT $DIR/$tfile-2 + fid=$($LFS path2fid $DIR/$tfile-2) + echo "cp /etc/passwd $DIR/.lustre/fid/$fid" + cp /etc/passwd $DIR/.lustre/fid/$fid && + error "create lov data thru .lustre should fail." + echo "cp /etc/passwd $DIR/$tfile-2" + cp /etc/passwd $DIR/$tfile-2 || error "copy to $DIR/$tfile-2 failed." + echo "diff /etc/passwd $DIR/.lustre/fid/$fid" + diff /etc/passwd $DIR/.lustre/fid/$fid || + error "diff /etc/passwd $DIR/.lustre/fid/$fid failed." + echo "Open-by-FID succeeded" } run_test 154 "Open-by-FID" @@ -8225,11 +8437,11 @@ run_test 163 "kernel <-> userspace comms" test_169() { # do directio so as not to populate the page cache log "creating a 10 Mb file" - multiop $DIR/$tfile oO_CREAT:O_DIRECT:O_RDWR:w$((10*1048576))c || error "multiop failed while creating a file" + $MULTIOP $DIR/$tfile oO_CREAT:O_DIRECT:O_RDWR:w$((10*1048576))c || error "multiop failed while creating a file" log "starting reads" dd if=$DIR/$tfile of=/dev/null bs=4096 & log "truncating the file" - multiop $DIR/$tfile oO_TRUNC:c || error "multiop failed while truncating the file" + $MULTIOP $DIR/$tfile oO_TRUNC:c || error "multiop failed while truncating the file" log "killing dd" kill %+ || true # reads might have finished echo "wait until dd is finished" @@ -8415,190 +8627,326 @@ test_181() { # bug 22177 } run_test 181 "Test open-unlinked dir ========================" -# OST pools tests -POOL=${POOL:-cea1} -TGT_COUNT=$OSTCOUNT -TGTPOOL_FIRST=1 -TGTPOOL_MAX=$(($TGT_COUNT - 1)) -TGTPOOL_STEP=2 -TGTPOOL_LIST=`seq $TGTPOOL_FIRST $TGTPOOL_STEP $TGTPOOL_MAX` -POOL_ROOT=${POOL_ROOT:-$DIR/d200.pools} -POOL_DIR_NAME=dir_tst -POOL_DIR=$POOL_ROOT/$POOL_DIR_NAME -POOL_FILE=$POOL_ROOT/file_tst +test_182() { + # disable MDC RPC lock wouldn't crash client + local fcount=1000 + local tcount=4 + mkdir -p $DIR/$tdir || error "creating dir $DIR/$tdir" +#define OBD_FAIL_MDC_RPCS_SEM 0x804 + $LCTL set_param fail_loc=0x804 + + for (( i=0; i < $tcount; i++ )) ; do + mkdir $DIR/$tdir/$i + createmany -o $DIR/$tdir/$i/f- $fcount & + done + wait + + for (( i=0; i < $tcount; i++ )) ; do + unlinkmany $DIR/$tdir/$i/f- $fcount & + done + wait + + rm -rf $DIR/$tdir + + $LCTL set_param fail_loc=0 +} +run_test 182 "Disable MDC RPCs semaphore wouldn't crash client ================" + +# OST pools tests check_file_in_pool() { - file=$1 - res=$($GETSTRIPE $file | grep 0x | cut -f2) + local file=$1 + local pool=$2 + local tlist="$3" + local res=$($GETSTRIPE $file | grep 0x | cut -f2) for i in $res do - found=$(echo :$TGTPOOL_LIST: | tr " " ":" | grep :$i:) - if [[ "$found" == "" ]] - then - echo "pool list: $TGTPOOL_LIST" - echo "striping: $res" - error "$file not allocated in $POOL" - return 1 - fi + for t in $tlist ; do + [ "$i" -eq "$t" ] && continue 2 + done + + echo "pool list: $tlist" + echo "striping: $res" + error_noexit "$file not allocated in $pool" + return 1 done return 0 } -trap "cleanup_pools $FSNAME" EXIT +pool_add() { + echo "Creating new pool" + local pool=$1 -test_200a() { - remote_mgs_nodsh && skip "remote MGS with nodsh" && return - create_pool $FSNAME.