X-Git-Url: https://git.whamcloud.com/?p=fs%2Flustre-release.git;a=blobdiff_plain;f=lustre%2Ftests%2Fsanity.sh;h=4859d9e20c65264ec351b51ab22986f3b9f8c5be;hp=c73694131b0f77d4b4d2fb8a8b8761ccf8507230;hb=c7ed812d6a87f0ffa3c0d4c1a9289497f680ccec;hpb=048a8740ae26e3406a7eab3bca383a90490cef93 diff --git a/lustre/tests/sanity.sh b/lustre/tests/sanity.sh index c736941..4859d9e 100755 --- a/lustre/tests/sanity.sh +++ b/lustre/tests/sanity.sh @@ -8,13 +8,15 @@ set -e ONLY=${ONLY:-"$*"} -# bug number for skipped test: 13297 2108 9789 3637 9789 3561 5188 -ALWAYS_EXCEPT=" 42a 42b 42c 42d 45 68b $SANITY_EXCEPT" +# bug number for skipped test: LU-9693 LU-6493 LU-9693 3561 5188 +ALWAYS_EXCEPT=" 42a 42b 42c 45 68b $SANITY_EXCEPT" # UPDATE THE COMMENT ABOVE WITH BUG NUMBERS WHEN CHANGING ALWAYS_EXCEPT! -# with LOD/OSP landing -# bug number for skipped tests: LU-2036 LU-8139 -ALWAYS_EXCEPT=" 76 101g $ALWAYS_EXCEPT" +# skipped tests: LU-8411 LU-9096 LU-9054 LU-10199 +ALWAYS_EXCEPT=" 407 253 312 56xb $ALWAYS_EXCEPT" + +# Check Grants after these tests +GRANT_CHECK_LIST="$GRANT_CHECK_LIST 42a 42b 42c 42d 42e 63a 63b 64a 64b 64c" is_sles11() # LU-4341 { @@ -38,6 +40,7 @@ SRCDIR=$(cd $(dirname $0); echo $PWD) export PATH=$PATH:/sbin TMP=${TMP:-/tmp} +OSC=${OSC:-"osc"} CC=${CC:-cc} CHECKSTAT=${CHECKSTAT:-"checkstat -v"} @@ -75,12 +78,12 @@ init_test_env $@ . ${CONFIG:=$LUSTRE/tests/cfg/${NAME}.sh} init_logging -# 5 12 (min)" -[ "$SLOW" = "no" ] && EXCEPT_SLOW="24D 27m 64b 68 71 115 300o" +# 5 12 (min)" +[ "$SLOW" = "no" ] && EXCEPT_SLOW="27m 64b 68 71 115 300o" if [ $(facet_fstype $SINGLEMDS) = "zfs" ]; then - # bug number for skipped test: LU-4536 LU-1957 - ALWAYS_EXCEPT="$ALWAYS_EXCEPT 65ic 180" + # bug number for skipped test: LU-1957 + ALWAYS_EXCEPT="$ALWAYS_EXCEPT 180" # 13 (min)" [ "$SLOW" = "no" ] && EXCEPT_SLOW="$EXCEPT_SLOW 51b" fi @@ -160,44 +163,41 @@ test_0c() { $LCTL get_param mdc.*.import | grep "target: $FSNAME-MDT" || error "bad target" } -run_test 0c "check import proc =============================" +run_test 0c "check import proc" test_1() { - test_mkdir -p $DIR/$tdir || error "mkdir $tdir failed" - test_mkdir -p $DIR/$tdir/d2 || error "mkdir $tdir/d2 failed" - test_mkdir $DIR/$tdir/d2 && error "we expect EEXIST, but not returned" + test_mkdir $DIR/$tdir + test_mkdir $DIR/$tdir/d2 + mkdir $DIR/$tdir/d2 && error "we expect EEXIST, but not returned" $CHECKSTAT -t dir $DIR/$tdir/d2 || error "$tdir/d2 is not a dir" rmdir $DIR/$tdir/d2 rmdir $DIR/$tdir $CHECKSTAT -a $DIR/$tdir || error "$tdir was not removed" } -run_test 1 "mkdir; remkdir; rmdir ==============================" +run_test 1 "mkdir; remkdir; rmdir" test_2() { - test_mkdir -p $DIR/$tdir || error "mkdir $tdir failed" + test_mkdir $DIR/$tdir touch $DIR/$tdir/$tfile || error "touch $tdir/$tfile failed" $CHECKSTAT -t file $DIR/$tdir/$tfile || error "$tdir/$tfile not a file" rm -r $DIR/$tdir $CHECKSTAT -a $DIR/$tdir/$tfile || error "$tdir/$file is not removed" } -run_test 2 "mkdir; touch; rmdir; check file ====================" +run_test 2 "mkdir; touch; rmdir; check file" test_3() { - test_mkdir -p $DIR/$tdir || error "mkdir $tdir failed" + test_mkdir $DIR/$tdir $CHECKSTAT -t dir $DIR/$tdir || error "$tdir is not a directory" touch $DIR/$tdir/$tfile $CHECKSTAT -t file $DIR/$tdir/$tfile || error "$tdir/$tfile not a file" rm -r $DIR/$tdir $CHECKSTAT -a $DIR/$tdir || error "$tdir is not removed" } -run_test 3 "mkdir; touch; rmdir; check dir =====================" +run_test 3 "mkdir; touch; rmdir; check dir" # LU-4471 - failed rmdir on remote directories still removes directory on MDT0 test_4() { - local MDTIDX=1 - - test_mkdir $DIR/$tdir || - error "Create remote directory failed" + test_mkdir -i 1 $DIR/$tdir touch $DIR/$tdir/$tfile || error "Create file under remote directory failed" @@ -212,13 +212,13 @@ test_4() { run_test 4 "mkdir; touch dir/file; rmdir; checkdir (expect error)" test_5() { - test_mkdir -p $DIR/$tdir || error "mkdir $tdir failed" - test_mkdir $DIR/$tdir/d2 || error "mkdir $tdir/d2 failed" + test_mkdir $DIR/$tdir + test_mkdir $DIR/$tdir/d2 chmod 0707 $DIR/$tdir/d2 || error "chmod 0707 $tdir/d2 failed" $CHECKSTAT -t dir -p 0707 $DIR/$tdir/d2 || error "$tdir/d2 not mode 707" $CHECKSTAT -t dir $DIR/$tdir/d2 || error "$tdir/d2 is not a directory" } -run_test 5 "mkdir .../d5 .../d5/d2; chmod .../d5/d2 ============" +run_test 5 "mkdir .../d5 .../d5/d2; chmod .../d5/d2" test_6a() { touch $DIR/$tfile || error "touch $DIR/$tfile failed" @@ -253,15 +253,15 @@ test_6e() { $CHECKSTAT -t file -u \#$UID -g \#$RUNAS_ID $DIR/$tfile || error "$tfile should be owned by UID $UID and GID $RUNAS_ID" } -run_test 6e "touch f6e; chgrp f6e; $RUNAS chgrp f6e (should return error) ==" +run_test 6e "touch+chgrp $tfile; $RUNAS chgrp $tfile (should return error)" test_6g() { [ $RUNAS_ID -eq $UID ] && skip_env "RUNAS_ID = UID = $UID" && return - test_mkdir $DIR/$tdir || error "mkdir $tfile failed" + test_mkdir $DIR/$tdir chmod 777 $DIR/$tdir || error "chmod 0777 $tdir failed" $RUNAS mkdir $DIR/$tdir/d || error "mkdir $tdir/d failed" chmod g+s $DIR/$tdir/d || error "chmod g+s $tdir/d failed" - test_mkdir $DIR/$tdir/d/subdir || error "mkdir $tdir/d/subdir failed" + test_mkdir $DIR/$tdir/d/subdir $CHECKSTAT -g \#$RUNAS_GID $DIR/$tdir/d/subdir || error "$tdir/d/subdir should be GID $RUNAS_GID" if [[ $MDSCOUNT -gt 1 ]]; then @@ -280,7 +280,7 @@ test_6g() { error "$tdir.remote should be mode 02755" fi } -run_test 6g "Is new dir in sgid dir inheriting group?" +run_test 6g "verify new dir in sgid dir inherits group" test_6h() { # bug 7331 [ $RUNAS_ID -eq $UID ] && skip_env "RUNAS_ID = UID = $UID" && return @@ -291,7 +291,7 @@ test_6h() { # bug 7331 $CHECKSTAT -t file -u \#$RUNAS_ID -g \#$RUNAS_GID $DIR/$tfile || error "$tdir/$tfile should be UID $RUNAS_UID GID $RUNAS_GID" } -run_test 6h "$RUNAS chown RUNAS_ID.0 .../f6h (should return error)" +run_test 6h "$RUNAS chown RUNAS_ID.0 .../$tfile (should return error)" test_7a() { test_mkdir $DIR/$tdir @@ -392,10 +392,10 @@ test_16() { rm -rf $DIR/$tdir/$tfile $CHECKSTAT -a $DIR/$tdir/$tfile || error "$tdir/$tfile not removed" } -run_test 16 "touch .../d16/f; rm -rf .../d16/f =================" +run_test 16 "touch .../d16/f; rm -rf .../d16/f" test_17a() { - test_mkdir -p $DIR/$tdir + test_mkdir $DIR/$tdir touch $DIR/$tdir/$tfile ln -s $DIR/$tdir/$tfile $DIR/$tdir/l-exist ls -l $DIR/$tdir @@ -406,10 +406,10 @@ test_17a() { rm -f $DIR/$tdir/l-exist $CHECKSTAT -a $DIR/$tdir/l-exist || error "$tdir/l-exist not removed" } -run_test 17a "symlinks: create, remove (real) ==================" +run_test 17a "symlinks: create, remove (real)" test_17b() { - test_mkdir -p $DIR/$tdir + test_mkdir $DIR/$tdir ln -s no-such-file $DIR/$tdir/l-dangle ls -l $DIR/$tdir $CHECKSTAT -l no-such-file $DIR/$tdir/l-dangle || @@ -419,33 +419,33 @@ test_17b() { rm -f $DIR/$tdir/l-dangle $CHECKSTAT -a $DIR/$tdir/l-dangle || error "$tdir/l-dangle not removed" } -run_test 17b "symlinks: create, remove (dangling) ==============" +run_test 17b "symlinks: create, remove (dangling)" test_17c() { # bug 3440 - don't save failed open RPC for replay - test_mkdir -p $DIR/$tdir + test_mkdir $DIR/$tdir ln -s foo $DIR/$tdir/$tfile cat $DIR/$tdir/$tfile && error "opened non-existent symlink" || true } -run_test 17c "symlinks: open dangling (should return error) ====" +run_test 17c "symlinks: open dangling (should return error)" test_17d() { - test_mkdir -p $DIR/$tdir + test_mkdir $DIR/$tdir ln -s foo $DIR/$tdir/$tfile touch $DIR/$tdir/$tfile || error "creating to new symlink" } -run_test 17d "symlinks: create dangling ========================" +run_test 17d "symlinks: create dangling" test_17e() { - test_mkdir -p $DIR/$tdir + test_mkdir $DIR/$tdir local foo=$DIR/$tdir/$tfile ln -s $foo $foo || error "create symlink failed" ls -l $foo || error "ls -l failed" ls $foo && error "ls not failed" || true } -run_test 17e "symlinks: create recursive symlink (should return error) ====" +run_test 17e "symlinks: create recursive symlink (should return error)" test_17f() { - test_mkdir -p $DIR/$tdir + test_mkdir $DIR/$tdir ln -s 1234567890/2234567890/3234567890/4234567890 $DIR/$tdir/111 ln -s 1234567890/2234567890/3234567890/4234567890/5234567890/6234567890 $DIR/$tdir/222 ln -s 1234567890/2234567890/3234567890/4234567890/5234567890/6234567890/7234567890/8234567890 $DIR/$tdir/333 @@ -454,7 +454,7 @@ test_17f() { ln -s 1234567890/2234567890/3234567890/4234567890/5234567890/6234567890/7234567890/8234567890/9234567890/a234567890/b234567890/c234567890/d234567890/f234567890/aaaaaaaaaa/bbbbbbbbbb/cccccccccc/dddddddddd/eeeeeeeeee/ffffffffff/ $DIR/$tdir/666 ls -l $DIR/$tdir } -run_test 17f "symlinks: long and very long symlink name ========================" +run_test 17f "symlinks: long and very long symlink name" # str_repeat(S, N) generate a string that is string S repeated N times str_repeat() { @@ -469,7 +469,7 @@ str_repeat() { # Long symlinks and LU-2241 test_17g() { - test_mkdir -p $DIR/$tdir + test_mkdir $DIR/$tdir local TESTS="59 60 61 4094 4095" # Fix for inode size boundary in 2.1.4 @@ -498,16 +498,16 @@ test_17h() { #bug 17378 remote_mds_nodsh && skip "remote MDS with nodsh" && return [ $PARALLEL == "yes" ] && skip "skip parallel run" && return local mdt_idx - test_mkdir -p $DIR/$tdir + test_mkdir $DIR/$tdir if [[ $MDSCOUNT -gt 1 ]]; then mdt_idx=$($LFS getdirstripe -i $DIR/$tdir) else mdt_idx=0 fi - $SETSTRIPE -c -1 $DIR/$tdir -#define OBD_FAIL_MDS_LOV_PREP_CREATE 0x141 - do_facet mds$((mdt_idx + 1)) lctl set_param fail_loc=0x80000141 - touch $DIR/$tdir/$tfile || true + $LFS setstripe -c -1 $DIR/$tdir + #define OBD_FAIL_MDS_LOV_PREP_CREATE 0x141 + do_facet mds$((mdt_idx + 1)) lctl set_param fail_loc=0x80000141 + touch $DIR/$tdir/$tfile || true } run_test 17h "create objects: lov_free_memmd() doesn't lbug" @@ -536,19 +536,19 @@ test_17k() { #bug 22301 rsync --help | grep -q xattr || skip_env "$(rsync --version | head -n1) does not support xattrs" [ $PARALLEL == "yes" ] && skip "skip parallel run" && return 0 - test_mkdir -p $DIR/$tdir - test_mkdir -p $DIR/$tdir.new + test_mkdir $DIR/$tdir + test_mkdir $DIR/$tdir.new touch $DIR/$tdir/$tfile ln -s $DIR/$tdir/$tfile $DIR/$tdir/$tfile.lnk rsync -av -X $DIR/$tdir/ $DIR/$tdir.new || error "rsync failed with xattrs enabled" } -run_test 17k "symlinks: rsync with xattrs enabled =========================" +run_test 17k "symlinks: rsync with xattrs enabled" test_17l() { # LU-279 [[ -z "$(which getfattr 2>/dev/null)" ]] && skip "no getfattr command" && return 0 - test_mkdir -p $DIR/$tdir + test_mkdir $DIR/$tdir touch $DIR/$tdir/$tfile ln -s $DIR/$tdir/$tfile $DIR/$tdir/$tfile.lnk for path in "$DIR/$tdir" "$DIR/$tdir/$tfile" "$DIR/$tdir/$tfile.lnk"; do @@ -561,12 +561,12 @@ test_17l() { # LU-279 done done } -run_test 17l "Ensure lgetxattr's returned xattr size is consistent ========" +run_test 17l "Ensure lgetxattr's returned xattr size is consistent" # LU-1540 test_17m() { local short_sym="0123456789" - local WDIR=$DIR/${tdir}m + local wdir=$DIR/$tdir local i remote_mds_nodsh && skip "remote MDS with nodsh" && return @@ -575,36 +575,36 @@ test_17m() { skip "MDS 2.2.0-2.2.93 do not NUL-terminate symlinks" && return [ "$(facet_fstype $SINGLEMDS)" != "ldiskfs" ] && - skip "only for ldiskfs MDT" && return 0 + skip "ldiskfs only test" && return 0 [ $PARALLEL == "yes" ] && skip "skip parallel run" && return - test_mkdir -p $WDIR + test_mkdir $wdir long_sym=$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" + 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 + ln -sf ${long_sym}"a5a5" $wdir/long-$i + ln -sf ${short_sym}"a5a5" $wdir/short-$i done echo "erase them" - rm -f $WDIR/* + rm -f $wdir/* sync wait_delete_completed 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 || error "long_sym failed" - ln -sf ${short_sym} $WDIR/short-$i || error "short_sym failed" + ln -sf ${long_sym} $wdir/long-$i || error "long_sym failed" + ln -sf ${short_sym} $wdir/short-$i || error "short_sym failed" done - local mds_index=$(($($LFS getstripe -M $WDIR) + 1)) + local mds_index=$(($($LFS getstripe -M $wdir) + 1)) local devname=$(mdsdevname $mds_index) echo "stop and checking mds${mds_index}:" @@ -613,7 +613,8 @@ test_17m() { run_e2fsck $(facet_active_host mds${mds_index}) $devname -n rc=$? - start mds${mds_index} $devname $MDS_MOUNT_OPTS || error "start failed" + start mds${mds_index} $devname $MDS_MOUNT_OPTS || + error "start mds${mds_index} failed" df $MOUNT > /dev/null 2>&1 [ $rc -eq 0 ] || error "e2fsck detected error for short/long symlink: rc=$rc" @@ -649,7 +650,7 @@ test_17n() { skip "MDS 2.2.0-2.2.93 do not NUL-terminate symlinks" && return [ "$(facet_fstype $SINGLEMDS)" != "ldiskfs" ] && - skip "only for ldiskfs MDT" && return 0 + skip "ldiskfs only test" && return 0 [[ $MDSCOUNT -lt 2 ]] && skip "needs >= 2 MDTs" && return @@ -700,26 +701,25 @@ test_17o() { [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.3.64) ] && skip "Need MDS version at least 2.3.64" && return - local WDIR=$DIR/${tdir}o + local wdir=$DIR/${tdir}o local mdt_index local rc=0 - test_mkdir -p $WDIR - mdt_index=$($LFS getstripe -M $WDIR) - mdt_index=$((mdt_index+1)) - - touch $WDIR/$tfile + test_mkdir $wdir + touch $wdir/$tfile + mdt_index=$($LFS getstripe -M $wdir/$tfile) + mdt_index=$((mdt_index + 1)) + cancel_lru_locks mdc #fail mds will wait the failover finish then set #following fail_loc to avoid interfer the recovery process. fail mds${mdt_index} #define OBD_FAIL_OSD_LMA_INCOMPAT 0x194 do_facet mds${mdt_index} lctl set_param fail_loc=0x194 - ls -l $WDIR/$tfile && rc=1 + ls -l $wdir/$tfile && rc=1 do_facet mds${mdt_index} lctl set_param fail_loc=0 - [[ $rc -ne 0 ]] && error "stat file should fail" - true + [[ $rc -eq 0 ]] || error "stat file should fail" } run_test 17o "stat file with incompat LMA feature" @@ -757,19 +757,16 @@ run_test 19d "cat .../f19 (should return error) ==============" test_20() { touch $DIR/$tfile rm $DIR/$tfile - log "1 done" touch $DIR/$tfile rm $DIR/$tfile - log "2 done" touch $DIR/$tfile rm $DIR/$tfile - log "3 done" $CHECKSTAT -a $DIR/$tfile || error "$tfile was not removed" } -run_test 20 "touch .../f ; ls -l ... ===========================" +run_test 20 "touch .../f ; ls -l ..." test_21() { - test_mkdir -p $DIR/$tdir + test_mkdir $DIR/$tdir [ -f $DIR/$tdir/dangle ] && rm -f $DIR/$tdir/dangle ln -s dangle $DIR/$tdir/link echo foo >> $DIR/$tdir/link @@ -778,43 +775,44 @@ test_21() { $CHECKSTAT -f -t file $DIR/$tdir/link || error "$tdir/link not linked to a file" } -run_test 21 "write to dangling link ============================" +run_test 21 "write to dangling link" test_22() { - WDIR=$DIR/$tdir - test_mkdir -p $DIR/$tdir - chown $RUNAS_ID:$RUNAS_GID $WDIR - (cd $WDIR || error "cd $WDIR failed"; - $RUNAS tar cf - /etc/hosts /etc/sysconfig/network | \ - $RUNAS tar xf -) - ls -lR $WDIR/etc || error "ls -lR $WDIR/etc failed" - $CHECKSTAT -t dir $WDIR/etc || error "checkstat -t dir failed" - $CHECKSTAT -u \#$RUNAS_ID -g \#$RUNAS_GID $WDIR/etc || error "checkstat -u failed" -} -run_test 22 "unpack tar archive as non-root user ===============" + local wdir=$DIR/$tdir + test_mkdir $wdir + chown $RUNAS_ID:$RUNAS_GID $wdir + (cd $wdir || error "cd $wdir failed"; + $RUNAS tar cf - /etc/hosts /etc/sysconfig/network | + $RUNAS tar xf -) + ls -lR $wdir/etc || error "ls -lR $wdir/etc failed" + $CHECKSTAT -t dir $wdir/etc || error "checkstat -t dir failed" + $CHECKSTAT -u \#$RUNAS_ID -g \#$RUNAS_GID $wdir/etc || + error "checkstat -u failed" +} +run_test 22 "unpack tar archive as non-root user" # was test_23 test_23a() { - test_mkdir -p $DIR/$tdir + test_mkdir $DIR/$tdir local file=$DIR/$tdir/$tfile openfile -f O_CREAT:O_EXCL $file || error "$file create failed" openfile -f O_CREAT:O_EXCL $file && error "$file recreate succeeded" || true } -run_test 23a "O_CREAT|O_EXCL in subdir ==========================" +run_test 23a "O_CREAT|O_EXCL in subdir" test_23b() { # bug 18988 - test_mkdir -p $DIR/$tdir + test_mkdir $DIR/$tdir local file=$DIR/$tdir/$tfile - rm -f $file - echo foo > $file || error "write filed" - echo bar >> $file || error "append filed" - $CHECKSTAT -s 8 $file || error "wrong size" - rm $file + rm -f $file + echo foo > $file || error "write filed" + echo bar >> $file || error "append filed" + $CHECKSTAT -s 8 $file || error "wrong size" + rm $file } -run_test 23b "O_APPEND check ==========================" +run_test 23b "O_APPEND check" # rename sanity test_24a() { @@ -956,10 +954,10 @@ test_24n() { run_test 24n "Statting the old file after renaming (Posix rename 2)" test_24o() { - test_mkdir -p $DIR/d24o - rename_many -s random -v -n 10 $DIR/d24o + test_mkdir $DIR/$tdir + rename_many -s random -v -n 10 $DIR/$tdir } -run_test 24o "rename of files during htree split ===============" +run_test 24o "rename of files during htree split" test_24p() { test_mkdir $DIR/R12a @@ -1038,53 +1036,73 @@ page_size() { } simple_cleanup_common() { + local rc=0 trap 0 + [ -z "$DIR" -o -z "$tdir" ] && return 0 + + local start=$SECONDS rm -rf $DIR/$tdir + rc=$? wait_delete_completed + echo "cleanup time $((SECONDS - start))" + return $rc } max_pages_per_rpc() { - $LCTL get_param -n mdc.*.max_pages_per_rpc | head -n1 + local mdtname="$(printf "MDT%04x" ${1:-0})" + $LCTL get_param -n mdc.*$mdtname*.max_pages_per_rpc } test_24v() { - local NRFILES=100000 - local FREE_INODES=$(mdt_free_inodes 0) - [[ $FREE_INODES -lt $NRFILES ]] && - skip "not enough free inodes $FREE_INODES required $NRFILES" && - return - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return - trap simple_cleanup_common EXIT - + local nrfiles=${COUNT:-100000} # Performance issue on ZFS see LU-4072 (c.f. LU-2887) - [ $(facet_fstype $SINGLEMDS) = "zfs" ] && NRFILES=10000 + [ $(facet_fstype $SINGLEMDS) = "zfs" ] && nrfiles=${COUNT:-10000} - test_mkdir -p $DIR/$tdir - createmany -m $DIR/$tdir/$tfile $NRFILES + local fname="$DIR/$tdir/$tfile" + test_mkdir "$(dirname $fname)" + # assume MDT0000 has the fewest inodes + local stripes=$($LFS getdirstripe -c $(dirname $fname)) + local free_inodes=$(($(mdt_free_inodes 0) * stripes)) + [[ $free_inodes -lt $nrfiles ]] && nrfiles=$free_inodes + + trap simple_cleanup_common EXIT + + createmany -m "$fname" $nrfiles cancel_lru_locks mdc lctl set_param mdc.*.stats clear - ls $DIR/$tdir >/dev/null || error "error in listing large dir" - + # was previously test_24D: LU-6101 + # readdir() returns correct number of entries after cursor reload + local num_ls=$(ls $DIR/$tdir | wc -l) + local num_uniq=$(ls $DIR/$tdir | sort -u | wc -l) + local num_all=$(ls -a $DIR/$tdir | wc -l) + if [ $num_ls -ne $nrfiles -o $num_uniq -ne $nrfiles -o \ + $num_all -ne $((nrfiles + 2)) ]; then + error "Expected $nrfiles files, got $num_ls " \ + "($num_uniq unique $num_all .