$POOL || return $? - [ $($LFS pool_list $FSNAME | grep -c $POOL) -eq 1 ] || - error "$POOL not in lfs pool_list" + create_pool $FSNAME.$pool || + { error_noexit "No pool created, result code $?"; return 1; } + [ $($LFS pool_list $FSNAME | grep -c $pool) -eq 1 ] || + { error_noexit "$pool not in lfs pool_list"; return 2; } } -run_test 200a "Create new pool ==========================================" -test_200b() { - remote_mgs_nodsh && skip "remote MGS with nodsh" && return - TGT=$(for i in $TGTPOOL_LIST; do printf "$FSNAME-OST%04x_UUID " $i; done) - do_facet mgs $LCTL pool_add $FSNAME.$POOL \ - $FSNAME-OST[$TGTPOOL_FIRST-$TGTPOOL_MAX/$TGTPOOL_STEP] - wait_update $HOSTNAME "lctl get_param -n lov.$FSNAME-*.pools.$POOL | sort -u | tr '\n' ' ' " "$TGT" || - error "Add to pool failed" - local lfscount=$($LFS pool_list $FSNAME.$POOL | grep -c "\-OST") - local addcount=$((($TGTPOOL_MAX - $TGTPOOL_FIRST) / $TGTPOOL_STEP + 1)) - [ $lfscount -eq $addcount ] || - error "lfs pool_list bad ost count $lfscount != $addcount" -} -run_test 200b "Add targets to a pool ====================================" - -test_200c() { - remote_mgs_nodsh && skip "remote MGS with nodsh" && return - mkdir -p $POOL_DIR - $SETSTRIPE -c 2 -p $POOL $POOL_DIR - [ $? = 0 ] || error "Cannot set pool $POOL to $POOL_DIR" - # b-19919 test relative path works well - mkdir -p $POOL_DIR/$POOL_DIR_NAME - cd $POOL_DIR - $SETSTRIPE -c 2 -p $POOL $POOL_DIR_NAME - [ $? = 0 ] || error "Cannot set pool $POOL to $POOL_DIR/$POOL_DIR_NAME" - $SETSTRIPE -c 2 -p $POOL ./$POOL_DIR_NAME - [ $? = 0 ] || error "Cannot set pool $POOL to $POOL_DIR/./$POOL_DIR_NAME" - $SETSTRIPE -c 2 -p $POOL ../$POOL_DIR_NAME - [ $? = 0 ] || error "Cannot set pool $POOL to $POOL_DIR/../$POOL_DIR_NAME" - $SETSTRIPE -c 2 -p $POOL ../$POOL_DIR_NAME/$POOL_DIR_NAME - [ $? = 0 ] || error "Cannot set pool $POOL to $POOL_DIR/../$POOL_DIR_NAME/$POOL_DIR_NAME" - rm -rf $POOL_DIR_NAME; cd - -} -run_test 200c "Set pool on a directory =================================" - -test_200d() { - remote_mgs_nodsh && skip "remote MGS with nodsh" && return - res=$($GETSTRIPE --pool $POOL_DIR) - [ $res = $POOL ] || error "Pool on $POOL_DIR is $res, not $POOL" +pool_add_targets() { + echo "Adding targets to pool" + local pool=$1 + local first=$2 + local last=$3 + local step=${4:-1} + + local list=$(seq $first $step $last) + + local t=$(for i in $list; do printf "$FSNAME-OST%04x_UUID " $i; done) + do_facet mgs $LCTL pool_add \ + $FSNAME.$pool $FSNAME-OST[$first-$last/$step] + wait_update $HOSTNAME "lctl get_param -n lov.$FSNAME-*.pools.$pool \ + | sort -u | tr '\n' ' ' " "$t" || { + error_noexit "Add to pool failed" + return 1 + } + local lfscount=$($LFS pool_list $FSNAME.$pool | grep -c "\-OST") + local addcount=$(((last - first) / step + 1)) + [ $lfscount -eq $addcount ] || { + error_noexit "lfs pool_list bad ost count" \ + "$lfscount != $addcount" + return 2 + } } -run_test 200d "Check pool on a directory ===============================" -test_200e() { - remote_mgs_nodsh && skip "remote MGS with nodsh" && return - failed=0 - for i in $(seq -w 1 $(($TGT_COUNT * 3))) +pool_set_dir() { + local pool=$1 + local tdir=$2 + echo "Setting pool on directory $tdir" + + $SETSTRIPE -c 2 -p $pool $tdir && return 0 + + error_noexit "Cannot set pool $pool to $tdir" + return 1 +} + +pool_check_dir() { + local pool=$1 + local tdir=$2 + echo "Checking pool on directory $tdir" + + local res=$($GETSTRIPE --pool $tdir | sed "s/\s*$//") + [ "$res" = "$pool" ] && return 0 + + error_noexit "Pool on '$tdir' is '$res', not '$pool'" + return 1 +} + +pool_dir_rel_path() { + echo "Testing relative path works well" + local pool=$1 + local tdir=$2 + local root=$3 + + mkdir -p $root/$tdir/$tdir + cd $root/$tdir + pool_set_dir $pool $tdir || return 1 + pool_set_dir $pool ./$tdir || return 2 + pool_set_dir $pool ../$tdir || return 3 + pool_set_dir $pool ../$tdir/$tdir || return 4 + rm -rf $tdir; cd - > /dev/null +} + +pool_alloc_files() { + echo "Checking files allocation from directory pool" + local pool=$1 + local tdir=$2 + local count=$3 + local tlist="$4" + + local failed=0 + for i in $(seq -w 1 $count) do - file=$POOL_DIR/file-$i + local file=$tdir/file-$i touch $file - check_file_in_pool $file - if [[ $? != 0 ]] - then - failed=$(($failed + 1)) - fi + check_file_in_pool $file $pool "$tlist" || \ + failed=$((failed + 1)) done - [ "$failed" = 0 ] || error "$failed files not allocated in $POOL" + [ "$failed" = 0 ] && return 0 + + error_noexit "$failed files not allocated in $pool" + return 1 } -run_test 200e "Check files allocation from directory pool ==============" -test_200f() { - remote_mgs_nodsh && skip "remote MGS with nodsh" && return - mkdir -p $POOL_FILE - failed=0 - for i in $(seq -w 1 $(($TGT_COUNT * 3))) +pool_create_files() { + echo "Creating files in pool" + local pool=$1 + local tdir=$2 + local count=$3 + local tlist="$4" + + mkdir -p $tdir + local failed=0 + for i in $(seq -w 1 $count) do - file=$POOL_FILE/spoo-$i - $SETSTRIPE -p $POOL $file - check_file_in_pool $file - if [[ $? != 0 ]] - then - failed=$(($failed + 1)) - fi + local file=$tdir/spoo-$i + $SETSTRIPE -p $pool $file + check_file_in_pool $file $pool "$tlist" || \ + failed=$((failed + 1)) done - [ "$failed" = 0 ] || error "$failed files not allocated in $POOL" + [ "$failed" = 0 ] && return 0 + + error_noexit "$failed files not allocated in $pool" + return 1 } -run_test 200f "Create files in a pool ===================================" -test_200g() { - remote_mgs_nodsh && skip "remote MGS with nodsh" && return - TGT=$($LCTL get_param -n lov.$FSNAME-clilov-*.pools.$POOL | tr '\n' ' ') - res=$($LFS df --pool $FSNAME.$POOL | awk '{print $1}' | grep "$FSNAME-OST" | tr '\n' ' ') - [ "$res" = "$TGT" ] || error "Pools OSTs '$TGT' is not '$res' that lfs df reports" +pool_lfs_df() { + echo "Checking 'lfs df' output" + local pool=$1 + + local t=$($LCTL get_param -n lov.$FSNAME-clilov-*.pools.$pool | + tr '\n' ' ') + local res=$($LFS df --pool $FSNAME.