&..)" + fi # LU-5 large readdir - # DIRENT_SIZE = 32 bytes for sizeof(struct lu_dirent) + - # 8 bytes for name(filename is mostly 5 in this test) + - # 8 bytes for luda_type + # dirent_size = 32 bytes for sizeof(struct lu_dirent) + + # N bytes for name (len($nrfiles) rounded to 8 bytes) + + # 8 bytes for luda_type (4 bytes rounded to 8 bytes) # take into account of overhead in lu_dirpage header and end mark in - # each page, plus one in RPC_NUM calculation. - DIRENT_SIZE=48 - RPC_SIZE=$(($(max_pages_per_rpc) * $(page_size))) - RPC_NUM=$(((NRFILES * DIRENT_SIZE + RPC_SIZE - 1) / RPC_SIZE + 1)) - mds_readpage=$(lctl get_param mdc.*MDT0000*.stats | - awk '/^mds_readpage/ {print $2}') - [[ $mds_readpage -gt $RPC_NUM ]] && - error "large readdir doesn't take effect" + # each page, plus one in rpc_num calculation. + local dirent_size=$((32 + (${#tfile} | 7) + 1 + 8)) + local page_entries=$((($(page_size) - 24) / dirent_size)) + local mdt_idx=$($LFS getdirstripe -i $(dirname $fname)) + local rpc_pages=$(max_pages_per_rpc $mdt_idx) + local rpc_max=$((nrfiles / (page_entries * rpc_pages) + stripes)) + local mds_readpage=$(calc_stats mdc.*.stats mds_readpage) + echo "readpages: $mds_readpage rpc_max: $rpc_max" + (( $mds_readpage < $rpc_max - 2 || $mds_readpage > $rpc_max + 1)) && + error "large readdir doesn't take effect: " \ + "$mds_readpage should be about $rpc_max" simple_cleanup_common } -run_test 24v "list directory with large files (handle hash collision, bug: 17560)" +run_test 24v "list large directory (test hash collision, b=17560)" test_24w() { # bug21506 SZ1=234852 @@ -1107,13 +1125,13 @@ test_24x() { local MDTIDX=1 local remote_dir=$DIR/$tdir/remote_dir - test_mkdir -p $DIR/$tdir + test_mkdir $DIR/$tdir $LFS mkdir -i $MDTIDX $remote_dir || error "create remote directory failed" - test_mkdir -p $DIR/$tdir/src_dir + test_mkdir $DIR/$tdir/src_dir touch $DIR/$tdir/src_file - test_mkdir -p $remote_dir/tgt_dir + test_mkdir $remote_dir/tgt_dir touch $remote_dir/tgt_file mrename $DIR/$tdir/src_dir $remote_dir/tgt_dir || @@ -1133,16 +1151,16 @@ run_test 24x "cross MDT rename/link" test_24y() { [[ $MDSCOUNT -lt 2 ]] && skip "needs >= 2 MDTs" && return [ $PARALLEL == "yes" ] && skip "skip parallel run" && return - local MDTIDX=1 local remote_dir=$DIR/$tdir/remote_dir + local mdtidx=1 - test_mkdir -p $DIR/$tdir - $LFS mkdir -i $MDTIDX $remote_dir || + test_mkdir $DIR/$tdir + $LFS mkdir -i $mdtidx $remote_dir || error "create remote directory failed" - test_mkdir -p $remote_dir/src_dir + test_mkdir $remote_dir/src_dir touch $remote_dir/src_file - test_mkdir -p $remote_dir/tgt_dir + test_mkdir $remote_dir/tgt_dir touch $remote_dir/tgt_file mrename $remote_dir/src_dir $remote_dir/tgt_dir || @@ -1162,7 +1180,8 @@ test_24A() { # LU-3182 local NFILES=5000 rm -rf $DIR/$tdir - test_mkdir -p $DIR/$tdir + test_mkdir $DIR/$tdir + trap simple_cleanup_common EXIT createmany -m $DIR/$tdir/$tfile $NFILES local t=$(ls $DIR/$tdir | wc -l) local u=$(ls $DIR/$tdir | sort -u | wc -l) @@ -1171,7 +1190,7 @@ test_24A() { # LU-3182 error "Expected $NFILES files, got $t ($u unique $v .&..)" fi - rm -rf $DIR/$tdir || error "Can not delete directories" + simple_cleanup_common || error "Can not delete directories" } run_test 24A "readdir() returns correct number of entries." @@ -1229,23 +1248,6 @@ test_24C() { } run_test 24C "check .. in striped dir" -test_24D() { # LU-6101 - local NFILES=50000 - - rm -rf $DIR/$tdir - mkdir -p $DIR/$tdir - createmany -m $DIR/$tdir/$tfile $NFILES - local t=$(ls $DIR/$tdir | wc -l) - local u=$(ls $DIR/$tdir | sort -u | wc -l) - local v=$(ls -ai $DIR/$tdir | sort -u | wc -l) - if [ $t -ne $NFILES -o $u -ne $NFILES -o $v -ne $((NFILES + 2)) ] ; then - error "Expected $NFILES files, got $t ($u unique $v .&..)" - fi - - rm -rf $DIR/$tdir || error "Can not delete directories" -} -run_test 24D "readdir() returns correct number of entries after cursor reload" - test_24E() { [[ $MDSCOUNT -lt 4 ]] && skip "needs >= 4 MDTs" && return [ $PARALLEL == "yes" ] && skip "skip parallel run" && return @@ -1302,8 +1304,8 @@ test_26a() { run_test 26a "multiple component symlink =======================" test_26b() { - test_mkdir -p $DIR/d26b/d26-2 - ln -s d26b/d26-2/foo $DIR/s26-2 + test_mkdir -p $DIR/$tdir/d26-2 + ln -s $tdir/d26-2/foo $DIR/s26-2 touch $DIR/s26-2 || error } run_test 26b "multiple component symlink at end of lookup ======" @@ -1316,27 +1318,27 @@ test_26c() { ln -s s26.2-2 $DIR/s26.2-3 chmod 0666 $DIR/s26.2-3/foo } -run_test 26c "chain of symlinks ================================" +run_test 26c "chain of symlinks" # recursive symlinks (bug 439) test_26d() { ln -s d26-3/foo $DIR/d26-3 } -run_test 26d "create multiple component recursive symlink ======" +run_test 26d "create multiple component recursive symlink" test_26e() { [ ! -h $DIR/d26-3 ] && test_26d rm $DIR/d26-3 } -run_test 26e "unlink multiple component recursive symlink ======" +run_test 26e "unlink multiple component recursive symlink" # recursive symlinks (bug 7022) test_26f() { - test_mkdir -p $DIR/$tdir - test_mkdir $DIR/$tdir/$tfile || error "mkdir $DIR/$tdir/$tfile failed" + test_mkdir $DIR/$tdir + test_mkdir $DIR/$tdir/$tfile cd $DIR/$tdir/$tfile || error "cd $DIR/$tdir/$tfile failed" - test_mkdir -p lndir/bar1 || error "mkdir lndir/bar1 failed" - test_mkdir $DIR/$tdir/$tfile/$tfile || error "mkdir $tfile failed" + test_mkdir -p lndir/bar1 + test_mkdir $DIR/$tdir/$tfile/$tfile cd $tfile || error "cd $tfile failed" ln -s .. dotdot || error "ln dotdot failed" ln -s dotdot/lndir lndir || error "ln lndir failed" @@ -1346,108 +1348,105 @@ test_26f() { rm -r $tfile || error "rm $tfile failed" $CHECKSTAT -a $DIR/$tfile || error "$tfile not gone" } -run_test 26f "rm -r of a directory which has recursive symlink =" +run_test 26f "rm -r of a directory which has recursive symlink" test_27a() { - echo '== stripe sanity ==============================================' - test_mkdir -p $DIR/d27 || error "mkdir failed" - $GETSTRIPE $DIR/d27 - $SETSTRIPE -c 1 $DIR/d27/f0 || error "setstripe failed" - $CHECKSTAT -t file $DIR/d27/f0 || error "checkstat failed" - log "== test_27a: write to one stripe file =========================" - cp /etc/hosts $DIR/d27/f0 || error + test_mkdir $DIR/$tdir + $LFS getstripe $DIR/$tdir + $LFS setstripe -c 1 $DIR/$tdir/$tfile || error "setstripe failed" + $CHECKSTAT -t file $DIR/$tdir/$tfile || error "checkstat failed" + cp /etc/hosts $DIR/$tdir/$tfile || error } -run_test 27a "one stripe file ==================================" +run_test 27a "one stripe file" test_27b() { - [[ $OSTCOUNT -lt 2 ]] && skip_env "skipping 2-stripe test" && return - test_mkdir -p $DIR/d27 - $SETSTRIPE -c 2 $DIR/d27/f01 || error "setstripe failed" - $GETSTRIPE -c $DIR/d27/f01 - [ $($GETSTRIPE -c $DIR/d27/f01) -eq 2 ] || + [[ $OSTCOUNT -lt 2 ]] && skip_env "needs >= 2 OSTs" && return + test_mkdir $DIR/$tdir + $LFS setstripe -c 2 $DIR/$tdir/$tfile || error "setstripe failed" + $LFS getstripe -c $DIR/$tdir/$tfile + [ $($LFS getstripe -c $DIR/$tdir/$tfile) -eq 2 ] || error "two-stripe file doesn't have two stripes" - dd if=/dev/zero of=$DIR/d27/f01 bs=4k count=4 || error "dd failed" + dd if=/dev/zero of=$DIR/$tdir/$tfile bs=4k count=4 || error "dd failed" } run_test 27b "create and write to two stripe file" test_27d() { - test_mkdir -p $DIR/d27 - $SETSTRIPE -c 0 -i -1 -S 0 $DIR/d27/fdef || error "setstripe failed" - $CHECKSTAT -t file $DIR/d27/fdef || error "checkstat failed" - dd if=/dev/zero of=$DIR/d27/fdef bs=4k count=4 || error + test_mkdir $DIR/$tdir + $LFS setstripe -c 0 -i -1 -S 0 $DIR/$tdir/$tfile || + error "setstripe failed" + $CHECKSTAT -t file $DIR/$tdir/$tfile || error "checkstat failed" + dd if=/dev/zero of=$DIR/$tdir/$tfile bs=4k count=4 || error "dd failed" } -run_test 27d "create file with default settings ================" +run_test 27d "create file with default settings" test_27e() { # LU-5839 adds check for existed layout before setting it [[ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.7.56) ]] && skip "Need MDS version at least 2.7.56" && return - test_mkdir -p $DIR/d27 - $SETSTRIPE -c 2 $DIR/d27/f12 || error "setstripe failed" - $SETSTRIPE -c 2 $DIR/d27/f12 && error "setstripe succeeded twice" - $CHECKSTAT -t file $DIR/d27/f12 || error "checkstat failed" + test_mkdir $DIR/$tdir + $LFS setstripe -c 2 $DIR/$tdir/$tfile || error "setstripe failed" + $LFS setstripe -c 2 $DIR/$tdir/$tfile && error "setstripe worked twice" + $CHECKSTAT -t file $DIR/$tdir/$tfile || error "checkstat failed" } -run_test 27e "setstripe existing file (should return error) ======" +run_test 27e "setstripe existing file (should return error)" test_27f() { test_mkdir $DIR/$tdir - $SETSTRIPE -S 100 -i 0 -c 1 $DIR/$tdir/$tfile && + $LFS setstripe -S 100 -i 0 -c 1 $DIR/$tdir/$tfile && error "$SETSTRIPE $DIR/$tdir/$tfile failed" $CHECKSTAT -t file $DIR/$tdir/$tfile && error "$CHECKSTAT -t file $DIR/$tdir/$tfile should fail" dd if=/dev/zero of=$DIR/$tdir/$tfile bs=4k count=4 || error "dd failed" - $GETSTRIPE $DIR/$tdir/$tfile || error "$GETSTRIPE failed" + $LFS getstripe $DIR/$tdir/$tfile || error "$LFS getstripe failed" } run_test 27f "setstripe with bad stripe size (should return error)" test_27g() { - test_mkdir -p $DIR/d27 - $MCREATE $DIR/d27/fnone || error "mcreate failed" - $GETSTRIPE $DIR/d27/fnone 2>&1 | grep "no stripe info" || - error "$DIR/d27/fnone has object" + test_mkdir $DIR/$tdir + $MCREATE $DIR/$tdir/$tfile || error "mcreate failed" + $LFS getstripe $DIR/$tdir/$tfile 2>&1 | grep "no stripe info" || + error "$DIR/$tdir/$tfile has object" } -run_test 27g "$GETSTRIPE with no objects" +run_test 27g "$LFS getstripe with no objects" test_27i() { test_mkdir $DIR/$tdir touch $DIR/$tdir/$tfile || error "touch failed" - [[ $($GETSTRIPE -c $DIR/$tdir/$tfile) -gt 0 ]] || + [[ $($LFS getstripe -c $DIR/$tdir/$tfile) -gt 0 ]] || error "missing objects" } -run_test 27i "$GETSTRIPE with some objects" +run_test 27i "$LFS getstripe with some objects" test_27j() { - test_mkdir -p $DIR/d27 - $SETSTRIPE -i $OSTCOUNT $DIR/d27/f27j && error "setstripe failed"||true + test_mkdir $DIR/$tdir + $LFS setstripe -i $OSTCOUNT $DIR/$tdir/$tfile && + error "setstripe failed" || true } run_test 27j "setstripe with bad stripe offset (should return error)" test_27k() { # bug 2844 - test_mkdir -p $DIR/d27 - FILE=$DIR/d27/f27k - LL_MAX_BLKSIZE=$((4 * 1024 * 1024)) - [ ! -d $DIR/d27 ] && test_mkdir -p $DIR d27 - $SETSTRIPE -S 67108864 $FILE || error "setstripe failed" - BLKSIZE=`stat $FILE | awk '/IO Block:/ { print $7 }'` - [ $BLKSIZE -le $LL_MAX_BLKSIZE ] || error "1:$BLKSIZE > $LL_MAX_BLKSIZE" - dd if=/dev/zero of=$FILE bs=4k count=1 - BLKSIZE=`stat $FILE | awk '/IO Block:/ { print $7 }'` - [ $BLKSIZE -le $LL_MAX_BLKSIZE ] || error "2:$BLKSIZE > $LL_MAX_BLKSIZE" -} -run_test 27k "limit i_blksize for broken user apps =============" + test_mkdir $DIR/$tdir + local file=$DIR/$tdir/$tfile + local ll_max_blksize=$((4 * 1024 * 1024)) + $LFS setstripe -S 67108864 $file || error "setstripe failed" + local blksize=$(stat $file | awk '/IO Block:/ { print $7 }') + [ $blksize -le $ll_max_blksize ] || error "1:$blksize > $ll_max_blksize" + dd if=/dev/zero of=$file bs=4k count=1 + blksize=$(stat $file | awk '/IO Block:/ { print $7 }') + [ $blksize -le $ll_max_blksize ] || error "2:$blksize > $ll_max_blksize" +} +run_test 27k "limit i_blksize for broken user apps" test_27l() { - test_mkdir -p $DIR/d27 - mcreate $DIR/f27l || error "creating file" - $RUNAS $SETSTRIPE -c 1 $DIR/f27l && \ + mcreate $DIR/$tfile || error "creating file" + $RUNAS $LFS setstripe -c 1 $DIR/$tfile && error "setstripe should have failed" || true } run_test 27l "check setstripe permissions (should return error)" test_27m() { - [[ $OSTCOUNT -lt 2 ]] && skip_env "$OSTCOUNT < 2 OSTs -- skipping" && - return + [[ $OSTCOUNT -lt 2 ]] && skip_env "needs >= 2 OSTs" && return ORIGFREE=$($LCTL get_param -n lov.$FSNAME-clilov-*.kbytesavail | head -n1) @@ -1456,30 +1455,33 @@ test_27m() { return fi trap simple_cleanup_common EXIT - test_mkdir -p $DIR/$tdir - $SETSTRIPE -i 0 -c 1 $DIR/$tdir/f27m_1 - dd if=/dev/zero of=$DIR/$tdir/f27m_1 bs=1024 count=$MAXFREE && + test_mkdir $DIR/$tdir + $LFS setstripe -i 0 -c 1 $DIR/$tdir/$tfile.1 + dd if=/dev/zero of=$DIR/$tdir/$tfile.1 bs=1024 count=$MAXFREE && error "dd should fill OST0" i=2 - while $SETSTRIPE -i 0 -c 1 $DIR/$tdir/f27m_$i; do + while $LFS setstripe -i 0 -c 1 $DIR/$tdir/$tfile.$i; do i=$((i + 1)) [ $i -gt 256 ] && break done i=$((i + 1)) - touch $DIR/$tdir/f27m_$i - [ `$GETSTRIPE $DIR/$tdir/f27m_$i | grep -A 10 obdidx | awk '{print $1}'| grep -w "0"` ] && + touch $DIR/$tdir/$tfile.$i + [ $($LFS getstripe $DIR/$tdir/$tfile.$i | grep -A 10 obdidx | + awk '{print $1}'| grep -w "0") ] && error "OST0 was full but new created file still use it" i=$((i + 1)) - touch $DIR/$tdir/f27m_$i - [ `$GETSTRIPE $DIR/$tdir/f27m_$i | grep -A 10 obdidx | awk '{print $1}'| grep -w "0"` ] && + touch $DIR/$tdir/$tfile.$i + [ $($LFS getstripe $DIR/$tdir/$tfile.$i | grep -A 10 obdidx | + awk '{print $1}'| grep -w "0") ] && error "OST0 was full but new created file still use it" simple_cleanup_common } -run_test 27m "create file while OST0 was full ==================" +run_test 27m "create file while OST0 was full" sleep_maxage() { - local DELAY=$(do_facet $SINGLEMDS lctl get_param -n lov.*.qos_maxage | head -n 1 | awk '{print $1 * 2}') - sleep $DELAY + local delay=$(do_facet $SINGLEMDS lctl get_param -n lov.*.qos_maxage | + head -n 1 | awk '{ print $1 * 2 }') + sleep $delay } # OSCs keep a NOSPC flag that will be reset after ~5s (qos_maxage) @@ -1537,7 +1539,7 @@ exhaust_all_precreations() { } test_27n() { - [[ $OSTCOUNT -lt 2 ]] && skip_env "too few OSTs" && return + [[ $OSTCOUNT -lt 2 ]] && skip_env "needs >= 2 OSTs" && return [ $PARALLEL == "yes" ] && skip "skip parallel run" && return remote_mds_nodsh && skip "remote MDS with nodsh" && return remote_ost_nodsh && skip "remote OST with nodsh" && return @@ -1545,15 +1547,15 @@ test_27n() { reset_enospc rm -f $DIR/$tdir/$tfile exhaust_precreations 0 0x80000215 - $SETSTRIPE -c -1 $DIR/$tdir + $LFS setstripe -c -1 $DIR/$tdir touch $DIR/$tdir/$tfile || error - $GETSTRIPE $DIR/$tdir/$tfile + $LFS getstripe $DIR/$tdir/$tfile reset_enospc } -run_test 27n "create file with some full OSTs ==================" +run_test 27n "create file with some full OSTs" test_27o() { - [[ $OSTCOUNT -lt 2 ]] && skip_env "too few OSTs" && return + [[ $OSTCOUNT -lt 2 ]] && skip_env "needs >= 2 OSTs" && return [ $PARALLEL == "yes" ] && skip "skip parallel run" && return remote_mds_nodsh && skip "remote MDS with nodsh" && return remote_ost_nodsh && skip "remote OST with nodsh" && return @@ -1567,17 +1569,17 @@ test_27o() { reset_enospc rm -rf $DIR/$tdir/* } -run_test 27o "create file with all full OSTs (should error) ====" +run_test 27o "create file with all full OSTs (should error)" test_27p() { - [[ $OSTCOUNT -lt 2 ]] && skip_env "too few OSTs" && return + [[ $OSTCOUNT -lt 2 ]] && skip_env "needs >= 2 OSTs" && return [ $PARALLEL == "yes" ] && skip "skip parallel run" && return remote_mds_nodsh && skip "remote MDS with nodsh" && return remote_ost_nodsh && skip "remote OST with nodsh" && return reset_enospc rm -f $DIR/$tdir/$tfile - test_mkdir -p $DIR/$tdir + test_mkdir $DIR/$tdir $MCREATE $DIR/$tdir/$tfile || error "mcreate failed" $TRUNCATE $DIR/$tdir/$tfile 80000000 || error "truncate failed" @@ -1586,14 +1588,14 @@ test_27p() { exhaust_precreations 0 0x80000215 echo foo >> $DIR/$tdir/$tfile || error "append failed" $CHECKSTAT -s 80000004 $DIR/$tdir/$tfile || error "checkstat failed" - $GETSTRIPE $DIR/$tdir/$tfile + $LFS getstripe $DIR/$tdir/$tfile reset_enospc } -run_test 27p "append to a truncated file with some full OSTs ===" +run_test 27p "append to a truncated file with some full OSTs" test_27q() { - [[ $OSTCOUNT -lt 2 ]] && skip_env "too few OSTs" && return + [[ $OSTCOUNT -lt 2 ]] && skip_env "needs >= 2 OSTs" && return [ $PARALLEL == "yes" ] && skip "skip parallel run" && return remote_mds_nodsh && skip "remote MDS with nodsh" && return remote_ost_nodsh && skip "remote OST with nodsh" && return @@ -1601,9 +1603,10 @@ test_27q() { reset_enospc rm -f $DIR/$tdir/$tfile - test_mkdir -p $DIR/$tdir + test_mkdir $DIR/$tdir $MCREATE $DIR/$tdir/$tfile || error "mcreate $DIR/$tdir/$tfile failed" - $TRUNCATE $DIR/$tdir/$tfile 80000000 ||error "truncate $DIR/$tdir/$tfile failed" + $TRUNCATE $DIR/$tdir/$tfile 80000000 || + error "truncate $DIR/$tdir/$tfile failed" $CHECKSTAT -s 80000000 $DIR/$tdir/$tfile || error "checkstat failed" exhaust_all_precreations 0x215 @@ -1613,10 +1616,10 @@ test_27q() { reset_enospc } -run_test 27q "append to truncated file with all OSTs full (should error) ===" +run_test 27q "append to truncated file with all OSTs full (should error)" test_27r() { - [[ $OSTCOUNT -lt 2 ]] && skip_env "too few OSTs" && return + [[ $OSTCOUNT -lt 2 ]] && skip_env "needs >= 2 OSTs" && return [ $PARALLEL == "yes" ] && skip "skip parallel run" && return remote_mds_nodsh && skip "remote MDS with nodsh" && return remote_ost_nodsh && skip "remote OST with nodsh" && return @@ -1625,18 +1628,18 @@ test_27r() { rm -f $DIR/$tdir/$tfile exhaust_precreations 0 0x80000215 - $SETSTRIPE -i 0 -c 2 $DIR/$tdir/$tfile # && error + $LFS setstripe -i 0 -c 2 $DIR/$tdir/$tfile # && error reset_enospc } run_test 27r "stripe file with some full OSTs (shouldn't LBUG) =" test_27s() { # bug 10725 - test_mkdir -p $DIR/$tdir + test_mkdir $DIR/$tdir local stripe_size=$((4096 * 1024 * 1024)) # 2^32 local stripe_count=0 [ $OSTCOUNT -eq 1 ] || stripe_count=2 - $SETSTRIPE -S $stripe_size -c $stripe_count $DIR/$tdir && + $LFS setstripe -S $stripe_size -c $stripe_count $DIR/$tdir && error "stripe width >= 2^32 succeeded" || true } @@ -1653,7 +1656,7 @@ test_27t() { # bug 10864 run_test 27t "check that utils parse path correctly" test_27u() { # bug 4900 - [[ $OSTCOUNT -lt 2 ]] && skip_env "too few OSTs" && return + [[ $OSTCOUNT -lt 2 ]] && skip_env "needs >= 2 OSTs" && return remote_mds_nodsh && skip "remote MDS with nodsh" && return local index local list=$(comma_list $(mdts_nodes)) @@ -1661,30 +1664,30 @@ test_27u() { # bug 4900 #define OBD_FAIL_MDS_OSC_PRECREATE 0x139 do_nodes $list $LCTL set_param fail_loc=0x139 test_mkdir -p $DIR/$tdir - rm -rf $DIR/$tdir/* + trap simple_cleanup_common EXIT createmany -o $DIR/$tdir/t- 1000 do_nodes $list $LCTL set_param fail_loc=0 TLOG=$TMP/$tfile.getstripe - $GETSTRIPE $DIR/$tdir > $TLOG + $LFS getstripe $DIR/$tdir > $TLOG OBJS=$(awk -vobj=0 '($1 == 0) { obj += 1 } END { print obj; }' $TLOG) unlinkmany $DIR/$tdir/t- 1000 + trap 0 [[ $OBJS -gt 0 ]] && error "$OBJS objects created on OST-0. See $TLOG" || pass } -run_test 27u "skip object creation on OSC w/o objects ==========" +run_test 27u "skip object creation on OSC w/o objects" test_27v() { # bug 4900 - [[ $OSTCOUNT -lt 2 ]] && skip_env "too few OSTs" && return + [[ $OSTCOUNT -lt 2 ]] && skip_env "needs >= 2 OSTs" && return [ $PARALLEL == "yes" ] && skip "skip parallel run" && return - remote_mds_nodsh && skip "remote MDS with nodsh" && return - remote_ost_nodsh && skip "remote OST with nodsh" && return + remote_mds_nodsh && skip "remote MDS with nodsh" && return + remote_ost_nodsh && skip "remote OST with nodsh" && return - exhaust_all_precreations 0x215 - reset_enospc + exhaust_all_precreations 0x215 + reset_enospc - test_mkdir -p $DIR/$tdir - $SETSTRIPE -c 1 $DIR/$tdir # 1 stripe / file + $LFS setstripe -c 1 $DIR/$tdir # 1 stripe / file touch $DIR/$tdir/$tfile #define OBD_FAIL_TGT_DELAY_PRECREATE 0x705 @@ -1703,51 +1706,52 @@ test_27v() { # bug 4900 sleep $((TIMEOUT / 2 - PROCESS)) reset_enospc } -run_test 27v "skip object creation on slow OST =================" +run_test 27v "skip object creation on slow OST" test_27w() { # bug 10997 - test_mkdir -p $DIR/$tdir || error "mkdir failed" - $SETSTRIPE -S 65536 $DIR/$tdir/f0 || error "setstripe failed" - [ $($GETSTRIPE -S $DIR/$tdir/f0) -ne 65536 ] && - error "stripe size $size != 65536" || true - [ $($GETSTRIPE -d $DIR/$tdir | grep -c "stripe_count") -ne 1 ] && - error "$GETSTRIPE -d $DIR/$tdir failed" || true + test_mkdir $DIR/$tdir + $LFS setstripe -S 65536 $DIR/$tdir/f0 || error "setstripe failed" + [ $($LFS getstripe -S $DIR/$tdir/f0) -ne 65536 ] && + error "stripe size $size != 65536" || true + [ $($LFS getstripe -d $DIR/$tdir | grep -c "stripe_count") -eq 0 ] && + error "$LFS getstripe -d $DIR/$tdir no 'stripe_count'" || true } -run_test 27w "check $SETSTRIPE -S option" +run_test 27w "check $LFS setstripe -S and getstrip -d options" test_27wa() { [[ $OSTCOUNT -lt 2 ]] && skip_env "skipping multiple stripe count/offset test" && return - test_mkdir -p $DIR/$tdir || error "mkdir failed" - for i in $(seq 1 $OSTCOUNT); do - offset=$((i - 1)) - $SETSTRIPE -c $i -i $offset $DIR/$tdir/f$i || - error "setstripe -c $i -i $offset failed" - count=$($GETSTRIPE -c $DIR/$tdir/f$i) - index=$($GETSTRIPE -i $DIR/$tdir/f$i) - [ $count -ne $i ] && error "stripe count $count != $i" || true - [ $index -ne $offset ] && - error "stripe offset $index != $offset" || true - done + test_mkdir $DIR/$tdir + for i in $(seq 1 $OSTCOUNT); do + offset=$((i - 1)) + $LFS setstripe -c $i -i $offset $DIR/$tdir/f$i || + error "setstripe -c $i -i $offset failed" + count=$($LFS getstripe -c $DIR/$tdir/f$i) + index=$($LFS getstripe -i $DIR/$tdir/f$i) + [ $count -ne $i ] && error "stripe count $count != $i" || true + [ $index -ne $offset ] && + error "stripe offset $index != $offset" || true + done } -run_test 27wa "check $SETSTRIPE -c -i options" +run_test 27wa "check $LFS setstripe -c -i options" test_27x() { remote_ost_nodsh && skip "remote OST with nodsh" && return - [[ $OSTCOUNT -lt 2 ]] && skip_env "$OSTCOUNT < 2 OSTs" && return + [[ $OSTCOUNT -lt 2 ]] && skip_env "needs >= 2 OSTs" && return [ $PARALLEL == "yes" ] && skip "skip parallel run" && return OFFSET=$(($OSTCOUNT - 1)) OSTIDX=0 local OST=$(ostname_from_index $OSTIDX) - test_mkdir -p $DIR/$tdir - $SETSTRIPE -c 1 $DIR/$tdir # 1 stripe per file + test_mkdir $DIR/$tdir + $LFS setstripe -c 1 $DIR/$tdir # 1 stripe per file do_facet ost$((OSTIDX + 1)) lctl set_param -n obdfilter.$OST.degraded 1 sleep_maxage createmany -o $DIR/$tdir/$tfile $OSTCOUNT - for i in `seq 0 $OFFSET`; do - [ `$GETSTRIPE $DIR/$tdir/$tfile$i | grep -A 10 obdidx | awk '{print $1}' | grep -w "$OSTIDX"` ] && + for i in $(seq 0 $OFFSET); do + [ $($LFS getstripe $DIR/$tdir/$tfile$i | grep -A 10 obdidx | + awk '{print $1}' | grep -w "$OSTIDX") ] && error "OST0 was degraded but new created file still use it" done do_facet ost$((OSTIDX + 1)) lctl set_param -n obdfilter.$OST.degraded 0 @@ -1755,18 +1759,17 @@ test_27x() { run_test 27x "create files while OST0 is degraded" test_27y() { - [[ $OSTCOUNT -lt 2 ]] && - skip_env "$OSTCOUNT < 2 OSTs -- skipping" && return - remote_mds_nodsh && skip "remote MDS with nodsh" && return - remote_ost_nodsh && skip "remote OST with nodsh" && return + [[ $OSTCOUNT -lt 2 ]] && skip_env "needs >= 2 OSTs" && return + remote_mds_nodsh && skip "remote MDS with nodsh" && return + remote_ost_nodsh && skip "remote OST with nodsh" && return [ $PARALLEL == "yes" ] && skip "skip parallel run" && return - local mdtosc=$(get_mdtosc_proc_path $SINGLEMDS $FSNAME-OST0000) - local last_id=$(do_facet $SINGLEMDS lctl get_param -n \ - osc.$mdtosc.prealloc_last_id) - local next_id=$(do_facet $SINGLEMDS lctl get_param -n \ - osc.$mdtosc.prealloc_next_id) - local fcount=$((last_id - next_id)) + local mdtosc=$(get_mdtosc_proc_path $SINGLEMDS $FSNAME-OST0000) + local last_id=$(do_facet $SINGLEMDS lctl get_param -n \ + osc.$mdtosc.prealloc_last_id) + local next_id=$(do_facet $SINGLEMDS lctl get_param -n \ + osc.$mdtosc.prealloc_next_id) + local fcount=$((last_id - next_id)) [[ $fcount -eq 0 ]] && skip "not enough space on OST0" && return [[ $fcount -gt $OSTCOUNT ]] && fcount=$OSTCOUNT @@ -1790,8 +1793,8 @@ test_27y() { done OSTIDX=$(index_from_ostuuid $OST) - test_mkdir -p $DIR/$tdir - $SETSTRIPE -c 1 $DIR/$tdir # 1 stripe / file + test_mkdir $DIR/$tdir + $LFS setstripe -c 1 $DIR/$tdir # 1 stripe / file for OSC in $MDS_OSCS; do OST=$(osc_to_ost $OSC) @@ -1819,16 +1822,16 @@ test_27y() { done # all osp devices get activated, hence -1 stripe count restored - local stripecnt=0 + local stripe_count=0 # sleep 2*lod_qos_maxage seconds waiting for lod qos to notice osp # devices get activated. sleep_maxage - $SETSTRIPE -c -1 $DIR/$tfile - stripecnt=$($GETSTRIPE -c $DIR/$tfile) + $LFS setstripe -c -1 $DIR/$tfile + stripe_count=$($LFS getstripe -c $DIR/$tfile) rm -f $DIR/$tfile - [ $stripecnt -ne $OSTCOUNT ] && - error "Of $OSTCOUNT OSTs, only $stripecnt is available" + [ $stripe_count -ne $OSTCOUNT ] && + error "Of $OSTCOUNT OSTs, only $stripe_count is available" return 0 } run_test 27y "create files while OST0 is degraded and the rest inactive" @@ -1877,17 +1880,18 @@ check_seq_oid() fi local obj_file="O/$seq/d$((oid %32))/$oid_hex" - local ff + local ff="" # # Don't unmount/remount the OSTs if we don't need to do that. # LU-2577 changes filter_fid to be smaller, so debugfs needs # update too, until that use mount/ll_decode_filter_fid/mount. # Re-enable when debugfs will understand new filter_fid. # - if false && [ $(facet_fstype ost$ost) == ldiskfs ]; then + if [ $(facet_fstype ost$ost) == ldiskfs ]; then ff=$(do_facet ost$ost "$DEBUGFS -c -R 'stat $obj_file' \ $dev 2>/dev/null" | grep "parent=") - else + fi + if [ -z "$ff" ]; then stop ost$ost mount_fstype ost$ost ff=$(do_facet ost$ost $LL_DECODE_FILTER_FID \ @@ -1903,57 +1907,65 @@ check_seq_oid() # /mnt/O/0/d23/23: objid=23 seq=0 parent=[0x200000400:0x1e:0x1] # fid: objid=23 seq=0 parent=[0x200000400:0x1e:0x0] stripe=1 - local ff_parent=$(echo $ff|sed -e 's/.*parent=.//') - local ff_pseq=$(echo $ff_parent | cut -d: -f1) - local ff_poid=$(echo $ff_parent | cut -d: -f2) + # + # fid: parent=[0x200000400:0x1e:0x0] stripe=1 stripe_count=2 \ + # stripe_size=1048576 component_id=1 component_start=0 \ + # component_end=33554432 + local ff_parent=$(sed -e 's/.*parent=.//' <<<$ff) + local ff_pseq=$(cut -d: -f1 <<<$ff_parent) + local ff_poid=$(cut -d: -f2 <<<$ff_parent) local ff_pstripe - if echo $ff_parent | grep -q 'stripe='; then - ff_pstripe=$(echo $ff_parent | sed -e 's/.*stripe=//') + if grep -q 'stripe=' <<<$ff; then + ff_pstripe=$(sed -e 's/.*stripe=//' -e 's/ .*//' <<<$ff) else - # # $LL_DECODE_FILTER_FID does not print "stripe="; look - # into f_ver in this case. See the comment on - # ff_parent. - # - ff_pstripe=$(echo $ff_parent | cut -d: -f3 | - sed -e 's/\]//') + # into f_ver in this case. See comment on ff_parent. + ff_pstripe=$(cut -d: -f3 <<<$ff_parent | sed -e 's/]//') fi - # compare lmm_seq and filter_fid->ff_parent.f_seq - [ $ff_pseq = $lmm_seq ] || - error "FF parent SEQ $ff_pseq != $lmm_seq" - # compare lmm_object_id and filter_fid->ff_parent.f_oid - [ $ff_poid = $lmm_oid ] || - error "FF parent OID $ff_poid != $lmm_oid" + # compare lmm_seq and filter_fid->ff_parent.f_seq + [ $ff_pseq = $lmm_seq ] || + error "FF parent SEQ $ff_pseq != $lmm_seq" + # compare lmm_object_id and filter_fid->ff_parent.f_oid + [ $ff_poid = $lmm_oid ] || + error "FF parent OID $ff_poid != $lmm_oid" (($ff_pstripe == $stripe_nr)) || - error "FF stripe $ff_pstripe != $stripe_nr" + error "FF stripe $ff_pstripe != $stripe_nr" - stripe_nr=$((stripe_nr + 1)) - done + stripe_nr=$((stripe_nr + 1)) + [ $(lustre_version_code client) -lt $(version_code 2.9.55) ] && + continue + if grep -q 'stripe_count=' <<<$ff; then + local ff_scnt=$(sed -e 's/.*stripe_count=//' \ + -e 's/ .*//' <<<$ff) + [ $lmm_count = $ff_scnt ] || + error "FF stripe count $lmm_count != $ff_scnt" + fi + done } test_27z() { - remote_ost_nodsh && skip "remote OST with nodsh" && return + remote_ost_nodsh && skip "remote OST with nodsh" && return [ $PARALLEL == "yes" ] && skip "skip parallel run" && return - test_mkdir -p $DIR/$tdir + test_mkdir $DIR/$tdir - $SETSTRIPE -c 1 -i 0 -S 64k $DIR/$tdir/$tfile-1 || - { error "setstripe -c -1 failed"; return 1; } - # We need to send a write to every object to get parent FID info set. - # This _should_ also work for setattr, but does not currently. - # touch $DIR/$tdir/$tfile-1 || - dd if=/dev/zero of=$DIR/$tdir/$tfile-1 bs=1M count=1 || - { error "dd $tfile-1 failed"; return 2; } - $SETSTRIPE -c -1 -i $((OSTCOUNT - 1)) -S 1M $DIR/$tdir/$tfile-2 || - { error "setstripe -c -1 failed"; return 3; } - dd if=/dev/zero of=$DIR/$tdir/$tfile-2 bs=1M count=$OSTCOUNT || - { error "dd $tfile-2 failed"; return 4; } + $LFS setstripe -c 1 -i 0 -S 64k $DIR/$tdir/$tfile-1 || + { error "setstripe -c -1 failed"; return 1; } + # We need to send a write to every object to get parent FID info set. + # This _should_ also work for setattr, but does not currently. + # touch $DIR/$tdir/$tfile-1 || + dd if=/dev/zero of=$DIR/$tdir/$tfile-1 bs=1M count=1 || + { error "dd $tfile-1 failed"; return 2; } + $LFS setstripe -c -1 -i $((OSTCOUNT - 1)) -S 1M $DIR/$tdir/$tfile-2 || + { error "setstripe -c -1 failed"; return 3; } + dd if=/dev/zero of=$DIR/$tdir/$tfile-2 bs=1M count=$OSTCOUNT || + { error "dd $tfile-2 failed"; return 4; } - # make sure write RPCs have been sent to OSTs - sync; sleep 5; sync + # make sure write RPCs have been sent to OSTs + sync; sleep 5; sync - check_seq_oid $DIR/$tdir/$tfile-1 || return 5 - check_seq_oid $DIR/$tdir/$tfile-2 || return 6 + check_seq_oid $DIR/$tdir/$tfile-1 || return 5 + check_seq_oid $DIR/$tdir/$tfile-2 || return 6 } run_test 27z "check SEQ/OID on the MDT and OST filesystems" @@ -1976,7 +1988,7 @@ test_27A() { # b=19102 run_test 27A "check filesystem-wide default LOV EA values" test_27B() { # LU-2523 - test_mkdir -p $DIR/$tdir + test_mkdir $DIR/$tdir rm -f $DIR/$tdir/f0 $DIR/$tdir/f1 touch $DIR/$tdir/f0 # open f1 with O_LOV_DELAY_CREATE @@ -2007,7 +2019,7 @@ test_27C() { #LU-2871 local i local j - test_mkdir -p $DIR/$tdir + test_mkdir $DIR/$tdir cd $DIR/$tdir for i in $(seq 0 $((OSTCOUNT - 1))); do # set stripe across all OSTs starting from OST$i @@ -2047,17 +2059,28 @@ test_27D() { local ost_list=$(seq $first_ost $ost_step $last_ost) local ost_range="$first_ost $last_ost $ost_step" - test_mkdir -p $DIR/$tdir + if ! combined_mgs_mds ; then + mount_mgs_client + fi + + test_mkdir $DIR/$tdir pool_add $POOL || error "pool_add failed" pool_add_targets $POOL $ost_range || error "pool_add_targets failed" local skip27D - [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.7.64) ] && - skip27D = "-s 29" + [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.8.55) ] && + skip27D += "-s 29" + [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.9.55) -o \ + $(lustre_version_code client) -lt $(version_code 2.9.55) ] && + skip27D += "-s 30,31" llapi_layout_test -d$DIR/$tdir -p$POOL -o$OSTCOUNT $skip27D || error "llapi_layout_test failed" - cleanup_pools || error "cleanup_pools failed" + destroy_test_pools || error "destroy test pools failed" + + if ! combined_mgs_mds ; then + umount_mgs_client + fi } run_test 27D "validate llapi_layout API" @@ -2093,13 +2116,12 @@ test_27E() { run_test 27E "check that default extended attribute size properly increases" test_27F() { # LU-5346/LU-7975 - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return - [[ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.8.51) ]] && skip "Need MDS version at least 2.8.51" && return + remote_ost_nodsh && skip "remote OST with nodsh" && return - test_mkdir -p $DIR/$tdir + test_mkdir $DIR/$tdir rm -f $DIR/$tdir/f0 $SETSTRIPE -c 2 $DIR/$tdir @@ -2389,7 +2411,7 @@ link_one() { } test_31o() { # LU-2901 - test_mkdir -p $DIR/$tdir + test_mkdir $DIR/$tdir for LOOP in $(seq 100); do rm -f $DIR/$tdir/$tfile* for THREAD in $(seq 8); do @@ -2426,10 +2448,10 @@ run_test 31p "remove of open striped directory" cleanup_test32_mount() { local rc=0 trap 0 - local loopdev=$(losetup -a | grep $EXT2_DEV | sed -ne 's/:.*$/p') + local loopdev=$(losetup -a | grep $EXT2_DEV | sed -ne 's/:.*$//p') $UMOUNT $DIR/$tdir/ext2-mountpoint || rc=$? losetup -d $loopdev || true - rm -rf $DIR/$tdir/ext2-mountpoint + rm -rf $DIR/$tdir return $rc } @@ -2478,52 +2500,52 @@ test_32d() { ls -al $DIR/$tdir/ext2-mountpoint/../d2/test_dir || error cleanup_test32_mount } -run_test 32d "open d32d/ext2-mountpoint/../d2/test_dir =========" +run_test 32d "open d32d/ext2-mountpoint/../d2/test_dir" test_32e() { - [ -e $DIR/d32e ] && rm -fr $DIR/d32e - test_mkdir -p $DIR/d32e/tmp - TMP_DIR=$DIR/d32e/tmp - ln -s $DIR/d32e $TMP_DIR/symlink11 - ln -s $TMP_DIR/symlink11 $TMP_DIR/../symlink01 - $CHECKSTAT -t link $DIR/d32e/tmp/symlink11 || error - $CHECKSTAT -t link $DIR/d32e/symlink01 || error + rm -fr $DIR/$tdir + test_mkdir -p $DIR/$tdir/tmp + local tmp_dir=$DIR/$tdir/tmp + ln -s $DIR/$tdir $tmp_dir/symlink11 + ln -s $tmp_dir/symlink11 $tmp_dir/../symlink01 + $CHECKSTAT -t link $DIR/$tdir/tmp/symlink11 || error "symlink11 bad" + $CHECKSTAT -t link $DIR/$tdir/symlink01 || error "symlink01 bad" } -run_test 32e "stat d32e/symlink->tmp/symlink->lustre-subdir ====" +run_test 32e "stat d32e/symlink->tmp/symlink->lustre-subdir" test_32f() { - [ -e $DIR/d32f ] && rm -fr $DIR/d32f - test_mkdir -p $DIR/d32f/tmp - TMP_DIR=$DIR/d32f/tmp - ln -s $DIR/d32f $TMP_DIR/symlink11 - ln -s $TMP_DIR/symlink11 $TMP_DIR/../symlink01 - ls $DIR/d32f/tmp/symlink11 || error - ls $DIR/d32f/symlink01 || error + rm -fr $DIR/$tdir + test_mkdir -p $DIR/$tdir/tmp + local tmp_dir=$DIR/$tdir/tmp + ln -s $DIR/$tdir $tmp_dir/symlink11 + ln -s $tmp_dir/symlink11 $tmp_dir/../symlink01 + ls $DIR/$tdir/tmp/symlink11 || error "symlink11 bad" + ls $DIR/$tdir/symlink01 || error "symlink01 bad" } -run_test 32f "open d32f/symlink->tmp/symlink->lustre-subdir ====" +run_test 32f "open d32f/symlink->tmp/symlink->lustre-subdir" test_32g() { - TMP_DIR=$DIR/$tdir/tmp - test_mkdir -p $DIR/$tdir/tmp + local tmp_dir=$DIR/$tdir/tmp + test_mkdir -p $tmp_dir test_mkdir $DIR/${tdir}2 - ln -s $DIR/${tdir}2 $TMP_DIR/symlink12 - ln -s $TMP_DIR/symlink12 $TMP_DIR/../symlink02 - $CHECKSTAT -t link $TMP_DIR/symlink12 || error - $CHECKSTAT -t link $DIR/$tdir/symlink02 || error - $CHECKSTAT -t dir -f $TMP_DIR/symlink12 || error - $CHECKSTAT -t dir -f $DIR/$tdir/symlink02 || error + ln -s $DIR/${tdir}2 $tmp_dir/symlink12 + ln -s $tmp_dir/symlink12 $tmp_dir/../symlink02 + $CHECKSTAT -t link $tmp_dir/symlink12 || error "symlink12 not a link" + $CHECKSTAT -t link $DIR/$tdir/symlink02 || error "symlink02 not a link" + $CHECKSTAT -t dir -f $tmp_dir/symlink12 || error "symlink12 not a dir" + $CHECKSTAT -t dir -f $DIR/$tdir/symlink02 || error "symlink12 not a dir" } run_test 32g "stat d32g/symlink->tmp/symlink->lustre-subdir/${tdir}2" test_32h() { rm -fr $DIR/$tdir $DIR/${tdir}2 - TMP_DIR=$DIR/$tdir/tmp - test_mkdir -p $DIR/$tdir/tmp + tmp_dir=$DIR/$tdir/tmp + test_mkdir -p $tmp_dir test_mkdir $DIR/${tdir}2 - ln -s $DIR/${tdir}2 $TMP_DIR/symlink12 - ln -s $TMP_DIR/symlink12 $TMP_DIR/../symlink02 - ls $TMP_DIR/symlink12 || error - ls $DIR/$tdir/symlink02 || error + ln -s $DIR/${tdir}2 $tmp_dir/symlink12 + ln -s $tmp_dir/symlink12 $tmp_dir/../symlink02 + ls $tmp_dir/symlink12 || error "listing symlink12" + ls $DIR/$tdir/symlink02 || error "listing symlink02" } run_test 32h "open d32h/symlink->tmp/symlink->lustre-subdir/${tdir}2" @@ -2613,59 +2635,49 @@ test_32o() { run_test 32o "stat d32o/symlink->tmp/symlink->lustre-root/$tfile" test_32p() { - log 32p_1 + log 32p_1 rm -fr $DIR/d32p - log 32p_2 + log 32p_2 rm -f $DIR/$tfile - log 32p_3 + log 32p_3 touch $DIR/$tfile - log 32p_4 + log 32p_4 test_mkdir -p $DIR/d32p/tmp - log 32p_5 + log 32p_5 TMP_DIR=$DIR/d32p/tmp - log 32p_6 + log 32p_6 ln -s $DIR/$tfile $TMP_DIR/symlink12 - log 32p_7 + log 32p_7 ln -s $TMP_DIR/symlink12 $TMP_DIR/../symlink02 - log 32p_8 + log 32p_8 cat $DIR/d32p/tmp/symlink12 || error - log 32p_9 + log 32p_9 cat $DIR/d32p/symlink02 || error - log 32p_10 + log 32p_10 } run_test 32p "open d32p/symlink->tmp/symlink->lustre-root/$tfile" -cleanup_testdir_mount() { - local rc=0 - trap 0 - local loopdev=$(losetup -a | grep $EXT2_DEV | sed -ne 's/:.