$pool | + awk '{print $1}' | + grep "$FSNAME-OST" | + tr '\n' ' ') + [ "$res" = "$t" ] && return 0 + + error_noexit "Pools OSTs '$t' is not '$res' that lfs df reports" + return 1 } -run_test 200g "lfs df a pool ============================================" -test_200h() { # b=24039 - mkdir -p $POOL_DIR || error "unable to create $POOL_DIR" +pool_file_rel_path() { + echo "Creating files in a pool with relative pathname" + local pool=$1 + local tdir=$2 - local file="/..$POOL_DIR/$tfile-1" - $SETSTRIPE -p $POOL $file || error "unable to create $file" + mkdir -p $tdir || + { error_noexit "unable to create $tdir"; return 1 ; } + local file="/..$tdir/$tfile-1" + $SETSTRIPE -p $pool $file || + { error_noexit "unable to create $file" ; return 2 ; } - cd $POOL_DIR - $SETSTRIPE -p $POOL $tfile-2 || \ - error "unable to create $tfile-2 in $POOL_DIR" + cd $tdir + $SETSTRIPE -p $pool $tfile-2 || { + error_noexit "unable to create $tfile-2 in $tdir" + return 3 + } } -run_test 200h "Create files in a pool with relative pathname ============" -test_201a() { - remote_mgs_nodsh && skip "remote MGS with nodsh" && return - TGT=$($LCTL get_param -n lov.$FSNAME-*.pools.$POOL | head -1) - do_facet mgs $LCTL pool_remove $FSNAME.$POOL $TGT - wait_update $HOSTNAME "lctl get_param -n lov.$FSNAME-*.pools.$POOL | grep $TGT" "" || - error "$TGT not removed from $FSNAME.$POOL" +pool_remove_first_target() { + echo "Removing first target from a pool" + local pool=$1 + + local pname="lov.$FSNAME-*.pools.$pool" + local t=$($LCTL get_param -n $pname | head -1) + do_facet mgs $LCTL pool_remove $FSNAME.$pool $t + wait_update $HOSTNAME "lctl get_param -n $pname | grep $t" "" || { + error_noexit "$t not removed from $FSNAME.$pool" + return 1 + } } -run_test 201a "Remove a target from a pool =============================" -test_201b() { - remote_mgs_nodsh && skip "remote MGS with nodsh" && return - for TGT in $($LCTL get_param -n lov.$FSNAME-*.pools.$POOL | sort -u) +pool_remove_all_targets() { + echo "Removing all targets from pool" + local pool=$1 + local file=$2 + local pname="lov.$FSNAME-*.pools.$pool" + for t in $($LCTL get_param -n $pname | sort -u) do - do_facet mgs $LCTL pool_remove $FSNAME.$POOL $TGT - done - wait_update $HOSTNAME "lctl get_param -n lov.$FSNAME-*.pools.$POOL" "" || - error "Pool $FSNAME.$POOL cannot be drained" - # striping on an empty/nonexistant pool should fall back to "pool of everything" - touch ${POOL_DIR}/$tfile || error "failed to use fallback striping for empty pool" + do_facet mgs $LCTL pool_remove $FSNAME.$pool $t + done + wait_update $HOSTNAME "lctl get_param -n $pname" "" || { + error_noexit "Pool $FSNAME.$pool cannot be drained" + return 1 + } + # striping on an empty/nonexistant pool should fall back + # to "pool of everything" + touch $file || { + error_noexit "failed to use fallback striping for empty pool" + return 2 + } # setstripe on an empty pool should fail - $SETSTRIPE -p $POOL ${POOL_FILE}/$tfile 2>/dev/null && \ - error "expected failure when creating file with empty pool" + $SETSTRIPE -p $pool $file 2>/dev/null && { + error_noexit "expected failure when creating file" \ + "with empty pool" + return 3 + } return 0 } -run_test 201b "Remove all targets from a pool ==========================" -test_201c() { - remote_mgs_nodsh && skip "remote MGS with nodsh" && return - do_facet mgs $LCTL pool_destroy $FSNAME.$POOL +pool_remove() { + echo "Destroying pool" + local pool=$1 + local file=$2 + + do_facet mgs $LCTL pool_destroy $FSNAME.$pool sleep 2 - # striping on an empty/nonexistant pool should fall back to "pool of everything" - touch ${POOL_DIR}/$tfile || error "failed to use fallback striping for missing pool" + # striping on an empty/nonexistant pool should fall back + # to "pool of everything" + touch $file || { + error_noexit "failed to use fallback striping for missing pool" + return 1 + } # setstripe on an empty pool should fail - $SETSTRIPE -p $POOL ${POOL_FILE}/$tfile 2>/dev/null && \ - error "expected failure when creating file with missing pool" + $SETSTRIPE -p $pool $file 2>/dev/null && { + error_noexit "expected failure when creating file" \ + "with missing pool" + return 2 + } # get param should return err once pool is gone - if wait_update $HOSTNAME "lctl get_param -n lov.$FSNAME-*.pools.$POOL 2>/dev/null || - echo foo" "foo"; then - remove_pool_from_list $FSNAME.$POOL + if wait_update $HOSTNAME "lctl get_param -n \ + lov.$FSNAME-*.pools.$pool 2>/dev/null || echo foo" "foo" + then + remove_pool_from_list $FSNAME.$pool return 0 fi - error "Pool $FSNAME.$POOL is not destroyed" + error_noexit "Pool $FSNAME.$pool is not destroyed" + return 3 } -run_test 201c "Remove a pool ============================================" -cleanup_pools $FSNAME +test_200() { + remote_mgs_nodsh && skip "remote MGS with nodsh" && return + + local POOL=${POOL:-cea1} + local POOL_ROOT=${POOL_ROOT:-$DIR/d200.pools} + local POOL_DIR_NAME=${POOL_DIR_NAME:-dir_tst} + # Pool OST targets + local first_ost=0 + local last_ost=$(($OSTCOUNT - 1)) + local ost_step=2 + local ost_list=$(seq $first_ost $ost_step $last_ost) + local ost_range="$first_ost $last_ost $ost_step" + local test_path=$POOL_ROOT/$POOL_DIR_NAME + local file_dir=$POOL_ROOT/file_tst + + local rc=0 + while : ; do + # former test_200a test_200b + pool_add $POOL || { rc=$? ; break; } + pool_add_targets $POOL $ost_range || { rc=$? ; break; } + # former test_200c test_200d + mkdir -p $test_path + pool_set_dir $POOL $test_path || { rc=$? ; break; } + pool_check_dir $POOL $test_path || { rc=$? ; break; } + pool_dir_rel_path $POOL $POOL_DIR_NAME $POOL_ROOT \ + || { rc=$? ; break; } + # former test_200e test_200f + local files=$((OSTCOUNT*3)) + pool_alloc_files $POOL $test_path $files "$ost_list" \ + || { rc=$? ; break; } + pool_create_files $POOL $file_dir $files "$ost_list" \ + || { rc=$? ; break; } + # former test_200g test_200h + pool_lfs_df $POOL || { rc=$? ; break; } + pool_file_rel_path $POOL $test_path || { rc=$? ; break; } + + # former test_201a test_201b test_201c + pool_remove_first_target $POOL || { rc=$? ; break; } + + local f=$test_path/$tfile + pool_remove_all_targets $POOL $f || { rc=$? ; break; } + pool_remove $POOL $f || { rc=$? ; break; } + break + done + + cleanup_pools + return $rc +} +run_test 200 "OST pools" # usage: default_attr default_attr() { @@ -8713,6 +9061,120 @@ test_204h() { } run_test 204h "Print