*$/p') - $UMOUNT $DIR/$tdir || rc=$? - losetup -d $loopdev || true - rm -rf $DIR/$tdir - return $rc -} - test_32q() { [ $PARALLEL == "yes" ] && skip "skip parallel run" && return [ -e $DIR/$tdir ] && rm -fr $DIR/$tdir - trap cleanup_testdir_mount EXIT - test_mkdir -p $DIR/$tdir - touch $DIR/$tdir/under_the_mount - mount -t ext2 -o loop $EXT2_DEV $DIR/$tdir - ls $DIR/$tdir | grep "\" && error - cleanup_testdir_mount + trap cleanup_test32_mount EXIT + test_mkdir -p $DIR/$tdir/ext2-mountpoint + touch $DIR/$tdir/ext2-mountpoint/under_the_mount + mount -t ext2 -o loop $EXT2_DEV $DIR/$tdir/ext2-mountpoint + ls $DIR/$tdir/ext2-mountpoint | grep "\" && error + cleanup_test32_mount } run_test 32q "stat follows mountpoints in Lustre (should return error)" test_32r() { [ $PARALLEL == "yes" ] && skip "skip parallel run" && return [ -e $DIR/$tdir ] && rm -fr $DIR/$tdir - trap cleanup_testdir_mount EXIT - test_mkdir -p $DIR/$tdir - touch $DIR/$tdir/under_the_mount - mount -t ext2 -o loop $EXT2_DEV $DIR/$tdir - ls $DIR/$tdir | grep -q under_the_mount && error || true - cleanup_testdir_mount + trap cleanup_test32_mount EXIT + test_mkdir -p $DIR/$tdir/ext2-mountpoint + touch $DIR/$tdir/ext2-mountpoint/under_the_mount + mount -t ext2 -o loop $EXT2_DEV $DIR/$tdir/ext2-mountpoint + ls $DIR/$tdir/ext2-mountpoint | grep -q under_the_mount && error || true + cleanup_test32_mount } run_test 32r "opendir follows mountpoints in Lustre (should return error)" @@ -2678,38 +2690,39 @@ test_33aa() { $RUNAS $OPENFILE -f O_RDWR $DIR/$tfile && error || true log 33_2 } -run_test 33aa "write file with mode 444 (should return error) ====" +run_test 33aa "write file with mode 444 (should return error)" test_33a() { - rm -fr $DIR/d33 - test_mkdir -p $DIR/d33 - chown $RUNAS_ID $DIR/d33 - $RUNAS $OPENFILE -f O_RDWR:O_CREAT -m 0444 $DIR/d33/f33|| error "create" - $RUNAS $OPENFILE -f O_RDWR:O_CREAT -m 0444 $DIR/d33/f33 && \ + rm -fr $DIR/$tdir + test_mkdir $DIR/$tdir + chown $RUNAS_ID $DIR/$tdir + $RUNAS $OPENFILE -f O_RDWR:O_CREAT -m 0444 $DIR/$tdir/$tfile || + error "$RUNAS create $tdir/$tfile failed" + $RUNAS $OPENFILE -f O_RDWR:O_CREAT -m 0444 $DIR/$tdir/$tfile && error "open RDWR" || true } run_test 33a "test open file(mode=0444) with O_RDWR (should return error)" test_33b() { - rm -fr $DIR/d33 - test_mkdir -p $DIR/d33 - chown $RUNAS_ID $DIR/d33 - $RUNAS $OPENFILE -f 1286739555 $DIR/d33/f33 || true + rm -fr $DIR/$tdir + test_mkdir $DIR/$tdir + chown $RUNAS_ID $DIR/$tdir + $RUNAS $OPENFILE -f 1286739555 $DIR/$tdir/$tfile || true } run_test 33b "test open file with malformed flags (No panic)" test_33c() { [ $PARALLEL == "yes" ] && skip "skip parallel run" && return - local ostnum - local ostname - local write_bytes - local all_zeros + local ostnum + local ostname + local write_bytes + local all_zeros - remote_ost_nodsh && skip "remote OST with nodsh" && return - all_zeros=: - rm -fr $DIR/d33 - test_mkdir -p $DIR/d33 - # Read: 0, Write: 4, create/destroy: 2/0, stat: 1, punch: 0 + remote_ost_nodsh && skip "remote OST with nodsh" && return + all_zeros=: + rm -fr $DIR/$tdir + test_mkdir $DIR/$tdir + # Read: 0, Write: 4, create/destroy: 2/0, stat: 1, punch: 0 sync for ostnum in $(seq $OSTCOUNT); do @@ -2732,10 +2745,10 @@ test_33c() { $all_zeros || return 0 - # Write four bytes - echo foo > $DIR/d33/bar - # Really write them - sync + # Write four bytes + echo foo > $DIR/$tdir/bar + # Really write them + sync # Total up write_bytes after writing. We'd better find non-zeros. for ostnum in $(seq $OSTCOUNT); do @@ -2770,7 +2783,7 @@ test_33d() { local MDTIDX=1 local remote_dir=$DIR/$tdir/remote_dir - test_mkdir -p $DIR/$tdir + test_mkdir $DIR/$tdir $LFS mkdir -i $MDTIDX $remote_dir || error "create remote directory failed" @@ -2989,19 +3002,19 @@ test_35a() { $RUNAS $DIR/f35a && error || true rm $DIR/f35a } -run_test 35a "exec file with mode 444 (should return and not leak) =====" +run_test 35a "exec file with mode 444 (should return and not leak)" test_36a() { rm -f $DIR/f36 utime $DIR/f36 || error } -run_test 36a "MDS utime check (mknod, utime) ===================" +run_test 36a "MDS utime check (mknod, utime)" test_36b() { echo "" > $DIR/f36 utime $DIR/f36 || error } -run_test 36b "OST utime check (open, utime) ====================" +run_test 36b "OST utime check (open, utime)" test_36c() { rm -f $DIR/d36/f36 @@ -3009,23 +3022,23 @@ test_36c() { chown $RUNAS_ID $DIR/d36 $RUNAS utime $DIR/d36/f36 || error } -run_test 36c "non-root MDS utime check (mknod, utime) ==========" +run_test 36c "non-root MDS utime check (mknod, utime)" test_36d() { [ ! -d $DIR/d36 ] && test_36c echo "" > $DIR/d36/f36 $RUNAS utime $DIR/d36/f36 || error } -run_test 36d "non-root OST utime check (open, utime) ===========" +run_test 36d "non-root OST utime check (open, utime)" test_36e() { [ $RUNAS_ID -eq $UID ] && skip_env "RUNAS_ID = UID = $UID -- skipping" && return - test_mkdir -p $DIR/$tdir + test_mkdir $DIR/$tdir touch $DIR/$tdir/$tfile - $RUNAS utime $DIR/$tdir/$tfile && \ + $RUNAS utime $DIR/$tdir/$tfile && error "utime worked, expected failure" || true } -run_test 36e "utime on non-owned file (should return error) ====" +run_test 36e "utime on non-owned file (should return error)" subr_36fh() { local fl="$1" @@ -3034,7 +3047,7 @@ subr_36fh() { export LANG=C LC_LANG=C # for date language DATESTR="Dec 20 2000" - test_mkdir -p $DIR/$tdir + test_mkdir $DIR/$tdir lctl set_param fail_loc=$fl date; date +%s cp /etc/hosts $DIR/$tdir/$tfile @@ -3068,7 +3081,7 @@ test_36g() { local fmd_before local fmd_after - test_mkdir -p $DIR/$tdir + test_mkdir $DIR/$tdir fmd_max_age=$(do_facet ost1 \ "lctl get_param -n obdfilter.*.client_cache_seconds 2> /dev/null | \ head -n 1") @@ -3145,10 +3158,10 @@ test_39a() { # was test_39 error "O_TRUNC didn't change timestamps" fi } -run_test 39a "mtime changed on create ===========================" +run_test 39a "mtime changed on create" test_39b() { - test_mkdir -p -c1 $DIR/$tdir + test_mkdir -c1 $DIR/$tdir cp -p /etc/passwd $DIR/$tdir/fopen cp -p /etc/passwd $DIR/$tdir/flink cp -p /etc/passwd $DIR/$tdir/funlink @@ -3529,19 +3542,19 @@ test_39o() { error "wrong links count $links1 != $links3" return 0 } -run_test 39o "directory cached attributes updated after create ========" +run_test 39o "directory cached attributes updated after create" test_39p() { [[ $MDSCOUNT -lt 2 ]] && skip "needs >= 2 MDTs" && return local MDTIDX=1 - TESTDIR=$DIR/$tdir/$tfile + TESTDIR=$DIR/$tdir/$tdir [ -e $TESTDIR ] && rm -rf $TESTDIR test_mkdir -p $TESTDIR cd $TESTDIR links1=2 ls - $LFS mkdir -i $MDTIDX $TESTDIR/remote_dir1 - $LFS mkdir -i $MDTIDX $TESTDIR/remote_dir2 + test_mkdir -i $MDTIDX $TESTDIR/remote_dir1 + test_mkdir -i $MDTIDX $TESTDIR/remote_dir2 ls links2=$(stat -c %h .) [ $(($links1 + 2)) != $links2 ] && @@ -3583,7 +3596,7 @@ test_41() { run_test 41 "test small file write + fstat =====================" count_ost_writes() { - lctl get_param -n osc.*.stats | + lctl get_param -n ${OSC}.*.stats | awk -vwrites=0 '/ost_write/ { writes += $2 } \ END { printf("%0.0f", writes) }' } @@ -3643,7 +3656,7 @@ setup_test42() { test_42a() { [ $PARALLEL == "yes" ] && skip "skip parallel run" && return setup_test42 - cancel_lru_locks osc + cancel_lru_locks $OSC stop_writeback sync; sleep 1; sync # just to be safe BEFOREWRITES=`count_ost_writes` @@ -3654,12 +3667,12 @@ test_42a() { error "$BEFOREWRITES < $AFTERWRITES" start_writeback } -run_test 42a "ensure that we don't flush on close ==============" +run_test 42a "ensure that we don't flush on close" test_42b() { [ $PARALLEL == "yes" ] && skip "skip parallel run" && return setup_test42 - cancel_lru_locks osc + cancel_lru_locks $OSC stop_writeback sync dd if=/dev/zero of=$DIR/f42b bs=1024 count=100 @@ -3695,21 +3708,21 @@ run_test 42b "test destroy of file with cached dirty data ======" # start the file with a full-file pw lock to match against # until the truncate. trunc_test() { - test=$1 - file=$DIR/$test - offset=$2 - cancel_lru_locks osc + test=$1 + file=$DIR/$test + offset=$2 + cancel_lru_locks $OSC stop_writeback # prime the file with 0,EOF PW to match touch $file $TRUNCATE $file 0 sync; sync # now the real test.. - dd if=/dev/zero of=$file bs=1024 count=100 - BEFOREWRITES=`count_ost_writes` - $TRUNCATE $file $offset - cancel_lru_locks osc - AFTERWRITES=`count_ost_writes` + dd if=/dev/zero of=$file bs=1024 count=100 + BEFOREWRITES=`count_ost_writes` + $TRUNCATE $file $offset + cancel_lru_locks $OSC + AFTERWRITES=`count_ost_writes` start_writeback } @@ -3741,7 +3754,7 @@ test_42e() { # bug22074 local max_dirty_mb local warmup_files - test_mkdir -p $DIR/${tdir}e + test_mkdir $DIR/${tdir}e $SETSTRIPE -c 1 $TDIR createmany -o $TDIR/f $files @@ -3808,63 +3821,60 @@ test_42e() { # bug22074 run_test 42e "verify sub-RPC writes are not done synchronously" test_43A() { # was test_43 - test_mkdir -p $DIR/$tdir + test_mkdir $DIR/$tdir cp -p /bin/ls $DIR/$tdir/$tfile $MULTIOP $DIR/$tdir/$tfile Ow_c & pid=$! # give multiop a chance to open sleep 1 - $DIR/$tdir/$tfile && error || true + $DIR/$tdir/$tfile && error "execute $DIR/$tdir/$tfile succeeded" || true kill -USR1 $pid } run_test 43A "execution of file opened for write should return -ETXTBSY" test_43a() { - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return - test_mkdir -p $DIR/$tdir - cp -p `which $MULTIOP` $DIR/$tdir/multiop || - cp -p multiop $DIR/$tdir/multiop + test_mkdir $DIR/$tdir + cp -p $(which $MULTIOP) $DIR/$tdir/multiop || + cp -p multiop $DIR/$tdir/multiop MULTIOP_PROG=$DIR/$tdir/multiop multiop_bg_pause $TMP/$tfile.junk O_c || - return 1 - MULTIOP_PID=$! - $MULTIOP $DIR/$tdir/multiop Oc && error "expected error, got success" - kill -USR1 $MULTIOP_PID || return 2 - wait $MULTIOP_PID || return 3 - rm $TMP/$tfile.junk $DIR/$tdir/multiop + error "multiop open $TMP/$tfile.junk failed" + MULTIOP_PID=$! + $MULTIOP $DIR/$tdir/multiop Oc && error "expected error, got success" + kill -USR1 $MULTIOP_PID || error "kill -USR1 PID $MULTIOP_PID failed" + wait $MULTIOP_PID || error "wait PID $MULTIOP_PID failed" } run_test 43a "open(RDWR) of file being executed should return -ETXTBSY" test_43b() { [ $PARALLEL == "yes" ] && skip "skip parallel run" && return - test_mkdir -p $DIR/$tdir - cp -p `which $MULTIOP` $DIR/$tdir/multiop || - cp -p multiop $DIR/$tdir/multiop + test_mkdir $DIR/$tdir + cp -p $(which $MULTIOP) $DIR/$tdir/multiop || + cp -p multiop $DIR/$tdir/multiop MULTIOP_PROG=$DIR/$tdir/multiop multiop_bg_pause $TMP/$tfile.junk O_c || - return 1 - MULTIOP_PID=$! - $TRUNCATE $DIR/$tdir/multiop 0 && error "expected error, got success" - kill -USR1 $MULTIOP_PID || return 2 - wait $MULTIOP_PID || return 3 - rm $TMP/$tfile.junk $DIR/$tdir/multiop + error "multiop open $TMP/$tfile.junk failed" + MULTIOP_PID=$! + $TRUNCATE $DIR/$tdir/multiop 0 && error "expected error, got success" + kill -USR1 $MULTIOP_PID || error "kill -USR1 PID $MULTIOP_PID failed" + wait $MULTIOP_PID || error "wait PID $MULTIOP_PID failed" } run_test 43b "truncate of file being executed should return -ETXTBSY" test_43c() { local testdir="$DIR/$tdir" - test_mkdir -p $DIR/$tdir + test_mkdir $testdir cp $SHELL $testdir/ - ( cd $(dirname $SHELL) && md5sum $(basename $SHELL) ) | \ - ( cd $testdir && md5sum -c) + ( cd $(dirname $SHELL) && md5sum $(basename $SHELL) ) | + ( cd $testdir && md5sum -c ) } -run_test 43c "md5sum of copy into lustre========================" +run_test 43c "md5sum of copy into lustre" test_44A() { # was test_44 - [[ $OSTCOUNT -lt 2 ]] && skip_env "skipping 2-stripe test" && return + [[ $OSTCOUNT -lt 2 ]] && skip_env "needs >= 2 OSTs" && return dd if=/dev/zero of=$DIR/f1 bs=4k count=1 seek=1023 dd if=$DIR/f1 bs=4k count=1 > /dev/null } -run_test 44A "zero length read from a sparse stripe =============" +run_test 44A "zero length read from a sparse stripe" test_44a() { local nstripe=$($LCTL lov_getconfig $DIR | grep default_stripe_count: | @@ -3911,7 +3921,7 @@ run_test 44a "test sparse pwrite ===============================" dirty_osc_total() { tot=0 - for d in `lctl get_param -n osc.*.cur_dirty_bytes`; do + for d in `lctl get_param -n ${OSC}.*.cur_dirty_bytes`; do tot=$(($tot + $d)) done echo $tot @@ -3975,15 +3985,15 @@ test_48a() { # bug 2399 test_mkdir $DIR/$tdir cd $DIR/$tdir mv $DIR/$tdir $DIR/$tdir.new || error "move directory failed" - test_mkdir $DIR/$tdir || error "recreate directory failed" + test_mkdir $DIR/$tdir touch foo || error "'touch foo' failed after recreating cwd" - test_mkdir bar || error "'mkdir foo' failed after recreating cwd" + test_mkdir bar touch .foo || error "'touch .foo' failed after recreating cwd" - test_mkdir .bar || error "'mkdir .foo' failed after recreating cwd" + test_mkdir .bar ls . > /dev/null || error "'ls .' failed after recreating cwd" ls .. > /dev/null || error "'ls ..' failed after removing cwd" cd . || error "'cd .' failed after recreating cwd" - test_mkdir . && error "'mkdir .' worked after recreating cwd" + mkdir . && error "'mkdir .' worked after recreating cwd" rmdir . && error "'rmdir .' worked after recreating cwd" ln -s . baz || error "'ln -s .' failed after recreating cwd" cd .. || error "'cd ..' failed after recreating cwd" @@ -3996,12 +4006,12 @@ test_48b() { # bug 2399 cd $DIR/$tdir rmdir $DIR/$tdir || error "remove cwd $DIR/$tdir failed" touch foo && error "'touch foo' worked after removing cwd" - test_mkdir foo && error "'mkdir foo' worked after removing cwd" + mkdir foo && error "'mkdir foo' worked after removing cwd" touch .foo && error "'touch .foo' worked after removing cwd" - test_mkdir .foo && error "'mkdir .foo' worked after removing cwd" + mkdir .foo && error "'mkdir .foo' worked after removing cwd" ls . > /dev/null && error "'ls .' worked after removing cwd" ls .. > /dev/null || error "'ls ..' failed after removing cwd" - test_mkdir . && error "'mkdir .' worked after removing cwd" + mkdir . && error "'mkdir .' worked after removing cwd" rmdir . && error "'rmdir .' worked after removing cwd" ln -s . foo && error "'ln -s .' worked after removing cwd" cd .. || echo "'cd ..' failed after removing cwd `pwd`" #bug 3517 @@ -4016,12 +4026,12 @@ test_48c() { # bug 2350 cd $DIR/$tdir/dir $TRACE rmdir $DIR/$tdir/dir || error "remove cwd $DIR/$tdir/dir failed" $TRACE touch foo && error "touch foo worked after removing cwd" - $TRACE test_mkdir foo && error "'mkdir foo' worked after removing cwd" + $TRACE mkdir foo && error "'mkdir foo' worked after removing cwd" touch .foo && error "touch .foo worked after removing cwd" - test_mkdir .foo && error "mkdir .foo worked after removing cwd" + mkdir .foo && error "mkdir .foo worked after removing cwd" $TRACE ls . && error "'ls .' worked after removing cwd" $TRACE ls .. || error "'ls ..' failed after removing cwd" - $TRACE test_mkdir . && error "'mkdir .' worked after removing cwd" + $TRACE mkdir . && error "'mkdir .' worked after removing cwd" $TRACE rmdir . && error "'rmdir .' worked after removing cwd" $TRACE ln -s . foo && error "'ln -s .' worked after removing cwd" $TRACE cd .. || echo "'cd ..' failed after removing cwd `pwd`" #bug 3415 @@ -4037,12 +4047,12 @@ test_48d() { # bug 2350 $TRACE rmdir $DIR/$tdir/dir || error "remove cwd $DIR/$tdir/dir failed" $TRACE rmdir $DIR/$tdir || error "remove parent $DIR/$tdir failed" $TRACE touch foo && error "'touch foo' worked after removing parent" - $TRACE test_mkdir foo && error "mkdir foo worked after removing parent" + $TRACE mkdir foo && error "mkdir foo worked after removing parent" touch .foo && error "'touch .foo' worked after removing parent" - test_mkdir .foo && error "mkdir .foo worked after removing parent" + mkdir .foo && error "mkdir .foo worked after removing parent" $TRACE ls . && error "'ls .' worked after removing parent" $TRACE ls .. && error "'ls ..' worked after removing parent" - $TRACE test_mkdir . && error "'mkdir .' worked after removing parent" + $TRACE mkdir . && error "'mkdir .' worked after removing parent" $TRACE rmdir . && error "'rmdir .' worked after removing parent" $TRACE ln -s . foo && error "'ln -s .' worked after removing parent" true @@ -4136,7 +4146,7 @@ test_51b() { # cleanup the directory rm -fr $dir - test_mkdir -p -c1 $dir + test_mkdir -c1 $dir $LFS df $LFS df -i @@ -4152,11 +4162,13 @@ test_51b() { [[ $numfree -lt $nrdirs ]] && skip "not enough blocks ($numfree)" && return - trap cleanup_print_lfsdf EXIT + trap cleanup_print_lfs_df EXIT # create files - createmany -d $dir/d $nrdirs || + createmany -d $dir/d $nrdirs || { + unlinkmany $dir/d $nrdirs error "failed to create $nrdirs subdirs in MDT$mdtidx:$dir" + } # really created : nrdirs=$(ls -U $dir | wc -l) @@ -4193,11 +4205,10 @@ run_test 51b "exceed 64k subdirectory nlink limit on create, verify unlink" test_51d() { [ $PARALLEL == "yes" ] && skip "skip parallel run" && return - [[ $OSTCOUNT -lt 3 ]] && - skip_env "skipping test with few OSTs" && return - test_mkdir -p $DIR/$tdir + [[ $OSTCOUNT -lt 3 ]] && skip_env "needs >= 3 OSTs" && return + test_mkdir $DIR/$tdir createmany -o $DIR/$tdir/t- 1000 - $GETSTRIPE $DIR/$tdir > $TMP/$tfile + $LFS getstripe $DIR/$tdir > $TMP/$tfile for N in $(seq 0 $((OSTCOUNT - 1))); do OBJS[$N]=$(awk -vobjs=0 '($1 == '$N') { objs += 1 } \ END { printf("%0.0f", objs) }' $TMP/$tfile) @@ -4231,12 +4242,12 @@ run_test 51d "check object distribution" test_51e() { if [ "$(facet_fstype $SINGLEMDS)" != ldiskfs ]; then - skip "Only applicable to ldiskfs-based MDTs" + skip "ldiskfs only test" return fi - test_mkdir -c1 $DIR/$tdir || error "create $tdir failed" - test_mkdir -c1 $DIR/$tdir/d0 || error "create d0 failed" + test_mkdir -c1 $DIR/$tdir + test_mkdir -c1 $DIR/$tdir/d0 touch $DIR/$tdir/d0/foo createmany -l $DIR/$tdir/d0/foo $DIR/$tdir/d0/f- 65001 && @@ -4246,9 +4257,40 @@ test_51e() { } run_test 51e "check file nlink limit" +test_51f() { + test_mkdir $DIR/$tdir + + local max=100000 + local ulimit_old=$(ulimit -n) + local spare=20 # number of spare fd's for scripts/libraries, etc. + local mdt=$(lfs getstripe -M $DIR/$tdir) + local numfree=$(lfs df -i $DIR/$tdir | awk '/MDT:'$mdt'/ { print $4 }') + + echo "MDT$mdt numfree=$numfree, max=$max" + [[ $numfree -gt $max ]] && numfree=$max || numfree=$((numfree * 7 / 8)) + if [ $((numfree + spare)) -gt $ulimit_old ]; then + while ! ulimit -n $((numfree + spare)); do + numfree=$((numfree * 3 / 4)) + done + echo "changed ulimit from $ulimit_old to $((numfree + spare))" + else + echo "left ulimit at $ulimit_old" + fi + + createmany -o -k -t 120 $DIR/$tdir/f $numfree || { + unlinkmany $DIR/$tdir/f $numfree + error "create+open $numfree files in $DIR/$tdir failed" + } + ulimit -n $ulimit_old + + # if createmany exits at 120s there will be fewer than $numfree files + unlinkmany $DIR/$tdir/f $numfree || true +} +run_test 51f "check many open files limit" + test_52a() { [ -f $DIR/$tdir/foo ] && chattr -a $DIR/$tdir/foo - test_mkdir -p $DIR/$tdir + test_mkdir $DIR/$tdir touch $DIR/$tdir/foo chattr +a $DIR/$tdir/foo || error "chattr +a failed" echo bar >> $DIR/$tdir/foo || error "append bar failed" @@ -4268,7 +4310,7 @@ run_test 52a "append-only flag test (should return errors)" test_52b() { [ -f $DIR/$tdir/foo ] && chattr -i $DIR/$tdir/foo - test_mkdir -p $DIR/$tdir + test_mkdir $DIR/$tdir touch $DIR/$tdir/foo chattr +i $DIR/$tdir/foo || error "chattr +i failed" cat test > $DIR/$tdir/foo && error "cat test worked" @@ -4355,6 +4397,8 @@ test_53() { run_test 53 "verify that MDS and OSTs agree on pre-creation ====" test_54a() { + perl -MSocket -e ';' || { skip "no Socket perl module installed" && return; } + $SOCKETSERVER $DIR/socket || error "$SOCKETSERVER $DIR/socket failed: $?" $SOCKETCLIENT $DIR/socket || @@ -4409,7 +4453,7 @@ test_54c() { losetup $loopdev $DIR/$tfile || error "can't set up $loopdev for $DIR/$tfile" mkfs.ext2 $loopdev || error "mke2fs on $loopdev" - test_mkdir -p $DIR/$tdir + test_mkdir $DIR/$tdir mount -t ext2 $loopdev $DIR/$tdir || error "error mounting $loopdev on $DIR/$tdir" dd if=/dev/zero of=$DIR/$tdir/tmp bs=$(get_page_size client) count=30 || @@ -4533,11 +4577,35 @@ test_56b() { } run_test 56b "check $LFS getdirstripe" -NUMFILES=3 -NUMDIRS=3 -setup_56() { - local LOCAL_NUMFILES="$1" - local LOCAL_NUMDIRS="$2" +test_56c() { + local ost_idx=0 + local ost_name=$(ostname_from_index $ost_idx) + + local old_status=$(ost_dev_status $ost_idx) + [[ -z "$old_status" ]] || + { skip_env "OST $ost_name is in $old_status status"; return 0; } + + do_facet ost1 $LCTL set_param -n obdfilter.$ost_name.degraded=1 + sleep_maxage + + local new_status=$(ost_dev_status $ost_idx) + [[ "$new_status" = "D" ]] || + error "OST $ost_name is in status of '$new_status', not 'D'" + + do_facet ost1 $LCTL set_param -n obdfilter.