raw stripe count and size =============" +# Figure out which job scheduler is being used, if any, +# or use a fake one +if [ -n "$SLURM_JOB_ID" ]; then # SLURM + JOBENV=SLURM_JOB_ID +elif [ -n "$LSB_JOBID" ]; then # Load Sharing Facility + JOBENV=LSB_JOBID +elif [ -n "$PBS_JOBID" ]; then # PBS/Maui/Moab + JOBENV=PBS_JOBID +elif [ -n "$LOADL_STEPID" ]; then # LoadLeveller + JOBENV=LOADL_STEP_ID +elif [ -n "$JOB_ID" ]; then # Sun Grid Engine + JOBENV=JOB_ID +else + JOBENV=FAKE_JOBID +fi + +verify_jobstats() { + local cmd=$1 + local target=$2 + + # clear old jobstats + do_facet $SINGLEMDS lctl set_param mdt.*.job_stats="clear" + do_facet ost0 lctl set_param obdfilter.*.job_stats="clear" + + # use a new JobID for this test, or we might see an old one + [ "$JOBENV" = "FAKE_JOBID" ] && FAKE_JOBID=test_id.$testnum.$RANDOM + + JOBVAL=${!JOBENV} + log "Test: $cmd" + log "Using JobID environment variable $JOBENV=$JOBVAL" + + if [ $JOBENV = "FAKE_JOBID" ]; then + FAKE_JOBID=$JOBVAL $cmd + else + $cmd + fi + + if [ "$target" = "mdt" -o "$target" = "both" ]; then + FACET="$SINGLEMDS" # will need to get MDS number for DNE + do_facet $FACET lctl get_param mdt.*.job_stats | + grep $JOBVAL || error "No job stats found on MDT $FACET" + fi + if [ "$target" = "ost" -o "$target" = "both" ]; then + FACET=ost0 + do_facet $FACET lctl get_param obdfilter.*.job_stats | + grep $JOBVAL || error "No job stats found on OST $FACET" + fi +} + +test_205() { # Job stats + [ -z "$(lctl get_param -n mdc.*.connect_flags | grep jobstats)" ] && + skip "Server doesn't support jobstats" && return 0 + + local cmd + OLD_JOBENV=`$LCTL get_param -n jobid_var` + if [ $OLD_JOBENV != $JOBENV ]; then + do_facet mgs $LCTL conf_param $FSNAME.sys.jobid_var=$JOBENV + wait_update $HOSTNAME "$LCTL get_param -n jobid_var" \ + $JOBENV || return 1 + fi + + # mkdir + cmd="mkdir $DIR/$tfile" + verify_jobstats "$cmd" "mdt" + # rmdir + cmd="rm -fr $DIR/$tfile" + verify_jobstats "$cmd" "mdt" + # mknod + cmd="mknod $DIR/$tfile c 1 3" + verify_jobstats "$cmd" "mdt" + # unlink + cmd="rm -f $DIR/$tfile" + verify_jobstats "$cmd" "mdt" + # open & close + cmd="$SETSTRIPE -i 0 -c 1 $DIR/$tfile" + verify_jobstats "$cmd" "mdt" + # setattr + cmd="touch $DIR/$tfile" + verify_jobstats "$cmd" "both" + # write + cmd="dd if=/dev/zero of=$DIR/$tfile bs=1M count=1 oflag=sync" + verify_jobstats "$cmd" "ost" + # read + cmd="dd if=$DIR/$tfile of=/dev/null bs=1M count=1 iflag=direct" + verify_jobstats "$cmd" "ost" + # truncate + cmd="$TRUNCATE $DIR/$tfile 0" + verify_jobstats "$cmd" "both" + # rename + cmd="mv -f $DIR/$tfile $DIR/jobstats_test_rename" + verify_jobstats "$cmd" "mdt" + + # cleanup + rm -f $DIR/jobstats_test_rename + + if [ $OLD_JOBENV != $JOBENV ]; then + do_facet mgs $LCTL conf_param $FSNAME.sys.jobid_var=$OLD_JOBENV + wait_update $HOSTNAME "$LCTL get_param -n jobid_var" \ + $OLD_JOBENV || return 1 + fi +} +run_test 205 "Verify job stats" + +# LU-1480, LU-1773 and LU-1657 +test_206() { + mkdir -p $DIR/$tdir + lfs setstripe -c -1 $DIR/$tdir +#define OBD_FAIL_LOV_INIT 0x1403 + $LCTL set_param