$ost_name.degraded=0 + sleep_maxage + + new_status=$(ost_dev_status $ost_idx) + [[ -z "$new_status" ]] || + error "OST $ost_name is in status of '$new_status', not ''" +} +run_test 56c "check 'lfs df' showing device status" + +NUMFILES=3 +NUMDIRS=3 +setup_56() { + local LOCAL_NUMFILES="$1" + local LOCAL_NUMDIRS="$2" local MKDIR_PARAMS="$3" local DIR_STRIPE_PARAMS="$4" @@ -4606,17 +4674,17 @@ test_56h() { "found $NUMS, expected $EXPECTED" done } -run_test 56h "check lfs find ! -name =============================" +run_test 56h "check lfs find ! -name" test_56i() { - tdir=${tdir}i - test_mkdir -p $DIR/$tdir - UUID=$(ostuuid_from_index 0 $DIR/$tdir) - CMD="$LFIND -ost $UUID $DIR/$tdir" - OUT=$($CMD) - [ -z "$OUT" ] || error "\"$CMD\" returned directory '$OUT'" + tdir=${tdir}i + test_mkdir $DIR/$tdir + UUID=$(ostuuid_from_index 0 $DIR/$tdir) + CMD="$LFIND -ost $UUID $DIR/$tdir" + OUT=$($CMD) + [ -z "$OUT" ] || error "'$CMD' returned directory '$OUT'" } -run_test 56i "check 'lfs find -ost UUID' skips directories =======" +run_test 56i "check 'lfs find -ost UUID' skips directories" test_56j() { TDIR=$DIR/${tdir}g @@ -4792,8 +4860,22 @@ run_test 56r "check lfs find -size works ==========================" test_56s() { # LU-611 TDIR=$DIR/${tdir}s - setup_56 $NUMFILES $NUMDIRS "-c $OSTCOUNT" + #LU-9369 + setup_56 0 $NUMDIRS + for i in $(seq 1 $NUMDIRS); do + $SETSTRIPE -c $((OSTCOUNT + 1)) $TDIR/dir$i/$tfile + done + EXPECTED=$NUMDIRS + CMD="$LFIND -c $OSTCOUNT $TDIR" + NUMS=$($CMD | wc -l) + [ $NUMS -eq $EXPECTED ] || { + $GETSTRIPE -R $TDIR + error "\"$CMD\" wrong: found $NUMS, expected $EXPECTED" + } + rm -rf $TDIR + + setup_56 $NUMFILES $NUMDIRS "-c $OSTCOUNT" if [[ $OSTCOUNT -gt 1 ]]; then $SETSTRIPE -c 1 $TDIR/$tfile.{0,1,2,3} ONESTRIPE=4 @@ -4846,6 +4928,21 @@ run_test 56s "check lfs find -stripe-count works" test_56t() { # LU-611 TDIR=$DIR/${tdir}t + + #LU-9369 + setup_56 0 $NUMDIRS + for i in $(seq 1 $NUMDIRS); do + $SETSTRIPE -S 4M $TDIR/dir$i/$tfile + done + EXPECTED=$NUMDIRS + CMD="$LFIND -S 4M $TDIR" + NUMS=$($CMD | wc -l) + [ $NUMS -eq $EXPECTED ] || { + $GETSTRIPE -R $TDIR + error "\"$CMD\" wrong: found $NUMS, expected $EXPECTED" + } + rm -rf $TDIR + setup_56 $NUMFILES $NUMDIRS "--stripe-size 512k" $SETSTRIPE -S 256k $TDIR/$tfile.{0,1,2,3} @@ -4955,59 +5052,63 @@ test_56v() { run_test 56v "check 'lfs find -mdt match with lfs getstripe -M' =======" test_56w() { - [[ $OSTCOUNT -lt 2 ]] && skip_env "$OSTCOUNT < 2 OSTs -- skipping" && - return + [[ $OSTCOUNT -lt 2 ]] && skip_env "needs >= 2 OSTs" && return [ $PARALLEL == "yes" ] && skip "skip parallel run" && return TDIR=$DIR/${tdir}w - rm -rf $TDIR || error "remove $TDIR failed" - setup_56 $NUMFILES $NUMDIRS "-c $OSTCOUNT" "-c1" - - local stripe_size - stripe_size=$($GETSTRIPE -S -d $TDIR) || - error "$GETSTRIPE -S -d $TDIR failed" - stripe_size=${stripe_size%% *} - - local file_size=$((stripe_size * OSTCOUNT)) - local file_num=$((NUMDIRS * NUMFILES + NUMFILES)) - local required_space=$((file_num * file_size)) - - local free_space=$($LCTL get_param -n lov.$FSNAME-clilov-*.kbytesavail | - head -n1) - [[ $free_space -le $((required_space / 1024)) ]] && - skip_env "need at least $required_space bytes free space," \ - "have $free_space kbytes" && return - - local dd_bs=65536 - local dd_count=$((file_size / dd_bs)) - - # write data into the files - local i - local j - local file - for i in $(seq 1 $NUMFILES); do - file=$TDIR/file$i - yes | dd bs=$dd_bs count=$dd_count of=$file >/dev/null 2>&1 || - error "write data into $file failed" - done - for i in $(seq 1 $NUMDIRS); do - for j in $(seq 1 $NUMFILES); do - file=$TDIR/dir$i/file$j - yes | dd bs=$dd_bs count=$dd_count of=$file \ - >/dev/null 2>&1 || - error "write data into $file failed" - done - done + rm -rf $TDIR || error "remove $TDIR failed" + setup_56 $NUMFILES $NUMDIRS "-c $OSTCOUNT" "-c1" - local expected=-1 - [[ $OSTCOUNT -gt 1 ]] && expected=$((OSTCOUNT - 1)) + local stripe_size + stripe_size=$($GETSTRIPE -S -d $TDIR) || + error "$GETSTRIPE -S -d $TDIR failed" + stripe_size=${stripe_size%% *} - # lfs_migrate file - local cmd="$LFS_MIGRATE -y -c $expected $TDIR/file1" - echo "$cmd" - eval $cmd || error "$cmd failed" + local file_size=$((stripe_size * OSTCOUNT)) + local file_num=$((NUMDIRS * NUMFILES + NUMFILES)) + local required_space=$((file_num * file_size)) + + local free_space=$($LCTL get_param -n lov.$FSNAME-clilov-*.kbytesavail | + head -n1) + [[ $free_space -le $((required_space / 1024)) ]] && + skip_env "need $required_space bytes, have $free_space KB" && + return + + local dd_bs=65536 + local dd_count=$((file_size / dd_bs)) + + # write data into the files + local i + local j + local file + for i in $(seq 1 $NUMFILES); do + file=$TDIR/file$i + yes | dd bs=$dd_bs count=$dd_count of=$file &>/dev/null || + error "write data into $file failed" + done + for i in $(seq 1 $NUMDIRS); do + for j in $(seq 1 $NUMFILES); do + file=$TDIR/dir$i/file$j + yes|dd bs=$dd_bs count=$dd_count of=$file &>/dev/null || + error "write data into $file failed" + done + done + + # $LFS_MIGRATE will fail if hard link migration is unsupported + if [[ $(lustre_version_code mds1) -gt $(version_code 2.5.55) ]]; then + createmany -l$TDIR/dir1/file1 $TDIR/dir1/link 200 || + error "creating links to $TDIR/dir1/file1 failed" + fi - check_stripe_count $TDIR/file1 $expected + local expected=-1 + [[ $OSTCOUNT -gt 1 ]] && expected=$((OSTCOUNT - 1)) + + # lfs_migrate file + local cmd="$LFS_MIGRATE -y -c $expected $TDIR/file1" + echo "$cmd" + eval $cmd || error "$cmd failed" + + check_stripe_count $TDIR/file1 $expected if [ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.6.90) ]; then @@ -5060,19 +5161,17 @@ run_test 56w "check lfs_migrate -c stripe_count works" test_56x() { check_swap_layouts_support && return 0 - [[ $OSTCOUNT -lt 2 ]] && - skip_env "need 2 OST, skipping test" && return - - local dir0=$DIR/$tdir/$testnum - test_mkdir -p $dir0 || error "creating dir $dir0" + [[ $OSTCOUNT -lt 2 ]] && skip_env "needs >= 2 OSTs" && return + local dir0=$DIR/$tdir local ref1=/etc/passwd local file1=$dir0/file1 - $SETSTRIPE -c 2 $file1 + test_mkdir $dir0 || error "creating dir $dir0" + $LFS setstripe -c 2 $file1 cp $ref1 $file1 $LFS migrate -c 1 $file1 || error "migrate failed rc = $?" - stripe=$($GETSTRIPE -c $file1) + stripe=$($LFS getstripe -c $file1) [[ $stripe == 1 ]] || error "stripe of $file1 is $stripe != 1" cmp $file1 $ref1 || error "content mismatch $file1 differs from $ref1" @@ -5083,19 +5182,18 @@ run_test 56x "lfs migration support" test_56xa() { check_swap_layouts_support && return 0 - [[ $OSTCOUNT -lt 2 ]] && - skip_env "need 2 OST, skipping test" && return + [[ $OSTCOUNT -lt 2 ]] && skip_env "needs >= 2 OSTs" && return local dir0=$DIR/$tdir/$testnum - test_mkdir -p $dir0 || error "creating dir $dir0" + test_mkdir -p $dir0 local ref1=/etc/passwd local file1=$dir0/file1 - $SETSTRIPE -c 2 $file1 + $LFS setstripe -c 2 $file1 cp $ref1 $file1 $LFS migrate --block -c 1 $file1 || error "migrate failed rc = $?" - local stripe=$($GETSTRIPE -c $file1) + local stripe=$($LFS getstripe -c $file1) [[ $stripe == 1 ]] || error "stripe of $file1 is $stripe != 1" cmp $file1 $ref1 || error "content mismatch $file1 differs from $ref1" @@ -5104,6 +5202,110 @@ test_56xa() { } run_test 56xa "lfs migration --block support" +check_migrate_links() { + local dir="$1" + local file1="$dir/file1" + local begin="$2" + local count="$3" + local total_count=$(($begin + $count - 1)) + local symlink_count=10 + local uniq_count=10 + + if [ ! -f "$file1" ]; then + echo -n "creating initial file..." + $LFS setstripe -c 1 -S "512k" "$file1" || + error "cannot setstripe initial file" + echo "done" + + echo -n "creating symlinks..." + for s in $(seq 1 $symlink_count); do + ln -s "$file1" "$dir/slink$s" || + error "cannot create symlinks" + done + echo "done" + + echo -n "creating nonlinked files..." + createmany -o "$dir/uniq" 1 10 &> /dev/null || + error "cannot create nonlinked files" + echo "done" + fi + + # create hard links + if [ ! -f "$dir/file$total_count" ]; then + echo -n "creating hard links $begin:$total_count..." + createmany -l"$file1" "$dir/file" "$begin" "$count" &> \ + /dev/null || error "cannot create hard links" + echo "done" + fi + + echo -n "checking number of hard links listed in xattrs..." + local fid=$($LFS getstripe -F "$file1") + local paths=($($LFS fid2path "$MOUNT" "$fid" 2> /dev/null)) + + echo "${#paths[*]}" + if [ ${#paths[*]} -lt $total_count -a "$begin" -eq 2 ]; then + echo "hard link list has unexpected size, skipping test" + return 0 + fi + if [ ${#paths[*]} -ge $total_count -a "$begin" -ne 2 ]; then + error "link names should exceed xattrs size" + fi + + echo -n "migrating files..." + local migrate_out=$($LFS_MIGRATE -y -S '1m' $dir) + local rc=$? + [ $rc -eq 0 ] || error "migrate failed rc = $rc" + echo "done" + + # make sure all links have been properly migrated + echo -n "verifying files..." + fid=$($LFS getstripe -F "$file1") || + error "cannot get fid for file $file1" + for i in $(seq 2 $total_count); do + local fid2=$($LFS getstripe -F $dir/file$i) + [ "$fid2" == "$fid" ] || + error "migrated hard link has mismatched FID" + done + + # make sure hard links were properly detected, and migration was + # performed only once for the entire link set; nonlinked files should + # also be migrated + local actual=$(grep -c 'done migrate' <<< "$migrate_out") + local expected=$(($uniq_count + 1)) + [ "$actual" -eq "$expected" ] || + error "hard links individually migrated ($actual != $expected)" + + # make sure the correct number of hard links are present + local hardlinks=$(stat -c '%h' "$file1") + [ $hardlinks -eq $total_count ] || + error "num hard links $hardlinks != $total_count" + echo "done" + + return 0 +} + +test_56xb() { + local dir0="$DIR/$tdir" + + test_mkdir "$dir0" || error "cannot create dir $dir0" + + echo "testing lfs migrate mode when all links fit within xattrs" + LFS_MIGRATE_RSYNC=false check_migrate_links "$dir0" 2 99 + + echo "testing rsync mode when all links fit within xattrs" + LFS_MIGRATE_RSYNC=true check_migrate_links "$dir0" 2 99 + + echo "testing lfs migrate mode when all links do not fit within xattrs" + LFS_MIGRATE_RSYNC=false check_migrate_links "$dir0" 101 100 + + echo "testing rsync mode when all links do not fit within xattrs" + LFS_MIGRATE_RSYNC=true check_migrate_links "$dir0" 101 100 + + # clean up + rm -rf $dir0 +} +run_test 56xb "lfs migration hard link support" + test_56y() { [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.4.53) ] && skip "No HSM $(lustre_build_version $SINGLEMDS) MDS < 2.4.53" && @@ -5111,7 +5313,7 @@ test_56y() { local res="" local dir0=$DIR/$tdir/$testnum - test_mkdir -p $dir0 || error "creating dir $dir0" + test_mkdir -p $dir0 local f1=$dir0/file1 local f2=$dir0/file2 @@ -5172,11 +5374,73 @@ test_56aa() { # LU-5937 } run_test 56aa "lfs find --size under striped dir" +test_56ba() { + # Create composite files with one component + TDIR=$DIR/$tdir/1Mfiles + setup_56 5 1 "--component-end 1M" + # Create composite files with three components + TDIR=$DIR/$tdir/2Mfiles + setup_56 5 2 "-E 2M -E 4M -E 6M" + TDIR=$DIR/$tdir + # Create non-composite files + createmany -o $TDIR/${tfile}- 10 + + local nfiles=$($LFIND --component-end 1M --type f $TDIR | wc -l) + [[ $nfiles == 10 ]] || + error "lfs find -E 1M found $nfiles != 10 files" + + nfiles=$($LFIND ! -E 1M --type f $TDIR | wc -l) + [[ $nfiles == 25 ]] || + error "lfs find ! -E 1M found $nfiles != 25 files" + + # All files have a component that starts at 0 + local nfiles=$($LFIND --component-start 0 --type f $TDIR | wc -l) + [[ $nfiles == 35 ]] || + error "lfs find --component-start 0 found $nfiles != 35 files" + + nfiles=$($LFIND --component-start 2M --type f $TDIR | wc -l) + [[ $nfiles == 15 ]] || + error "$LFIND --component-start 2M found $nfiles != 15 files" + + # All files created here have a componenet that does not starts at 2M + nfiles=$($LFIND ! --component-start 2M --type f $TDIR | wc -l) + [[ $nfiles == 35 ]] || + error "$LFIND ! --component-start 2M found $nfiles != 35 files" + + # Find files with a specified number of components + local nfiles=$($LFIND --component-count 3 --type f $TDIR | wc -l) + [[ $nfiles == 15 ]] || + error "lfs find --component-count 3 found $nfiles != 15 files" + + # Remember non-composite files have a component count of zero + local nfiles=$($LFIND --component-count 0 --type f $TDIR | wc -l) + [[ $nfiles == 10 ]] || + error "lfs find --component-count 0 found $nfiles != 10 files" + + nfiles=$($LFIND ! --component-count 3 --type f $TDIR | wc -l) + [[ $nfiles == 20 ]] || + error "$LFIND ! --component-count 3 found $nfiles != 20 files" + + # All files have a flag called "init" + local nfiles=$($LFIND --component-flags init --type f $TDIR | wc -l) + [[ $nfiles == 35 ]] || + error "$LFIND --component-flags init found $nfiles != 35 files" + + # Multi-component files will have a component not initialized + local nfiles=$($LFIND ! --component-flags init --type f $TDIR | wc -l) + [[ $nfiles == 15 ]] || + error "$LFIND !--component-flags init found $nfiles != 15 files" + + rm -rf $TDIR + +} +run_test 56ba "test lfs find --component-end, -start, -count, and -flags" + test_57a() { [ $PARALLEL == "yes" ] && skip "skip parallel run" && return # note test will not do anything if MDS is not local if [ "$(facet_fstype $SINGLEMDS)" != ldiskfs ]; then - skip "Only applicable to ldiskfs-based MDTs" + skip "ldiskfs only test" return fi @@ -5197,7 +5461,7 @@ run_test 57a "verify MDS filesystem created with large inodes ==" test_57b() { [ $PARALLEL == "yes" ] && skip "skip parallel run" && return if [ "$(facet_fstype $SINGLEMDS)" != ldiskfs ]; then - skip "Only applicable to ldiskfs-based MDTs" + skip "ldiskfs only test" return fi @@ -5209,7 +5473,7 @@ test_57b() { local FILEN=$dir/f$FILECOUNT rm -rf $dir || error "removing $dir" - test_mkdir -p -c1 $dir || error "creating $dir" + test_mkdir -c1 $dir local mdtidx=$($LFS getstripe -M $dir) local mdtname=MDT$(printf %04x $mdtidx) local facet=mds$((mdtidx + 1)) @@ -5285,7 +5549,7 @@ test_60a() { log "$TEST60_HEAD - from kernel mode" do_facet mgs "$LCTL set_param debug=warning; $LCTL dk > /dev/null" - do_facet mgs sh run-llog.sh + do_facet mgs "sh run-llog.sh" || error "run-llog.sh failed" do_facet mgs $LCTL dk > $TMP/$tfile # LU-6388: test llog_reader @@ -5362,7 +5626,23 @@ test_60a() { start mgs $(mgsdevname) $MGS_MOUNT_OPTS || error "start mgs failed" $pass || error "test failed, see FAILED test_60a messages for specifics" } -run_test 60a "llog_test run from kernel module and test llog_reader ==========" +run_test 60a "llog_test run from kernel module and test llog_reader" + +test_60aa() { + # test old logid format + if [ $(lustre_version_code mgs) -le $(version_code 3.1.53) ]; then + do_facet mgs $LCTL dl | grep MGS + do_facet mgs "$LCTL --device %MGS llog_print \\\\\\\$$FSNAME-client" || + error "old llog_print failed" + fi + + # test new logid format + if [ $(lustre_version_code mgs) -ge $(version_code 2.9.53) ]; then + do_facet mgs "$LCTL --device MGS llog_print $FSNAME-client" || + error "new llog_print failed" + fi +} +run_test 60aa "llog_print works with FIDs and simple names" test_60b() { # bug 6411 [ $PARALLEL == "yes" ] && skip "skip parallel run" && return @@ -5501,25 +5781,25 @@ test_64b () { [ $PARALLEL == "yes" ] && skip "skip parallel run" && return sh oos.sh $MOUNT || error "oos.sh failed: $?" } -run_test 64b "check out-of-space detection on client ===========" +run_test 64b "check out-of-space detection on client" test_64c() { $LCTL set_param osc.*OST0000-osc-[^mM]*.cur_grant_bytes=0 } -run_test 64c "verify grant shrink ========================------" +run_test 64c "verify grant shrink" # bug 1414 - set/get directories' stripe info test_65a() { [ $PARALLEL == "yes" ] && skip "skip parallel run" && return - test_mkdir -p $DIR/$tdir + test_mkdir $DIR/$tdir touch $DIR/$tdir/f1 $LVERIFY $DIR/$tdir $DIR/$tdir/f1 || error "lverify failed" } -run_test 65a "directory with no stripe info ====================" +run_test 65a "directory with no stripe info" test_65b() { [ $PARALLEL == "yes" ] && skip "skip parallel run" && return - test_mkdir -p $DIR/$tdir + test_mkdir $DIR/$tdir local STRIPESIZE=$($GETSTRIPE -S $DIR/$tdir) $SETSTRIPE -S $((STRIPESIZE * 2)) -i 0 -c 1 $DIR/$tdir || @@ -5531,21 +5811,20 @@ run_test 65b "directory setstripe -S stripe_size*2 -i 0 -c 1" test_65c() { [ $PARALLEL == "yes" ] && skip "skip parallel run" && return - if [[ $OSTCOUNT -gt 1 ]]; then - test_mkdir -p $DIR/$tdir - local STRIPESIZE=$($GETSTRIPE -S $DIR/$tdir) + [ $OSTCOUNT -lt 2 ] && skip "need at least 2 OSTs" && return + test_mkdir $DIR/$tdir + local stripesize=$($GETSTRIPE -S $DIR/$tdir) - $SETSTRIPE -S $(($STRIPESIZE * 4)) -i 1 \ - -c $(($OSTCOUNT - 1)) $DIR/$tdir || error "setstripe" - touch $DIR/$tdir/f3 - $LVERIFY $DIR/$tdir $DIR/$tdir/f3 || error "lverify failed" - fi + $LFS setstripe -S $((stripesize * 4)) -i 1 \ + -c $((OSTCOUNT - 1)) $DIR/$tdir || error "setstripe" + touch $DIR/$tdir/f3 + $LVERIFY $DIR/$tdir $DIR/$tdir/f3 || error "lverify failed" } run_test 65c "directory setstripe -S stripe_size*4 -i 1 -c $((OSTCOUNT-1))" test_65d() { [ $PARALLEL == "yes" ] && skip "skip parallel run" && return - test_mkdir -p $DIR/$tdir + test_mkdir $DIR/$tdir local STRIPECOUNT=$($GETSTRIPE -c $DIR/$tdir) local STRIPESIZE=$($GETSTRIPE -S $DIR/$tdir) @@ -5566,7 +5845,7 @@ run_test 65d "directory setstripe -S stripe_size -c stripe_count" test_65e() { [ $PARALLEL == "yes" ] && skip "skip parallel run" && return - test_mkdir -p $DIR/$tdir + test_mkdir $DIR/$tdir $SETSTRIPE $DIR/$tdir || error "setstripe" $GETSTRIPE -v $DIR/$tdir | grep "Default" || @@ -5574,38 +5853,38 @@ test_65e() { touch $DIR/$tdir/f6 $LVERIFY $DIR/$tdir $DIR/$tdir/f6 || error "lverify failed" } -run_test 65e "directory setstripe defaults =======================" +run_test 65e "directory setstripe defaults" test_65f() { [ $PARALLEL == "yes" ] && skip "skip parallel run" && return - test_mkdir -p $DIR/${tdir}f + test_mkdir $DIR/${tdir}f $RUNAS $SETSTRIPE $DIR/${tdir}f && error "setstripe succeeded" || true } run_test 65f "dir setstripe permission (should return error) ===" test_65g() { [ $PARALLEL == "yes" ] && skip "skip parallel run" && return - test_mkdir -p $DIR/$tdir + test_mkdir $DIR/$tdir local STRIPESIZE=$($GETSTRIPE -S $DIR/$tdir) - $SETSTRIPE -S $((STRIPESIZE * 2)) -i 0 -c 1 $DIR/$tdir || - error "setstripe" - $SETSTRIPE -d $DIR/$tdir || error "setstripe" - $GETSTRIPE -v $DIR/$tdir | grep "Default" || + $LFS setstripe -S $((STRIPESIZE * 2)) -i 0 -c 1 $DIR/$tdir || + error "setstripe -S failed" + $LFS setstripe -d $DIR/$tdir || error "setstripe -d failed" + $LFS getstripe -v $DIR/$tdir | grep "Default" || error "delete default stripe failed" } -run_test 65g "directory setstripe -d ===========================" +run_test 65g "directory setstripe -d" test_65h() { [ $PARALLEL == "yes" ] && skip "skip parallel run" && return - test_mkdir -p $DIR/$tdir + test_mkdir $DIR/$tdir local STRIPESIZE=$($GETSTRIPE -S $DIR/$tdir) - $SETSTRIPE -S $((STRIPESIZE * 2)) -i 0 -c 1 $DIR/$tdir || - error "setstripe" - test_mkdir -p $DIR/$tdir/dd1 - [ $($GETSTRIPE -c $DIR/$tdir) == $($GETSTRIPE -c $DIR/$tdir/dd1) ] || - error "stripe info inherit failed" + $LFS setstripe -S $((STRIPESIZE * 2)) -i 0 -c 1 $DIR/$tdir || + error "setstripe -S failed" + test_mkdir $DIR/$tdir/dd1 + [ $($LFS getstripe -c $DIR/$tdir) = $($GETSTRIPE -c $DIR/$tdir/dd1) ] || + error "stripe info inherit failed" } run_test 65h "directory stripe info inherit ====================" @@ -5647,7 +5926,7 @@ run_test 65j "set default striping on root directory (bug 6367)=" test_65k() { # bug11679 [ $PARALLEL == "yes" ] && skip "skip parallel run" && return - [[ $OSTCOUNT -lt 2 ]] && skip_env "too few OSTs" && return + [[ $OSTCOUNT -lt 2 ]] && skip_env "needs >= 2 OSTs" && return remote_mds_nodsh && skip "remote MDS with nodsh" && return local disable_precreate=true @@ -5732,23 +6011,6 @@ test_66() { } run_test 66 "update inode blocks count on client ===============" -LLOOP= -LLITELOOPLOAD= -cleanup_68() { - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return - trap 0 - if [ ! -z "$LLOOP" ]; then - if swapon -s | grep -q $LLOOP; then - swapoff $LLOOP || error "swapoff failed" - fi - - $LCTL blockdev_detach $LLOOP || error "detach failed" - rm -f $LLOOP - unset LLOOP - fi - rm -f $DIR/f68* -} - meminfo() { awk '($1 == "'$1':") { print $2 }' /proc/meminfo } @@ -5787,7 +6049,7 @@ test_69() { run_test 69 "verify oa2dentry return -ENOENT doesn't LBUG ======" test_71() { - test_mkdir -p $DIR/$tdir + test_mkdir $DIR/$tdir $LFS setdirstripe -D -c$MDSCOUNT $DIR/$tdir sh rundbench -C -D $DIR/$tdir 2 || error "dbench failed!" } @@ -5923,21 +6185,9 @@ num_inodes() { awk '/lustre_inode_cache/ {print $2; exit}' /proc/slabinfo } -get_inode_slab_tunables() { - awk '/lustre_inode_cache/ {print $9," ",$10," ",$11; exit}' /proc/slabinfo -} - -set_inode_slab_tunables() { - echo "lustre_inode_cache $1" > /proc/slabinfo -} - test_76() { # Now for bug 20433, added originally in bug 1443 [ $PARALLEL == "yes" ] && skip "skip parallel run" && return - local SLAB_SETTINGS=$(get_inode_slab_tunables) local CPUS=$(getconf _NPROCESSORS_ONLN 2>/dev/null) - # we cannot set limit below 1 which means 1 inode in each - # per-cpu cache is still allowed - set_inode_slab_tunables "1 1 0" cancel_lru_locks osc BEFORE_INODES=$(num_inodes) echo "before inodes: $BEFORE_INODES" @@ -5960,7 +6210,6 @@ test_76() { # Now for bug 20433, added originally in bug 1443 error "inode slab grew from $BEFORE_INODES to $AFTER_INODES" fi done - set_inode_slab_tunables "$SLAB_SETTINGS" } run_test 76 "confirm clients recycle inodes properly ====" @@ -6034,6 +6283,96 @@ test_77b() { # bug 10889 } run_test 77b "checksum error on client write, read" +cleanup_77c() { + trap 0 + set_checksums 0 + $LCTL set_param osc.