fail_loc=0xa0001403 + $LCTL set_param fail_val=1 + touch $DIR/$tdir/$tfile || true +} +run_test 206 "fail lov_init_raid0() doesn't lbug" + test_212() { size=`date +%s` size=$((size % 8192 + 1)) @@ -8860,10 +9322,10 @@ test_215() { # for bugs 18102, 21079, 21517 # /proc/sys/lnet/nis should look like this: # nid status alive refs peer rtr max tx min # where nid is a string like 192.168.1.1@tcp2, status is up/down, - # alive is numeric (0 or >0 or <0), refs > 0, peer >= 0, + # alive is numeric (0 or >0 or <0), refs >= 0, peer >= 0, # rtr >= 0, max >=0, tx and min are numeric (0 or >0 or <0). L1="^nid +status +alive +refs +peer +rtr +max +tx +min$" - BR="^$NID +(up|down) +$I +$P +$N +$N +$N +$I +$I$" + BR="^$NID +(up|down) +$I +$N +$N +$N +$N +$I +$I$" create_lnet_proc_files "nis" check_lnet_proc_entry "nis.out" "/proc/sys/lnet/nis" "$BR" "$L1" check_lnet_proc_entry "nis.sys" "lnet.nis" "$BR" "$L1" @@ -8931,11 +9393,11 @@ run_test 217 "check lctl ping for hostnames with hiphen ('-')" test_218() { # do directio so as not to populate the page cache log "creating a 10 Mb file" - multiop $DIR/$tfile oO_CREAT:O_DIRECT:O_RDWR:w$((10*1048576))c || error "multiop failed while creating a file" + $MULTIOP $DIR/$tfile oO_CREAT:O_DIRECT:O_RDWR:w$((10*1048576))c || error "multiop failed while creating a file" log "starting reads" dd if=$DIR/$tfile of=/dev/null bs=4096 & log "truncating the file" - multiop $DIR/$tfile oO_TRUNC:c || error "multiop failed while truncating the file" + $MULTIOP $DIR/$tfile oO_TRUNC:c || error "multiop failed while truncating the file" log "killing dd" kill %+ || true # reads might have finished echo "wait until dd is finished" @@ -9184,6 +9646,148 @@ test_226 () { } run_test 226 "call path2fid and fid2path on files of all type" +# LU-1299 Executing or running ldd on a truncated executable does not +# cause an out-of-memory condition. +test_227() { + dd if=`which date` of=$MOUNT/date bs=1k count=1 + chmod +x $MOUNT/date + + $MOUNT/date > /dev/null + ldd $MOUNT/date > /dev/null + rm -f $MOUNT/date +} +run_test 227 "running truncated executable does not cause OOM" + +# LU-1512 try to reuse idle OI blocks +test_228a() { + [ "$FSTYPE" != "ldiskfs" ] && skip "non-ldiskfs backend" && return + + local MDT_DEV=$(mdsdevname ${SINGLEMDS//mds/}) + local myDIR=$DIR/$tdir + + mkdir -p $myDIR + #define OBD_FAIL_SEQ_EXHAUST 0x1002 + $LCTL set_param fail_loc=0x80001002 + createmany -o $myDIR/t- 10000 + $LCTL set_param fail_loc=0 + # The guard is current the largest FID holder + touch $myDIR/guard + local SEQ=$($LFS path2fid $myDIR/guard | awk -F ':' '{print $1}' | + tr -d '[') + local IDX=$(($SEQ % 64)) + + do_facet $SINGLEMDS sync + # Make sure journal flushed. + sleep 6 + local blk1=$(do_facet $SINGLEMDS \ + "$DEBUGFS -c -R \\\"stat oi.16.${IDX}\\\" $MDT_DEV" | + grep Blockcount | awk '{print $4}') + + # Remove old files, some OI blocks will become idle. + unlinkmany $myDIR/t- 10000 + # Create new files, idle OI blocks should be reused. + createmany -o $myDIR/t- 2000 + do_facet $SINGLEMDS sync + # Make sure journal flushed. + sleep 6 + local blk2=$(do_facet $SINGLEMDS \ + "$DEBUGFS -c -R \\\"stat oi.16.