*osc-[^mM]*.checksum_dump=0 + $check_ost && + do_facet ost1 $LCTL set_param obdfilter.*-OST*.checksum_dump=0 + [ -n $osc_file_prefix ] && rm -f ${osc_file_prefix}* + $check_ost && [ -n $ost_file_prefix ] && + do_facet ost1 rm -f ${ost_file_prefix}\* +} + +test_77c() { + [ $PARALLEL == "yes" ] && skip "skip parallel run" && return + $GSS && skip "could not run with gss" && return + + local bad1 + local osc_file_prefix + local osc_file + local check_ost=false + local ost_file_prefix + local ost_file + local orig_cksum + local dump_cksum + local fid + + # ensure corruption will occur on first OSS/OST + $LFS setstripe -i 0 $DIR/$tfile + + [ ! -f $F77_TMP ] && setup_f77 + dd if=$F77_TMP of=$DIR/$tfile bs=1M count=$F77SZ conv=sync || + error "dd write error: $?" + fid=$($LFS path2fid $DIR/$tfile) + + if [ $(lustre_version_code ost1) -ge $(version_code 2.9.57) ] + then + check_ost=true + ost_file_prefix=$(do_facet ost1 $LCTL get_param -n debug_path) + ost_file_prefix=${ost_file_prefix}-checksum_dump-ost-\\${fid} + else + echo "OSS do not support bulk pages dump upon error" + fi + + osc_file_prefix=$($LCTL get_param -n debug_path) + osc_file_prefix=${osc_file_prefix}-checksum_dump-osc-\\${fid} + + trap cleanup_77c EXIT + + set_checksums 1 + # enable bulk pages dump upon error on Client + $LCTL set_param osc.*osc-[^mM]*.checksum_dump=1 + # enable bulk pages dump upon error on OSS + $check_ost && + do_facet ost1 $LCTL set_param obdfilter.*-OST*.checksum_dump=1 + + # flush Client cache to allow next read to reach OSS + cancel_lru_locks osc + + #define OBD_FAIL_OSC_CHECKSUM_RECEIVE 0x408 + $LCTL set_param fail_loc=0x80000408 + dd if=$DIR/$tfile of=/dev/null bs=1M || error "dd read error: $?" + $LCTL set_param fail_loc=0 + + rm -f $DIR/$tfile + + # check cksum dump on Client + osc_file=$(ls ${osc_file_prefix}*) + [ -n "$osc_file" ] || error "no checksum dump file on Client" + # OBD_FAIL_OSC_CHECKSUM_RECEIVE corrupts with "bad1" at start of file + bad1=$(dd if=$osc_file bs=1 count=4 2>/dev/null) || error "dd error: $?" + [ $bad1 == "bad1" ] || error "unexpected corrupt pattern" + orig_cksum=$(dd if=$F77_TMP bs=1 skip=4 count=1048572 2>/dev/null | + cksum) + dump_cksum=$(dd if=$osc_file bs=1 skip=4 2>/dev/null | cksum) + [[ "$orig_cksum" == "$dump_cksum" ]] || + error "dump content does not match on Client" + + $check_ost || skip "No need to check cksum dump on OSS" + + # check cksum dump on OSS + ost_file=$(do_facet ost1 ls ${ost_file_prefix}\*) + [ -n "$ost_file" ] || error "no checksum dump file on OSS" + orig_cksum=$(dd if=$F77_TMP bs=1048576 count=1 2>/dev/null | cksum) + dump_cksum=$(do_facet ost1 dd if=$ost_file 2>/dev/null \| cksum) + [[ "$orig_cksum" == "$dump_cksum" ]] || + error "dump content does not match on OSS" + + cleanup_77c +} +run_test 77c "checksum error on client read with debug" + test_77d() { # bug 10889 [ $PARALLEL == "yes" ] && skip "skip parallel run" && return $GSS && skip "could not run with gss" && return @@ -6100,22 +6439,6 @@ test_77g() { # bug 10889 } run_test 77g "checksum error on OST write, read" -test_77i() { # bug 13805 - [ $PARALLEL == "yes" ] && skip "skip parallel run" && return - $GSS && skip "could not run with gss" && return - #define OBD_FAIL_OSC_CONNECT_CKSUM 0x40b - lctl set_param fail_loc=0x40b - remount_client $MOUNT - lctl set_param fail_loc=0 - 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" = "adler" ] || error "algo set to $algo instead of adler" - done - remount_client $MOUNT -} -run_test 77i "client not supporting OSD_CONNECT_CKSUM" - test_77j() { # bug 13805 [ $PARALLEL == "yes" ] && skip "skip parallel run" && return $GSS && skip "could not run with gss" && return @@ -6123,10 +6446,15 @@ test_77j() { # bug 13805 lctl set_param fail_loc=0x40c remount_client $MOUNT lctl set_param fail_loc=0 - sleep 2 # wait async osc connect to finish - 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'` + # wait async osc connect to finish and reflect updated state value + local i + for (( i=0; i < OSTCOUNT; i++ )) ; do + wait_osc_import_state client ost$((i+1)) FULL + done + + 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" = "adler" ] || error "algo set to $algo instead of adler" done remount_client $MOUNT @@ -6304,80 +6632,59 @@ test_82() { # LU-1031 wait $MULTIPID1 wait $MULTIPID2 } -run_test 82 "Basic grouplock test ===============================" +run_test 82 "Basic grouplock test" + +test_83() { + local sfile="/boot/System.map-$(uname -r)" + # define OBD_FAIL_LLITE_PTASK_IO_FAIL 0x140d + $LCTL set_param fail_loc=0x140d + cp $sfile $DIR/$tfile || error "write failed" + diff -c $sfile $DIR/$tfile || error "files are different" + $LCTL set_param fail_loc=0 + rm -f $DIR/$tfile +} +run_test 83 "Short write in ptask ===============================" -test_99a() { +test_99() { [ -z "$(which cvs 2>/dev/null)" ] && skip_env "could not find cvs" && return - test_mkdir -p $DIR/d99cvsroot - chown $RUNAS_ID $DIR/d99cvsroot - local oldPWD=$PWD # bug 13584, use $TMP as working dir - cd $TMP + test_mkdir $DIR/$tdir.cvsroot + chown $RUNAS_ID $DIR/$tdir.cvsroot - $RUNAS cvs -d $DIR/d99cvsroot init || error "cvs init failed" - cd $oldPWD -} -run_test 99a "cvs init =========================================" + cd $TMP + $RUNAS cvs -d $DIR/$tdir.cvsroot init || error "cvs init failed" -test_99b() { - [ -z "$(which cvs 2>/dev/null)" ] && - skip_env "could not find cvs" && return - [ ! -d $DIR/d99cvsroot ] && test_99a cd /etc/init.d # some versions of cvs import exit(1) when asked to import links or # files they can't read. ignore those files. - TOIGNORE=$(find . -type l -printf '-I %f\n' -o \ - ! -perm /4 -printf '-I %f\n') - $RUNAS cvs -d $DIR/d99cvsroot import -m "nomesg" $TOIGNORE \ - d99reposname vtag rtag -} -run_test 99b "cvs import =======================================" + local toignore=$(find . -type l -printf '-I %f\n' -o \ + ! -perm /4 -printf '-I %f\n') + $RUNAS cvs -d $DIR/$tdir.cvsroot import -m "nomesg" $toignore \ + $tdir.reposname vtag rtag -test_99c() { - [ -z "$(which cvs 2>/dev/null)" ] && skip_env "could not find cvs" && return - [ ! -d $DIR/d99cvsroot ] && test_99b cd $DIR - test_mkdir -p $DIR/d99reposname - chown $RUNAS_ID $DIR/d99reposname - $RUNAS cvs -d $DIR/d99cvsroot co d99reposname -} -run_test 99c "cvs checkout =====================================" + test_mkdir $DIR/$tdir.reposname + chown $RUNAS_ID $DIR/$tdir.reposname + $RUNAS cvs -d $DIR/$tdir.cvsroot co $tdir.reposname -test_99d() { - [ -z "$(which cvs 2>/dev/null)" ] && skip_env "could not find cvs" && return - [ ! -d $DIR/d99cvsroot ] && test_99c - cd $DIR/d99reposname + cd $DIR/$tdir.reposname $RUNAS touch foo99 $RUNAS cvs add -m 'addmsg' foo99 -} -run_test 99d "cvs add ==========================================" - -test_99e() { - [ -z "$(which cvs 2>/dev/null)" ] && skip_env "could not find cvs" && return - [ ! -d $DIR/d99cvsroot ] && test_99c - cd $DIR/d99reposname $RUNAS cvs update -} -run_test 99e "cvs update =======================================" - -test_99f() { - [ -z "$(which cvs 2>/dev/null)" ] && skip_env "could not find cvs" && return - [ ! -d $DIR/d99cvsroot ] && test_99d - cd $DIR/d99reposname $RUNAS cvs commit -m 'nomsg' foo99 - rm -fr $DIR/d99cvsroot + rm -fr $DIR/$tdir.cvsroot } -run_test 99f "cvs commit =======================================" +run_test 99 "cvs strange file/directory operations" test_100() { [ $PARALLEL == "yes" ] && skip "skip parallel run" && return - [ "$NETTYPE" = tcp ] || \ - { skip "TCP secure port test, not useful for NETTYPE=$NETTYPE" && \ + [[ "$NETTYPE" =~ tcp ]] || + { skip "TCP secure port test, not useful for NETTYPE=$NETTYPE" && return ; } remote_ost_nodsh && skip "remote OST with nodsh" && return remote_mds_nodsh && skip "remote MDS with nodsh" && return - remote_servers || \ + remote_servers || { skip "useless for local single node setup" && return; } netstat -tna | ( rc=1; while read PROT SND RCV LOCAL REMOTE STAT; do @@ -6423,7 +6730,7 @@ cleanup_101a() { test_101a() { [ $PARALLEL == "yes" ] && skip "skip parallel run" && return - [ $MDSCOUNT -ge 2 ] && skip "skip now for >= 2 MDTs" && return #LU-4322 + [ $MDSCOUNT -ge 2 ] && skip "needs < 2 MDTs" && return #LU-4322 local s local discard local nreads=10000 @@ -6454,10 +6761,10 @@ test_101a() { fi rm -f $DIR/$tfile || true } -run_test 101a "check read-ahead for random reads ================" +run_test 101a "check read-ahead for random reads" setup_test101bc() { - test_mkdir -p $DIR/$tdir + test_mkdir $DIR/$tdir local STRIPE_SIZE=$1 local FILE_LENGTH=$2 STRIPE_OFFSET=0 @@ -6512,8 +6819,7 @@ ra_check_101() { test_101b() { [ $PARALLEL == "yes" ] && skip "skip parallel run" && return - [[ $OSTCOUNT -lt 2 ]] && - skip_env "skipping stride IO stride-ahead test" && return + [[ $OSTCOUNT -lt 2 ]] && skip_env "needs >= 2 OSTs" && return local STRIPE_SIZE=1048576 local STRIDE_SIZE=$((STRIPE_SIZE*OSTCOUNT)) if [ $SLOW == "yes" ]; then @@ -6643,7 +6949,7 @@ test_101e() { done echo "Cancel LRU locks on lustre client to flush the client cache" - cancel_lru_locks osc + cancel_lru_locks $OSC echo "Reset readahead stats" $LCTL set_param -n llite.*.read_ahead_stats 0 @@ -6695,67 +7001,81 @@ test_101f() { } run_test 101f "check mmap read performance" -test_101g() { - local rpcs - local osts=$(get_facets OST) - local list=$(comma_list $(osts_nodes)) - local p="$TMP/$TESTSUITE-$TESTNAME.parameters" - - save_lustre_params $osts "obdfilter.*.brw_size" > $p +test_101g_brw_size_test() { + local mb=$1 + local pages=$((mb * 1048576 / $(page_size))) - $LFS setstripe -c 1 $DIR/$tfile + $LCTL set_param osc.*.max_pages_per_rpc=${mb}M || + { error "unable to set max_pages_per_rpc=${mb}M"; return 1; } + for mp in $($LCTL get_param -n osc.*.max_pages_per_rpc); do + [ $mp -ne $pages ] && error "max_pages_per_rpc $mp != $pages" && + return 2 + done - if [ $(lustre_version_code ost1) -ge $(version_code 2.8.52) ]; then - set_osd_param $list '' brw_size 16M + $LCTL set_param -n osc.*.rpc_stats=0 - echo "remount client to enable large RPC size" - remount_client $MOUNT || error "remount_client failed" + # 10 RPCs should be enough for the test + local count=10 + dd if=/dev/zero of=$DIR/$tfile bs=${mb}M count=$count || + { error "dd write ${mb} MB blocks failed"; return 3; } + cancel_lru_locks osc + dd of=/dev/null if=$DIR/$tfile bs=${mb}M count=$count || + { error "dd write ${mb} MB blocks failed"; return 4; } - for mp in $($LCTL get_param -n osc.*.max_pages_per_rpc); do - [ "$mp" -eq 4096 ] || - error "max_pages_per_rpc not correctly set" - done + # calculate number of full-sized read and write RPCs + rpcs=($($LCTL get_param -n 'osc.*.rpc_stats' | + sed -n '/pages per rpc/,/^$/p' | + awk '/'$pages':/ { reads += $2; writes += $6 }; \ + END { print reads,writes }')) + [ ${rpcs[0]} -ne $count ] && error "${rpcs[0]} != $count read RPCs" && + return 5 + [ ${rpcs[1]} -ne $count ] && error "${rpcs[1]} != $count write RPCs" && + return 6 - $LCTL set_param -n osc.*.rpc_stats=0 + return 0 +} - # 10*16 MiB should be enough for the test - dd if=/dev/zero of=$DIR/$tfile bs=16M count=10 - cancel_lru_locks osc - dd of=/dev/null if=$DIR/$tfile bs=16M count=10 +test_101g() { + local rpcs + local osts=$(get_facets OST) + local list=$(comma_list $(osts_nodes)) + local p="$TMP/$TESTSUITE-$TESTNAME.parameters" + local brw_size="obdfilter.*.brw_size" + + $LFS setstripe -i 0 -c 1 $DIR/$tfile + + local orig_mb=$(do_facet ost1 $LCTL get_param -n $brw_size | head -n 1) + if [ $(lustre_version_code ost1) -ge $(version_code 2.8.52) -a \ + $(lustre_version_code client) -ge $(version_code 2.8.52) ]; then + [ $(lustre_version_code ost1) -ge $(version_code 2.9.52) ] && + suffix="M" + if [[ $orig_mb -lt 16 ]]; then + save_lustre_params $osts "$brw_size" > $p + do_nodes $list $LCTL set_param -n $brw_size=16$suffix || + error "set 16MB RPC size failed" + + echo "remount client to enable new RPC size" + remount_client $MOUNT || error "remount_client failed" + fi - # calculate 16 MiB RPCs - rpcs=$($LCTL get_param 'osc.*.rpc_stats' | - sed -n '/pages per rpc/,/^$/p' | - awk 'BEGIN { sum = 0 }; /4096:/ { sum += $2 }; - END { print sum }') - echo $rpcs RPCs - [ "$rpcs" -eq 10 ] || error "not all RPCs are 16 MiB BRW rpcs" + test_101g_brw_size_test 16 || error "16MB RPC test failed" + # should be able to set brw_size=12, but no rpc_stats for that + test_101g_brw_size_test 8 || error "8MB RPC test failed" fi - echo "set RPC size to 4MB" - - $LCTL set_param -n osc.*.max_pages_per_rpc=4M osc.*.rpc_stats=0 - dd if=/dev/zero of=$DIR/$tfile bs=4M count=25 - cancel_lru_locks osc - dd of=/dev/null if=$DIR/$tfile bs=4M count=25 - - # calculate 4 MiB RPCs - rpcs=$($LCTL get_param 'osc.*.rpc_stats' | - sed -n '/pages per rpc/,/^$/p' | - awk 'BEGIN { sum = 0 }; /1024:/ { sum += $2 }; - END { print sum }') - echo $rpcs RPCs - [ "$rpcs" -eq 25 ] || error "not all RPCs are 4 MiB BRW rpcs" + test_101g_brw_size_test 4 || error "4MB RPC test failed" - restore_lustre_params < $p - remount_client $MOUNT || error "remount_client failed" + if [[ $orig_mb -lt 16 ]]; then + restore_lustre_params < $p + remount_client $MOUNT || error "remount_client restore failed" + fi rm -f $p $DIR/$tfile } run_test 101g "Big bulk(4/16 MiB) readahead" setup_test102() { - test_mkdir -p $DIR/$tdir + test_mkdir $DIR/$tdir chown $RUNAS_ID $DIR/$tdir STRIPE_SIZE=65536 STRIPE_OFFSET=1 @@ -6777,7 +7097,7 @@ setup_test102() { done cd $DIR - $1 $TAR cf $TMP/f102.tar $tdir --xattrs + $1 tar cf $TMP/f102.tar $tdir --xattrs } cleanup_test102() { @@ -6836,13 +7156,15 @@ test_102a() { setfattr -x user.author1 $testfile || error "$testfile error deleting user.author1" - getfattr -d -m user $testfile 2> /dev/null | grep "user.author1" && - error "$testfile did not delete trusted.name1 xattr" - - # b10667: setting lustre special xattr be silently discarded echo "set lustre special xattr ..." - setfattr -n "trusted.lov" -v "invalid value" $testfile || - error "$testfile allowed setting trusted.lov" + $LFS setstripe -c1 $testfile + local lovea=$(getfattr -n "trusted.lov" -e hex $testfile | + awk -F "=" '/trusted.lov/ { print $2 }' ) + setfattr -n "trusted.lov" -v $lovea $testfile || + error "$testfile doesn't ignore setting trusted.lov again" + setfattr -n "trusted.lov" -v "invalid_value" $testfile && + error "$testfile allow setting invalid trusted.lov" + rm -f $testfile } run_test 102a "user xattr test ==================================" @@ -6852,7 +7174,7 @@ test_102b() { # b10930: get/set/list trusted.lov xattr echo "get/set/list trusted.lov xattr ..." - [[ $OSTCOUNT -lt 2 ]] && skip_env "skipping 2-stripe test" && return + [[ $OSTCOUNT -lt 2 ]] && skip_env "needs >= 2 OSTs" && return local testfile=$DIR/$tfile $SETSTRIPE -S 65536 -i 1 -c $OSTCOUNT $testfile || error "setstripe failed" @@ -6883,8 +7205,8 @@ test_102c() { # b10930: get/set/list lustre.lov xattr echo "get/set/list lustre.lov xattr ..." - [[ $OSTCOUNT -lt 2 ]] && skip_env "skipping 2-stripe test" && return - test_mkdir -p $DIR/$tdir + [[ $OSTCOUNT -lt 2 ]] && skip_env "needs >= 2 OSTs" && return + test_mkdir $DIR/$tdir chown $RUNAS_ID $DIR/$tdir local testfile=$DIR/$tdir/$tfile $RUNAS $SETSTRIPE -S 65536 -i 1 -c $OSTCOUNT $testfile || @@ -6934,39 +7256,35 @@ compare_stripe_info1() { return 0 } -find_lustre_tar() { - [ -n "$(which tar 2>/dev/null)" ] && - strings $(which tar) | grep -q "lustre" && echo tar +have_xattrs_include() { + tar --help | grep -q xattrs-include && + echo --xattrs-include="lustre.*" } test_102d() { [ $PARALLEL == "yes" ] && skip "skip parallel run" && return - # b10930: tar test for trusted.lov xattr - TAR=$(find_lustre_tar) - [ -z "$TAR" ] && skip_env "lustre-aware tar is not installed" && return - [[ $OSTCOUNT -lt 2 ]] && skip_env "skipping N-stripe test" && return + [[ $OSTCOUNT -lt 2 ]] && skip_env "needs >= 2 OSTs" && return + XINC=$(have_xattrs_include) setup_test102 - test_mkdir -p $DIR/d102d - $TAR xf $TMP/f102.tar -C $DIR/d102d --xattrs - cd $DIR/d102d/$tdir + tar xf $TMP/f102.tar -C $DIR/$tdir --xattrs $XINC + cd $DIR/$tdir/$tdir compare_stripe_info1 } -run_test 102d "tar restore stripe info from tarfile,not keep osts ===========" +run_test 102d "tar restore stripe info from tarfile,not keep osts" test_102f() { [ $PARALLEL == "yes" ] && skip "skip parallel run" && return - # b10930: tar test for trusted.lov xattr - TAR=$(find_lustre_tar) - [ -z "$TAR" ] && skip_env "lustre-aware tar is not installed" && return - [[ $OSTCOUNT -lt 2 ]] && skip_env "skipping N-stripe test" && return + [[ $OSTCOUNT -lt 2 ]] && skip_env "needs >= 2 OSTs" && return + XINC=$(have_xattrs_include) setup_test102 - test_mkdir -p $DIR/d102f + test_mkdir $DIR/$tdir.restore cd $DIR - $TAR cf - --xattrs $tdir | $TAR xf - --xattrs -C $DIR/d102f - cd $DIR/d102f/$tdir + tar cf - --xattrs $tdir | tar xf - \ + -C $DIR/$tdir.restore --xattrs $XINC + cd $DIR/$tdir.restore/$tdir compare_stripe_info1 } -run_test 102f "tar copy files, not keep osts ===========" +run_test 102f "tar copy files, not keep osts" grow_xattr() { local xsize=${1:-1024} # in bytes @@ -7036,14 +7354,12 @@ run_test 102i "lgetxattr test on symbolic link ============" test_102j() { [ $PARALLEL == "yes" ] && skip "skip parallel run" && return - TAR=$(find_lustre_tar) - [ -z "$TAR" ] && skip_env "lustre-aware tar is not installed" && return - [[ $OSTCOUNT -lt 2 ]] && skip_env "skipping N-stripe test" && return + [[ $OSTCOUNT -lt 2 ]] && skip_env "needs >= 2 OSTs" && return + XINC=$(have_xattrs_include) setup_test102 "$RUNAS" - test_mkdir -p $DIR/d102j - chown $RUNAS_ID $DIR/d102j - $RUNAS $TAR xf $TMP/f102.tar -C $DIR/d102j --xattrs - cd $DIR/d102j/$tdir + chown $RUNAS_ID $DIR/$tdir + $RUNAS tar xf $TMP/f102.tar -C $DIR/$tdir --xattrs $XINC + cd $DIR/$tdir/$tdir compare_stripe_info1 "$RUNAS" } run_test 102j "non-root tar restore stripe info from tarfile, not keep osts ===" @@ -7051,28 +7367,28 @@ run_test 102j "non-root tar restore stripe info from tarfile, not keep osts ===" test_102k() { [ -z "$(which setfattr 2>/dev/null)" ] && skip "could not find setfattr" && return - touch $DIR/$tfile - # b22187 just check that does not crash for regular file. - setfattr -n trusted.lov $DIR/$tfile - # b22187 'setfattr -n trusted.lov' should work as remove LOV EA for directories - local test_kdir=$DIR/d102k - test_mkdir $test_kdir - local default_size=`$GETSTRIPE -S $test_kdir` - local default_count=`$GETSTRIPE -c $test_kdir` - local default_offset=`$GETSTRIPE -i $test_kdir` + touch $DIR/$tfile + # b22187 just check that does not crash for regular file. + setfattr -n trusted.lov $DIR/$tfile + # b22187 'setfattr -n trusted.lov' should remove LOV EA for directories + local test_kdir=$DIR/$tdir + test_mkdir $test_kdir + local default_size=$($LFS getstripe -S $test_kdir) + local default_count=$($LFS getstripe -c $test_kdir) + local default_offset=$($LFS getstripe -i $test_kdir) $SETSTRIPE -S 65536 -i 0 -c $OSTCOUNT $test_kdir || - error 'dir setstripe failed' - setfattr -n trusted.lov $test_kdir - local stripe_size=`$GETSTRIPE -S $test_kdir` - local stripe_count=`$GETSTRIPE -c $test_kdir` - local stripe_offset=`$GETSTRIPE -i $test_kdir` - [ $stripe_size -eq $default_size ] || - error "stripe size $stripe_size != $default_size" - [ $stripe_count -eq $default_count ] || - error "stripe count $stripe_count != $default_count" - [ $stripe_offset -eq $default_offset ] || - error "stripe offset $stripe_offset != $default_offset" - rm -rf $DIR/$tfile $test_kdir + error 'dir setstripe failed' + setfattr -n trusted.lov $test_kdir + local stripe_size=$($LFS getstripe -S $test_kdir) + local stripe_count=$($LFS getstripe -c $test_kdir) + local stripe_offset=$($LFS getstripe -i $test_kdir) + [ $stripe_size -eq $default_size ] || + error "stripe size $stripe_size != $default_size" + [ $stripe_count -eq $default_count ] || + error "stripe count $stripe_count != $default_count" + [ $stripe_offset -eq $default_offset ] || + error "stripe offset $stripe_offset != $default_offset" + rm -rf $DIR/$tfile $test_kdir } run_test 102k "setfattr without parameter of value shouldn't cause a crash" @@ -7157,8 +7473,13 @@ test_102n() { # LU-4101 mdt: protect internal xattrs # Try to set a garbage xattr. value=0sVGhlIHF1aWNrIGJyb3duIGZveCBqdW1wcyBvdmVyIGl0c2VsZi4= - setfattr --name=trusted.$name --value="$value" $file1 || - error "setxattr 'trusted.$name' failed" + if [[ x$name == "xlov" ]]; then + setfattr --name=trusted.lov --value="$value" $file1 && + error "setxattr invalid 'trusted.lov' success" + else + setfattr --name=trusted.$name --value="$value" $file1 || + error "setxattr invalid 'trusted.$name' failed" + fi # Try to remove the xattr from $file1. We don't care if this # appears to succeed or fail, we just don't want there to be @@ -7246,7 +7567,7 @@ test_102r() { rmdir $DIR/$tdir || error "rmdir" #striped directory - test_mkdir -p $DIR/$tdir || error "make striped dir" + test_mkdir $DIR/$tdir setfattr -n user.$(basename $tdir) $DIR/$tdir || error "setfattr dir" getfattr -n user.$(basename $tdir) $DIR/$tdir || error "getfattr dir" setfattr -x user.$(basename $tdir) $DIR/$tdir || @@ -7341,43 +7662,6 @@ test_103a() { } run_test 103a "acl test =========================================" -test_103b() { - remote_mds_nodsh && skip "remote MDS with nodsh" && return - local noacl=false - local MDT_DEV=$(mdsdevname ${SINGLEMDS//mds/}) - local mountopts=$MDS_MOUNT_OPTS - - if [[ "$MDS_MOUNT_OPTS" =~ "noacl" ]]; then - noacl=true - else - # stop the MDT - stop $SINGLEMDS || error "failed to stop MDT." - # remount the MDT - if [ -z "$MDS_MOUNT_OPTS" ]; then - MDS_MOUNT_OPTS="-o noacl" - else - MDS_MOUNT_OPTS="${MDS_MOUNT_OPTS},noacl" - fi - start $SINGLEMDS $MDT_DEV $MDS_MOUNT_OPTS || - error "failed to start MDT." - MDS_MOUNT_OPTS=$mountopts - fi - - touch $DIR/$tfile - setfacl -m u:bin:rw $DIR/$tfile && error "setfacl should fail" - - if ! $noacl; then - # stop the MDT - stop $SINGLEMDS || error "failed to stop MDT." - # remount the MDT - start $SINGLEMDS $MDT_DEV $MDS_MOUNT_OPTS || - error "failed to start MDT." - fi - - true -} -run_test 103b "MDS mount option 'noacl'" - test_103c() { mkdir -p $DIR/$tdir cp -rp $DIR/$tdir $DIR/$tdir.bak @@ -7457,11 +7741,11 @@ test_105c() { fi rm -f $DIR/$tfile } -run_test 105c "lockf when mounted without -o flock test ========" +run_test 105c "lockf when mounted without -o flock test" test_105d() { # bug 15924 [ $PARALLEL == "yes" ] && skip "skip parallel run" && return - test_mkdir -p $DIR/$tdir + test_mkdir $DIR/$tdir flock_is_enabled || { skip "mount w/o flock enabled" && return; } #define OBD_FAIL_LDLM_CP_CB_WAIT 0x315 $LCTL set_param fail_loc=0x80000315 @@ -7474,10 +7758,10 @@ test_105e() { # bug 22660 && 22040 touch $DIR/$tfile flocks_test 3 $DIR/$tfile } -run_test 105e "Two conflicting flocks from same process =======" +run_test 105e "Two conflicting flocks from same process" test_106() { #bug 10921 - test_mkdir -p $DIR/$tdir + test_mkdir $DIR/$tdir $DIR/$tdir && error "exec $DIR/$tdir succeeded" chmod 777 $DIR/$tdir || error "chmod $DIR/$tdir failed" } @@ -7518,10 +7802,9 @@ test_107() { run_test 107 "Coredump on SIG" test_110() { - test_mkdir -p $DIR/$tdir - test_mkdir $DIR/$tdir/$(str_repeat 'a' 255) || - error "mkdir with 255 char failed" - test_mkdir $DIR/$tdir/$(str_repeat 'b' 256) && + test_mkdir $DIR/$tdir + test_mkdir $DIR/$tdir/$(str_repeat 'a' 255) + $LFS mkdir -c $MDSCOUNT $DIR/$tdir/$(str_repeat 'b' 256) && error "mkdir with 256 char should fail, but did not" touch $DIR/$tdir/$(str_repeat 'x' 255) || error "create with 255 char failed" @@ -7533,58 +7816,141 @@ test_110() { } run_test 110 "filename length checking" +# +# Purpose: To verify dynamic thread (OSS) creation. +# test_115() { [ $PARALLEL == "yes" ] && skip "skip parallel run" && return - OSTIO_pre=$(ps -e | grep ll_ost_io | awk '{ print $4 }'| sort -n | - tail -1 | cut -c11-20) - [ -z "$OSTIO_pre" ] && skip "no OSS threads" && return + remote_ost_nodsh && skip "remote OST with nodsh" && return + + # Lustre does not stop service threads once they are started. + # Reset number of running threads to default. + stopall + setupall + + local OSTIO_pre + local save_params="$TMP/sanity-$TESTNAME.parameters" + + # Get ll_ost_io count before I/O + OSTIO_pre=$(do_facet ost1 \ + "$LCTL get_param ost.OSS.ost_io.threads_started | cut -d= -f2") + # Exit if lustre is not running (ll_ost_io not running). + [ -z "$OSTIO_pre" ] && error "no OSS threads" + echo "Starting with $OSTIO_pre threads" + local thread_max=$((OSTIO_pre * 2)) + local rpc_in_flight=$((thread_max * 2)) + # Number of I/O Process proposed to be started. + local nfiles + local facets=$(get_facets OST) - NUMTEST=20000 - NUMFREE=$(df -i -P $DIR | tail -n 1 | awk '{ print $4 }') - [[ $NUMFREE -lt $NUMTEST ]] && NUMTEST=$(($NUMFREE - 1000)) - echo "$NUMTEST creates/unlinks" - test_mkdir -p $DIR/$tdir - createmany -o $DIR/$tdir/$tfile $NUMTEST - unlinkmany $DIR/$tdir/$tfile $NUMTEST + save_lustre_params client "osc.*OST*.max_rpcs_in_flight" > $save_params + save_lustre_params $facets "ost.OSS.ost_io.threads_max" >> $save_params - OSTIO_post=$(ps -e | grep ll_ost_io | awk '{ print $4 }' | sort -n | - tail -1 | cut -c11-20) + # Set in_flight to $rpc_in_flight + $LCTL set_param osc.*OST*.max_rpcs_in_flight=$rpc_in_flight || + error "Failed to set max_rpcs_in_flight to $rpc_in_flight" + nfiles=${rpc_in_flight} + # Set ost thread_max to $thread_max + do_facet ost1 "$LCTL set_param ost.OSS.ost_io.threads_max=$thread_max" - # don't return an error - [ $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 + # 5 Minutes should be sufficient for max number of OSS + # threads(thread_max) to be created. + local timeout=300 - echo "Started with $OSTIO_pre threads, ended with $OSTIO_post" + # Start I/O. + local WTL=${WTL:-"$LUSTRE/tests/write_time_limit"} + test_mkdir $DIR/$tdir + for i in $(seq $nfiles); do + local file=$DIR/$tdir/${tfile}-$i + $LFS setstripe -c -1 -i 0 $file + ($WTL $file $timeout)& + done + + # I/O Started - Wait for thread_started to reach thread_max or report + # error if thread_started is more than thread_max. + echo "Waiting for thread_started to reach thread_max" + local thread_started=0 + local end_time=$((SECONDS + timeout)) + + while [ $SECONDS -le $end_time ] ; do + echo -n "." + # Get ost i/o thread_started count. + thread_started=$(do_facet ost1 \ + "$LCTL get_param \ + ost.OSS.ost_io.threads_started | cut -d= -f2") + # Break out if thread_started is equal/greater than thread_max + if [[ $thread_started -ge $thread_max ]]; then + echo ll_ost_io thread_started $thread_started, \ + equal/greater than thread_max $thread_max + break + fi + sleep 1 + done + + # Cleanup - We have the numbers, Kill i/o jobs if running. + jobcount=($(jobs -p)) + for i in $(seq 0 $((${#jobcount[@]}-1))) + do + kill -9 ${jobcount[$i]} + if [ $? -ne 0 ] ; then + echo Warning: \ + Failed to Kill \'WTL\(I/O\)\' with pid ${jobcount[$i]} + fi + done + + # Cleanup files left by WTL binary. + for i in $(seq $nfiles); do + local file=$DIR/$tdir/${tfile}-$i + rm -rf $file + if [ $? -ne 0 ] ; then + echo "Warning: Failed to delete file $file" + fi + done + + restore_lustre_params <$save_params + rm -f $save_params || echo "Warning: delete file '$save_params' failed" + + # Error out if no new thread has started or Thread started is greater + # than thread max. + if [[ $thread_started -le $OSTIO_pre || + $thread_started -gt $thread_max ]]; then + error "ll_ost_io: thread_started $thread_started" \ + "OSTIO_pre $OSTIO_pre, thread_max $thread_max." \ + "No new thread started or thread started greater " \ + "than thread_max." + fi } run_test 115 "verify dynamic thread creation====================" free_min_max () { wait_delete_completed AVAIL=($(lctl get_param -n osc.*[oO][sS][cC]-[^M]*.kbytesavail)) - echo OST kbytes available: ${AVAIL[@]} - MAXI=0; MAXV=${AVAIL[0]} - MINI=0; MINV=${AVAIL[0]} + echo "OST kbytes available: ${AVAIL[@]}" + MAXV=${AVAIL[0]} + MAXI=0 + MINV=${AVAIL[0]} + MINI=0 for ((i = 0; i < ${#AVAIL[@]}; i++)); do #echo OST $i: ${AVAIL[i]}kb if [[ ${AVAIL[i]} -gt $MAXV ]]; then - MAXV=${AVAIL[i]}; MAXI=$i + MAXV=${AVAIL[i]} + MAXI=$i fi if [[ ${AVAIL[i]} -lt $MINV ]]; then - MINV=${AVAIL[i]}; MINI=$i + MINV=${AVAIL[i]} + MINI=$i fi done - echo Min free space: OST $MINI: $MINV - echo Max free space: OST $MAXI: $MAXV + echo "Min free space: OST $MINI: $MINV" + echo "Max free space: OST $MAXI: $MAXV" } test_116a() { # was previously test_116() [ $PARALLEL == "yes" ] && skip "skip parallel run" && return remote_mds_nodsh && skip "remote MDS with nodsh" && return - [[ $OSTCOUNT -lt 2 ]] && skip_env "$OSTCOUNT < 2 OSTs" && return + [[ $OSTCOUNT -lt 2 ]] && skip_env "needs >= 2 OSTs" && return echo -n "Free space priority " do_facet $SINGLEMDS lctl get_param -n lo*.*-mdtlov.qos_prio_free | @@ -7600,9 +7966,9 @@ test_116a() { # was previously test_116() # Check if we need to generate uneven OSTs test_mkdir -p $DIR/$tdir/OST${MINI} - local FILL=$(($MINV / 4)) - local DIFF=$(($MAXV - $MINV)) - local DIFF2=$(($DIFF * 100 / $MINV)) + local FILL=$((MINV / 4)) + local DIFF=$((MAXV - MINV)) + local DIFF2=$((DIFF * 100 / MINV)) local threshold=$(do_facet $SINGLEMDS \ lctl get_param -n *.*MDT0000-mdtlov.qos_threshold_rr | head -n1) @@ -7616,15 +7982,15 @@ test_116a() { # was previously test_116() else # generate uneven OSTs. Write 2% over the QOS threshold value echo "no" - DIFF=$(($threshold - $DIFF2 + 2)) - DIFF2=$(( ($MINV * $DIFF)/100 )) - echo "Fill ${DIFF}% remaining space in OST${MINI} with ${DIFF2}KB" + DIFF=$((threshold - DIFF2 + 2)) + DIFF2=$((MINV * DIFF / 100)) + echo "Fill $DIFF% remaining space in OST$MINI with ${DIFF2}KB" $SETSTRIPE -i $MINI -c 1 $DIR/$tdir/OST${MINI} || error "setstripe failed" - DIFF=$(($DIFF2 / 2048)) + DIFF=$((DIFF2 / 2048)) i=0 while [ $i -lt $DIFF ]; do - i=$(($i + 1)) + i=$((i + 1)) dd if=/dev/zero of=$DIR/$tdir/OST${MINI}/$tfile-$i \ bs=2M count=1 2>/dev/null echo -n . @@ -7635,10 +8001,10 @@ test_116a() { # was previously test_116() free_min_max fi - DIFF=$(($MAXV - $MINV)) - DIFF2=$(($DIFF * 100 / $MINV)) - echo -n "diff=${DIFF}=${DIFF2}% must be > ${threshold}% for QOS mode..." - if [[ $DIFF2 -gt $threshold ]]; then + DIFF=$((MAXV - MINV)) + DIFF2=$((DIFF * 100 / MINV)) + echo -n "diff=$DIFF=$DIFF2% must be > $threshold% for QOS mode..." + if [ $DIFF2 -gt $threshold ]; then echo "ok" else echo "failed - QOS mode won't be used" @@ -7647,12 +8013,14 @@ test_116a() { # was previously test_116() return fi - MINI1=$MINI; MINV1=$MINV - MAXI1=$MAXI; MAXV1=$MAXV + MINI1=$MINI + MINV1=$MINV + MAXI1=$MAXI + MAXV1=$MAXV # now fill using QOS $SETSTRIPE -c 1 $DIR/$tdir - FILL=$(($FILL / 200)) + FILL=$((FILL / 200)) if [ $FILL -gt 600 ]; then FILL=600 fi @@ -7670,31 +8038,31 @@ test_116a() { # was previously test_116() echo "Note: free space may not be updated, so measurements might be off" free_min_max - DIFF2=$(($MAXV - $MINV)) + DIFF2=$((MAXV - MINV)) echo "free space delta: orig $DIFF final $DIFF2" [ $DIFF2 -gt $DIFF ] && echo "delta got worse!" - DIFF=$(($MINV1 - ${AVAIL[$MINI1]})) + DIFF=$((MINV1 - ${AVAIL[$MINI1]})) echo "Wrote ${DIFF}KB to smaller OST $MINI1" - DIFF2=$(($MAXV1 - ${AVAIL[$MAXI1]})) + DIFF2=$((MAXV1 - ${AVAIL[$MAXI1]})) echo "Wrote ${DIFF2}KB to larger OST $MAXI1" if [[ $DIFF -gt 0 ]]; then - FILL=$(($DIFF2 * 100 / $DIFF - 100)) + FILL=$((DIFF2 * 100 / DIFF - 100)) echo "Wrote ${FILL}% more data to larger OST $MAXI1" fi # Figure out which files were written where UUID=$(lctl get_param -n lov.${FSNAME}-clilov-*.target_obd | - awk '/'$MINI1': / {print $2; exit}') + awk '/'$MINI1': / {print $2; exit}') echo $UUID MINC=$($GETSTRIPE --ost $UUID $DIR/$tdir | grep $DIR | wc -l) echo "$MINC files created on smaller OST $MINI1" UUID=$(lctl get_param -n lov.${FSNAME}-clilov-*.target_obd | - awk '/'$MAXI1': / {print $2; exit}') + awk '/'$MAXI1': / {print $2; exit}') echo $UUID MAXC=$($GETSTRIPE --ost $UUID $DIR/$tdir | grep $DIR | wc -l) echo "$MAXC files created on larger OST $MAXI1" if [[ $MINC -gt 0 ]]; then - FILL=$(($MAXC * 100 / $MINC - 100)) + FILL=$((MAXC * 100 / MINC - 100)) echo "Wrote ${FILL}% more files to larger OST $MAXI1" fi [[ $MAXC -gt $MINC ]] || @@ -8099,7 +8467,7 @@ test_118k() #define OBD_FAIL_OST_BRW_WRITE_BULK 0x20e set_nodes_failloc "$(osts_nodes)" 0x20e - test_mkdir -p $DIR/$tdir + test_mkdir $DIR/$tdir for ((i=0;i<10;i++)); do (dd if=/dev/zero of=$DIR/$tdir/$tfile-$i bs=1M count=10 || \ @@ -8119,16 +8487,16 @@ test_118l() { [ $PARALLEL == "yes" ] && skip "skip parallel run" && return # LU-646 - test_mkdir -p $DIR/$tdir + test_mkdir $DIR/$tdir $MULTIOP $DIR/$tdir Dy || error "fsync dir failed" rm -rf $DIR/$tdir } -run_test 118l "fsync dir =========" +run_test 118l "fsync dir" test_118m() # LU-3066 { [ $PARALLEL == "yes" ] && skip "skip parallel run" && return - test_mkdir -p $DIR/$tdir + test_mkdir $DIR/$tdir $MULTIOP $DIR/$tdir DY || error "fdatasync dir failed" rm -rf $DIR/$tdir } @@ -8154,7 +8522,7 @@ run_test 119a "Short directIO read must return actual read amount" test_119b() # bug 11737 { - [ "$OSTCOUNT" -lt "2" ] && skip_env "skipping 2-stripe test" && return + [[ $OSTCOUNT -lt 2 ]] && skip_env "needs >= 2 OSTs" && return $SETSTRIPE -c 2 $DIR/$tfile || error "setstripe failed" dd if=/dev/zero of=$DIR/$tfile bs=1M count=1 seek=1 || error "dd failed" @@ -8203,9 +8571,9 @@ run_test 119d "The DIO path should try to send a new rpc once one is completed" test_120a() { [ $PARALLEL == "yes" ] && skip "skip parallel run" && return remote_mds_nodsh && skip "remote MDS with nodsh" && return - test_mkdir -p $DIR/$tdir - [ -z "`lctl get_param -n mdc.*.connect_flags | grep early_lock_cancel`" ] && \ - skip "no early lock cancel on server" && return 0 + test_mkdir $DIR/$tdir + $LCTL get_param -n mdc.*.connect_flags | grep -q early_lock_cancel || + { skip "no early lock cancel on server"; return 0; } lru_resize_disable mdc lru_resize_disable osc @@ -8235,13 +8603,13 @@ run_test 120a "Early Lock Cancel: mkdir test" test_120b() { [ $PARALLEL == "yes" ] && skip "skip parallel run" && return remote_mds_nodsh && skip "remote MDS with nodsh" && return - test_mkdir $DIR/$tdir - [ -z "$(lctl get_param -n mdc.*.connect_flags | grep early_lock_cancel)" ] && \ - skip "no early lock cancel on server" && return 0 - lru_resize_disable mdc - lru_resize_disable osc - cancel_lru_locks mdc - stat $DIR/$tdir > /dev/null + test_mkdir $DIR/$tdir + $LCTL get_param -n mdc.*.connect_flags | grep -q early_lock_cancel || + { skip "no early lock cancel on server"; return 0; } + lru_resize_disable mdc + lru_resize_disable osc + cancel_lru_locks mdc + stat $DIR/$tdir > /dev/null can1=$(do_facet $SINGLEMDS \ "$LCTL get_param -n ldlm.services.ldlm_canceld.stats" | awk '/ldlm_cancel/ {print $2}') @@ -8264,12 +8632,12 @@ test_120c() { [ $PARALLEL == "yes" ] && skip "skip parallel run" && return remote_mds_nodsh && skip "remote MDS with nodsh" && return test_mkdir -c1 $DIR/$tdir - [ -z "$(lctl get_param -n mdc.*.connect_flags | grep early_lock_cancel)" ] && \ - skip "no early lock cancel on server" && return 0 - lru_resize_disable mdc - lru_resize_disable osc - test_mkdir -p -c1 $DIR/$tdir/d1 - test_mkdir -p -c1 $DIR/$tdir/d2 + $LCTL get_param -n mdc.