${IDX}\\\" $MDT_DEV" | + grep Blockcount | awk '{print $4}') + + [ $blk1 == $blk2 ] || error "old blk1=$blk1, new blk2=$blk2, unmatched!" +} +run_test 228a "try to reuse idle OI blocks" + +test_228b() { + [ "$FSTYPE" != "ldiskfs" ] && skip "non-ldiskfs backend" && return + + local MDT_DEV=$(mdsdevname ${SINGLEMDS//mds/}) + local myDIR=$DIR/$tdir + + mkdir -p $myDIR + #define OBD_FAIL_SEQ_EXHAUST 0x1002 + $LCTL set_param fail_loc=0x80001002 + createmany -o $myDIR/t- 10000 + $LCTL set_param fail_loc=0 + # The guard is current the largest FID holder + touch $myDIR/guard + local SEQ=$($LFS path2fid $myDIR/guard | awk -F ':' '{print $1}' | + tr -d '[') + local IDX=$(($SEQ % 64)) + + do_facet $SINGLEMDS sync + # Make sure journal flushed. + sleep 6 + local blk1=$(do_facet $SINGLEMDS \ + "$DEBUGFS -c -R \\\"stat oi.16.${IDX}\\\" $MDT_DEV" | + grep Blockcount | awk '{print $4}') + + # Remove old files, some OI blocks will become idle. + unlinkmany $myDIR/t- 10000 + + # stop the MDT + stop $SINGLEMDS || error "Fail to stop MDT." + # remount the MDT + start $SINGLEMDS $MDT_DEV $MDS_MOUNT_OPTS || error "Fail to start MDT." + + df $MOUNT || error "Fail to df." + # Create new files, idle OI blocks should be reused. + createmany -o $myDIR/t- 2000 + do_facet $SINGLEMDS sync + # Make sure journal flushed. + sleep 6 + local blk2=$(do_facet $SINGLEMDS \ + "$DEBUGFS -c -R \\\"stat oi.16.${IDX}\\\" $MDT_DEV" | + grep Blockcount | awk '{print $4}') + + [ $blk1 == $blk2 ] || error "old blk1=$blk1, new blk2=$blk2, unmatched!" +} +run_test 228b "idle OI blocks can be reused after MDT restart" + +#LU-1881 +test_228c() { + [ "$FSTYPE" != "ldiskfs" ] && skip "non-ldiskfs backend" && return + + local MDT_DEV=$(mdsdevname ${SINGLEMDS//mds/}) + local myDIR=$DIR/$tdir + + mkdir -p $myDIR + #define OBD_FAIL_SEQ_EXHAUST 0x1002 + $LCTL set_param fail_loc=0x80001002 + # 20000 files can guarantee there are index nodes in the OI file + createmany -o $myDIR/t- 20000 + $LCTL set_param fail_loc=0 + # The guard is current the largest FID holder + touch $myDIR/guard + local SEQ=$($LFS path2fid $myDIR/guard | awk -F ':' '{print $1}' | + tr -d '[') + local IDX=$(($SEQ % 64)) + + do_facet $SINGLEMDS sync + # Make sure journal flushed. + sleep 6 + local blk1=$(do_facet $SINGLEMDS \ + "$DEBUGFS -c -R \\\"stat oi.16.${IDX}\\\" $MDT_DEV" | + grep Blockcount | awk '{print $4}') + + # Remove old files, some OI blocks will become idle. + unlinkmany $myDIR/t- 20000 + rm -f $myDIR/guard + # The OI file should become empty now + + # Create new files, idle OI blocks should be reused. + createmany -o $myDIR/t- 2000 + do_facet $SINGLEMDS sync + # Make sure journal flushed. + sleep 6 + local blk2=$(do_facet $SINGLEMDS \ + "$DEBUGFS -c -R \\\"stat oi.16.${IDX}\\\" $MDT_DEV" | + grep Blockcount | awk '{print $4}') + + [ $blk1 == $blk2 ] || error "old blk1=$blk1, new blk2=$blk2, unmatched!" +} +run_test 228c "NOT shrink the last entry in OI index node to recycle idle leaf" + # # tests that do cleanup/setup should be run at the end #