*.connect_flags | grep -q early_lock_cancel || + { skip "no early lock cancel on server"; return 0; } + lru_resize_disable mdc + lru_resize_disable osc + test_mkdir -c1 $DIR/$tdir/d1 + test_mkdir -c1 $DIR/$tdir/d2 touch $DIR/$tdir/d1/f1 cancel_lru_locks mdc stat $DIR/$tdir/d1 $DIR/$tdir/d2 $DIR/$tdir/d1/f1 > /dev/null @@ -8294,9 +8662,9 @@ run_test 120c "Early Lock Cancel: link test" test_120d() { [ $PARALLEL == "yes" ] && skip "skip parallel run" && return remote_mds_nodsh && skip "remote MDS with nodsh" && return - test_mkdir -p -c1 $DIR/$tdir - [ -z "$(lctl get_param -n mdc.*.connect_flags | grep early_lock_cancel)" ] && \ - skip "no early lock cancel on server" && return 0 + test_mkdir -c1 $DIR/$tdir + $LCTL get_param -n mdc.*.connect_flags | grep -q early_lock_cancel || + { skip "no early lock cancel on server"; return 0; } lru_resize_disable mdc lru_resize_disable osc touch $DIR/$tdir @@ -8322,12 +8690,12 @@ run_test 120d "Early Lock Cancel: setattr test" test_120e() { [ $PARALLEL == "yes" ] && skip "skip parallel run" && return - ! $($LCTL get_param -n mdc.*.connect_flags | grep -q early_lock_can) && - skip "no early lock cancel on server" && return 0 + $LCTL get_param -n mdc.*.connect_flags | grep -q early_lock_cancel || + { skip "no early lock cancel on server"; return 0; } remote_mds_nodsh && skip "remote MDS with nodsh" && return local dlmtrace_set=false - test_mkdir -p -c1 $DIR/$tdir + test_mkdir -c1 $DIR/$tdir lru_resize_disable mdc lru_resize_disable osc ! $LCTL get_param debug | grep -q dlmtrace && @@ -8365,14 +8733,14 @@ run_test 120e "Early Lock Cancel: unlink test" test_120f() { [ $PARALLEL == "yes" ] && skip "skip parallel run" && return - [ -z "`lctl get_param -n mdc.*.connect_flags | grep early_lock_cancel`" ] && \ - skip "no early lock cancel on server" && return 0 + $LCTL get_param -n mdc.*.connect_flags | grep -q early_lock_cancel || + { skip "no early lock cancel on server"; return 0; } remote_mds_nodsh && skip "remote MDS with nodsh" && return - test_mkdir -p -c1 $DIR/$tdir - lru_resize_disable mdc - lru_resize_disable osc - test_mkdir -p -c1 $DIR/$tdir/d1 - test_mkdir -p -c1 $DIR/$tdir/d2 + test_mkdir -c1 $DIR/$tdir + lru_resize_disable mdc + lru_resize_disable osc + test_mkdir -c1 $DIR/$tdir/d1 + test_mkdir -c1 $DIR/$tdir/d2 dd if=/dev/zero of=$DIR/$tdir/d1/f1 count=1 dd if=/dev/zero of=$DIR/$tdir/d2/f2 count=1 cancel_lru_locks mdc @@ -8405,17 +8773,17 @@ run_test 120f "Early Lock Cancel: rename test" test_120g() { [ $PARALLEL == "yes" ] && skip "skip parallel run" && return - [ -z "`lctl get_param -n mdc.*.connect_flags | grep early_lock_cancel`" ] && \ - skip "no early lock cancel on server" && return 0 + $LCTL get_param -n mdc.*.connect_flags | grep -q early_lock_cancel || + { skip "no early lock cancel on server"; return 0; } remote_mds_nodsh && skip "remote MDS with nodsh" && return - lru_resize_disable mdc - lru_resize_disable osc - count=10000 - echo create $count files - test_mkdir -p $DIR/$tdir - cancel_lru_locks mdc - cancel_lru_locks osc - t0=`date +%s` + lru_resize_disable mdc + lru_resize_disable osc + count=10000 + echo create $count files + test_mkdir $DIR/$tdir + cancel_lru_locks mdc + cancel_lru_locks osc + t0=$(date +%s) can0=$(do_facet $SINGLEMDS \ "$LCTL get_param -n ldlm.services.ldlm_canceld.stats" | @@ -8465,14 +8833,14 @@ run_test 121 "read cancel race =========" test_123a() { # was test 123, statahead(bug 11401) [ $PARALLEL == "yes" ] && skip "skip parallel run" && return - SLOWOK=0 - if [ -z "$(grep "processor.*: 1" /proc/cpuinfo)" ]; then - log "testing on UP system. Performance may be not as good as expected." - SLOWOK=1 - fi + SLOWOK=0 + if ! grep -q "processor.*: 1" /proc/cpuinfo; then + log "testing UP system. Performance may be lower than expected." + SLOWOK=1 + fi - rm -rf $DIR/$tdir - test_mkdir -p $DIR/$tdir + rm -rf $DIR/$tdir + test_mkdir $DIR/$tdir NUMFREE=$(df -i -P $DIR | tail -n 1 | awk '{ print $4 }') [[ $NUMFREE -gt 100000 ]] && NUMFREE=100000 || NUMFREE=$((NUMFREE-1000)) MULT=10 @@ -8548,7 +8916,7 @@ run_test 123a "verify statahead work" test_123b () { # statahead(bug 15027) [ $PARALLEL == "yes" ] && skip "skip parallel run" && return - test_mkdir -p $DIR/$tdir + test_mkdir $DIR/$tdir createmany -o $DIR/$tdir/$tfile-%d 1000 cancel_lru_locks mdc @@ -8568,20 +8936,20 @@ run_test 123b "not panic with network error in statahead enqueue (bug 15027)" test_124a() { [ $PARALLEL == "yes" ] && skip "skip parallel run" && return - [ -z "$($LCTL get_param -n mdc.*.connect_flags | grep lru_resize)" ] && - skip "no lru resize on server" && return 0 - local NR=2000 - test_mkdir -p $DIR/$tdir || error "failed to create $DIR/$tdir" + $LCTL get_param -n mdc.*.connect_flags | grep -q lru_resize || + { skip "no lru resize on server"; return 0; } + local NR=2000 + test_mkdir $DIR/$tdir - log "create $NR files at $DIR/$tdir" - createmany -o $DIR/$tdir/f $NR || - error "failed to create $NR files in $DIR/$tdir" + log "create $NR files at $DIR/$tdir" + createmany -o $DIR/$tdir/f $NR || + error "failed to create $NR files in $DIR/$tdir" - cancel_lru_locks mdc - ls -l $DIR/$tdir > /dev/null + cancel_lru_locks mdc + ls -l $DIR/$tdir > /dev/null - local NSDIR="" - local LRU_SIZE=0 + local NSDIR="" + local LRU_SIZE=0 for VALUE in $($LCTL get_param ldlm.namespaces.*mdc-*.lru_size); do local PARAM=$(echo ${VALUE[0]} | cut -d "=" -f1) LRU_SIZE=$($LCTL get_param -n $PARAM) @@ -8670,8 +9038,8 @@ get_max_pool_limit() test_124b() { [ $PARALLEL == "yes" ] && skip "skip parallel run" && return - [ -z "$($LCTL get_param -n mdc.*.connect_flags | grep lru_resize)" ] && - skip "no lru resize on server" && return 0 + $LCTL get_param -n mdc.*.connect_flags | grep -q lru_resize || + { skip "no lru resize on server"; return 0; } LIMIT=$(get_max_pool_limit) @@ -8688,8 +9056,7 @@ test_124b() { fi lru_resize_disable mdc - test_mkdir -p $DIR/$tdir/disable_lru_resize || - error "failed to create $DIR/$tdir/disable_lru_resize" + test_mkdir -p $DIR/$tdir/disable_lru_resize createmany -o $DIR/$tdir/disable_lru_resize/f $NR log "doing ls -la $DIR/$tdir/disable_lru_resize 3 times" @@ -8711,9 +9078,8 @@ test_124b() { log "lru_size = $(lctl get_param -n ldlm.namespaces.*mdc*.lru_size)" unlinkmany $DIR/$tdir/disable_lru_resize/f $NR - lru_resize_enable mdc - test_mkdir -p $DIR/$tdir/enable_lru_resize || - error "failed to create $DIR/$tdir/enable_lru_resize" + lru_resize_enable mdc + test_mkdir -p $DIR/$tdir/enable_lru_resize createmany -o $DIR/$tdir/enable_lru_resize/f $NR log "doing ls -la $DIR/$tdir/enable_lru_resize 3 times" @@ -8747,13 +9113,13 @@ run_test 124b "lru resize (performance test) =======================" test_124c() { [ $PARALLEL == "yes" ] && skip "skip parallel run" && return - [ -z "$($LCTL get_param -n mdc.*.connect_flags | grep lru_resize)" ] && - skip "no lru resize on server" && return 0 + $LCTL get_param -n mdc.*.connect_flags | grep -q lru_resize || + { skip "no lru resize on server"; return 0; } # cache ununsed locks on client local nr=100 cancel_lru_locks mdc - test_mkdir -p $DIR/$tdir || error "failed to create $DIR/$tdir" + test_mkdir $DIR/$tdir createmany -o $DIR/$tdir/f $nr || error "failed to create $nr files in $DIR/$tdir" ls -l $DIR/$tdir > /dev/null @@ -8765,7 +9131,7 @@ test_124c() { echo "unused=$unused, max_age=$max_age, recalc_p=$recalc_p" # set lru_max_age to 1 sec - $LCTL set_param $nsdir.lru_max_age=1000 # jiffies + $LCTL set_param $nsdir.lru_max_age=1000 # milliseconds echo "sleep $((recalc_p * 2)) seconds..." sleep $((recalc_p * 2)) @@ -8778,19 +9144,22 @@ test_124c() { run_test 124c "LRUR cancel very aged locks" test_125() { # 13358 - [ -z "$(lctl get_param -n llite.*.client_type | grep local)" ] && skip "must run as local client" && return - [ -z "$(lctl get_param -n mdc.*-mdc-*.connect_flags | grep acl)" ] && skip "must have acl enabled" && return + $LCTL get_param -n llite.*.client_type | grep -q local || + { skip "must run as local client"; return; } + $LCTL get_param -n mdc.*-mdc-*.connect_flags | grep -q acl || + { skip "must have acl enabled"; return; } [ -z "$(which setfacl)" ] && skip "must have setfacl tool" && return - test_mkdir -p $DIR/d125 || error "mkdir failed" - $SETSTRIPE -S 65536 -c -1 $DIR/d125 || error "setstripe failed" - setfacl -R -m u:bin:rwx $DIR/d125 || error "setfacl $DIR/d125 failed" - ls -ld $DIR/d125 || error "cannot access $DIR/d125" + test_mkdir $DIR/$tdir + $LFS setstripe -S 65536 -c -1 $DIR/$tdir || error "setstripe failed" + setfacl -R -m u:bin:rwx $DIR/$tdir || error "setfacl $DIR/$tdir failed" + ls -ld $DIR/$tdir || error "cannot access $DIR/$tdir" } run_test 125 "don't return EPROTO when a dir has a non-default striping and ACLs" test_126() { # bug 12829/13455 - [ -z "$(lctl get_param -n llite.*.client_type | grep local)" ] && skip "must run as local client" && return - [ "$UID" != 0 ] && skip_env "skipping $TESTNAME (must run as root)" && return + $LCTL get_param -n llite.*.client_type | grep -q local || + { skip "must run as local client"; return; } + [ "$UID" != 0 ] && skip_env "$TESTNAME (must run as root)" && return $GSS && skip "must run as gss disabled" && return $RUNAS -u 0 -g 1 touch $DIR/$tfile || error "touch failed" @@ -8901,18 +9270,16 @@ set_dir_limits () { local canondev local node - local LDPROC=/proc/fs/ldiskfs + local ldproc=/proc/fs/ldiskfs local facets=$(get_facets MDS) for facet in ${facets//,/ }; do canondev=$(ldiskfs_canon \ *.$(convert_facet2label $facet).mntdev $facet) - do_facet $facet "test -e $LDPROC/$canondev/max_dir_size" || - LDPROC=/sys/fs/ldiskfs - do_facet $facet "echo $1 >$LDPROC/$canondev/max_dir_size" - do_facet $facet "test -e $LDPROC/$canondev/warning_dir_size" || - LDPROC=/sys/fs/ldiskfs - do_facet $facet "echo $2 >$LDPROC/$canondev/warning_dir_size" + do_facet $facet "test -e $ldproc/$canondev/max_dir_size" || + ldproc=/sys/fs/ldiskfs + do_facet $facet "echo $1 >$ldproc/$canondev/max_dir_size" + do_facet $facet "echo $2 >$ldproc/$canondev/warning_dir_size" done } @@ -8925,70 +9292,65 @@ check_mds_dmesg() { } test_129() { + [[ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.5.56) ]] || + { skip "Need MDS version with at least 2.5.56"; return 0; } + [ $PARALLEL == "yes" ] && skip "skip parallel run" && return if [ "$(facet_fstype $SINGLEMDS)" != ldiskfs ]; then - skip "Only applicable to ldiskfs-based MDTs" + skip "ldiskfs only test" return fi remote_mds_nodsh && skip "remote MDS with nodsh" && return local ENOSPC=28 local EFBIG=27 - local has_warning=0 + local has_warning=false rm -rf $DIR/$tdir mkdir -p $DIR/$tdir # block size of mds1 - local MDT_DEV=$(mdsdevname ${SINGLEMDS//mds/}) - local MDSBLOCKSIZE=$($LCTL get_param -n mdc.*MDT0000*.blocksize) - local MAX=$((MDSBLOCKSIZE * 5)) - set_dir_limits $MAX $MAX - local I=$(stat -c%s "$DIR/$tdir") - local J=0 - while [[ $I -le $MAX ]]; do - $MULTIOP $DIR/$tdir/$J Oc + local maxsize=$(($($LCTL get_param -n mdc.*MDT0000*.blocksize) * 5)) + set_dir_limits $maxsize $maxsize + local dirsize=$(stat -c%s "$DIR/$tdir") + local nfiles=0 + while [[ $dirsize -le $maxsize ]]; do + $MULTIOP $DIR/$tdir/file_base_$nfiles Oc rc=$? - if [ $has_warning -eq 0 ]; then - check_mds_dmesg '"is approaching"' && - has_warning=1 + if ! $has_warning; then + check_mds_dmesg '"is approaching"' && has_warning=true fi - #check two errors ENOSPC for new version of ext4 max_dir_size patch - #mainline kernel commit df981d03eeff7971ac7e6ff37000bfa702327ef1 - #and EFBIG for previous versions + # check two errors: + # ENOSPC for new ext4 max_dir_size (kernel commit df981d03ee) + # EFBIG for previous versions included in ldiskfs series if [ $rc -eq $EFBIG -o $rc -eq $ENOSPC ]; then set_dir_limits 0 0 echo "return code $rc received as expected" - createmany -o $DIR/$tdir/$J_file_ 1000 || + createmany -o $DIR/$tdir/file_extra_$nfiles. 5 || error_exit "create failed w/o dir size limit" check_mds_dmesg '"has reached"' || - error_exit "has reached message should be output" + error_exit "reached message should be output" [ $has_warning -eq 0 ] && error_exit "warning message should be output" - I=$(stat -c%s "$DIR/$tdir") + dirsize=$(stat -c%s "$DIR/$tdir") - if [ $(lustre_version_code $SINGLEMDS) -lt \ - $(version_code 2.4.51) ] - then - [[ $I -eq $MAX ]] && return 0 - else - [[ $I -gt $MAX ]] && return 0 - fi - error_exit "current dir size $I, previous limit $MAX" + [[ $dirsize -ge $maxsize ]] && return 0 + error_exit "current dir size $dirsize, " \ + "previous limit $maxsize" elif [ $rc -ne 0 ]; then set_dir_limits 0 0 - error_exit "return code $rc received instead of expected " \ - "$EFBIG or $ENOSPC, files in dir $I" + error_exit "return $rc received instead of expected " \ + "$EFBIG or $ENOSPC, files in dir $dirsize" fi - J=$((J+1)) - I=$(stat -c%s "$DIR/$tdir") + nfiles=$((nfiles + 1)) + dirsize=$(stat -c%s "$DIR/$tdir") done set_dir_limits 0 0 - error "exceeded dir size limit $MAX($MDSCOUNT) : $I bytes" + error "exceeded dir size limit $maxsize($MDSCOUNT) : $dirsize bytes" } run_test 129 "test directory size limit ========================" @@ -9049,8 +9411,7 @@ test_130a() { run_test 130a "FIEMAP (1-stripe file)" test_130b() { - [ "$OSTCOUNT" -lt "2" ] && - skip_env "skipping FIEMAP on $OSTCOUNT-stripe file" && return + [ "$OSTCOUNT" -lt "2" ] && skip_env "needs >= 2 OSTs" && return local filefrag_op=$(filefrag -e 2>&1 | grep "invalid option") [ -n "$filefrag_op" ] && skip_env "filefrag does not support FIEMAP" && @@ -9110,8 +9471,7 @@ test_130b() { run_test 130b "FIEMAP ($OSTCOUNT-stripe file)" test_130c() { - [ "$OSTCOUNT" -lt "2" ] && - skip_env "skipping FIEMAP on 2-stripe file" && return + [[ $OSTCOUNT -lt 2 ]] && skip_env "needs >= 2 OSTs" && return filefrag_op=$(filefrag -e 2>&1 | grep "invalid option") [ -n "$filefrag_op" ] && skip "filefrag does not support FIEMAP" && @@ -9177,8 +9537,7 @@ test_130c() { run_test 130c "FIEMAP (2-stripe file with hole)" test_130d() { - [ "$OSTCOUNT" -lt "3" ] && - skip_env "skipping FIEMAP on N-stripe file test" && return + [[ $OSTCOUNT -lt 3 ]] && skip_env "needs >= 3 OSTs" && return filefrag_op=$(filefrag -e 2>&1 | grep "invalid option") [ -n "$filefrag_op" ] && skip "filefrag does not support FIEMAP" && @@ -9192,8 +9551,8 @@ test_130d() { [ "$(facet_fstype ost$(($($GETSTRIPE -i $fm_file) + 1)))" = "zfs" ] && skip "ORI-366/LU-1941: FIEMAP unimplemented on ZFS" && return - local actual_stripecnt=$($GETSTRIPE -c $fm_file) - dd if=/dev/zero of=$fm_file bs=1M count=$actual_stripecnt || + local actual_stripe_count=$($GETSTRIPE -c $fm_file) + dd if=/dev/zero of=$fm_file bs=1M count=$actual_stripe_count || error "dd failed on $fm_file" filefrag -ves $fm_file || error "filefrag $fm_file failed" @@ -9225,7 +9584,7 @@ test_130d() { (( tot_len += ext_len )) last_lun=$frag_lun done - if (( num_luns != actual_stripecnt || tot_len != 1024 )); then + if (( num_luns != actual_stripe_count || tot_len != 1024 )); then cleanup_130 error "FIEMAP on $fm_file failed; returned wrong number of " \ "luns or wrong len for OST $last_lun" @@ -9239,8 +9598,7 @@ test_130d() { run_test 130d "FIEMAP (N-stripe file)" test_130e() { - [ "$OSTCOUNT" -lt "2" ] && - skip_env "skipping continuation FIEMAP test" && return + [[ $OSTCOUNT -lt 2 ]] && skip_env "needs >= 2 OSTs" && return filefrag_op=$(filefrag -e 2>&1 | grep "invalid option") [ -n "$filefrag_op" ] && skip "filefrag does not support FIEMAP" && return @@ -9304,19 +9662,24 @@ run_test 130e "FIEMAP (test continuation FIEMAP calls)" # Test for writev/readv test_131a() { - rwv -f $DIR/$tfile -w -n 3 524288 1048576 1572864 || \ - error "writev test failed" - rwv -f $DIR/$tfile -r -v -n 2 1572864 1048576 || \ - error "readv failed" + rwv -f $DIR/$tfile -w -n 3 524288 1048576 1572864 || + error "writev test failed" + rwv -f $DIR/$tfile -r -v -n 2 1572864 1048576 || + error "readv failed" rm -f $DIR/$tfile } run_test 131a "test iov's crossing stripe boundary for writev/readv" test_131b() { - rwv -f $DIR/$tfile -w -a -n 3 524288 1048576 1572864 || \ - error "append writev test failed" - rwv -f $DIR/$tfile -w -a -n 2 1572864 1048576 || \ - error "append writev test failed" + local fsize=$((524288 + 1048576 + 1572864)) + rwv -f $DIR/$tfile -w -a -n 3 524288 1048576 1572864 && + $CHECKSTAT -t file $DIR/$tfile -s $fsize || + error "append writev test failed" + + ((fsize += 1572864 + 1048576)) + rwv -f $DIR/$tfile -w -a -n 2 1572864 1048576 && + $CHECKSTAT -t file $DIR/$tfile -s $fsize || + error "append writev test failed" rm -f $DIR/$tfile } run_test 131b "test append writev" @@ -9346,22 +9709,27 @@ test_131e() { run_test 131e "test read hitting hole" check_stats() { + local facet=$1 + local op=$2 + local want=${3:-0} local res - local count - case $1 in - $SINGLEMDS) res=`do_facet $SINGLEMDS $LCTL get_param mdt.$FSNAME-MDT0000.md_stats | grep "$2"` + + case $facet in + mds*) res=$(do_facet $facet \ + $LCTL get_param mdt.$FSNAME-MDT0000.md_stats | grep "$op") ;; - ost) res=`do_facet ost1 $LCTL get_param obdfilter.$FSNAME-OST0000.stats | grep "$2"` + ost*) res=$(do_facet $facet \ + $LCTL get_param obdfilter.$FSNAME-OST0000.stats | grep "$op") ;; - *) error "Wrong argument $1" ;; + *) error "Wrong facet '$facet'" ;; esac echo $res - [ -z "$res" ] && error "The counter for $2 on $1 was not incremented" + [ "$res" ] || error "The counter for $op on $facet was not incremented" # if the argument $3 is zero, it means any stat increment is ok. - if [[ $3 -gt 0 ]]; then - count=$(echo $res | awk '{ print $2 }') - [[ $count -ne $3 ]] && - error "The $2 counter on $1 is wrong - expected $3" + if [[ $want -gt 0 ]]; then + local count=$(echo $res | awk '{ print $2 }') + [[ $count -ne $want ]] && + error "The $op counter on $facet is $count, not $want" fi } @@ -9445,11 +9813,11 @@ test_133c() { [ $PARALLEL == "yes" ] && skip "skip parallel run" && return remote_ost_nodsh && skip "remote OST with nodsh" && return remote_mds_nodsh && skip "remote MDS with nodsh" && return - local testdir=$DIR/${tdir}/stats_testdir - test_mkdir -p ${testdir} || error "mkdir failed" + local testdir=$DIR/$tdir/stats_testdir + test_mkdir -p $testdir # verify obdfilter stats. - $SETSTRIPE -c 1 -i 0 ${testdir}/${tfile} + $SETSTRIPE -c 1 -i 0 $testdir/$tfile sync cancel_lru_locks osc wait_delete_completed @@ -9458,22 +9826,23 @@ test_133c() { do_facet $SINGLEMDS $LCTL set_param mdt.*.md_stats=clear do_facet ost1 $LCTL set_param obdfilter.*.stats=clear - dd if=/dev/zero of=${testdir}/${tfile} conv=notrunc bs=512k count=1 || error "dd failed" + dd if=/dev/zero of=$testdir/$tfile conv=notrunc bs=512k count=1 || + error "dd failed" sync cancel_lru_locks osc - check_stats ost "write" 1 + check_stats ost1 "write" 1 - dd if=${testdir}/${tfile} of=/dev/null bs=1k count=1 || error "dd failed" - check_stats ost "read" 1 + dd if=$testdir/$tfile of=/dev/null bs=1k count=1 || error "dd failed" + check_stats ost1 "read" 1 - > ${testdir}/${tfile} || error "truncate failed" - check_stats ost "punch" 1 + > $testdir/$tfile || error "truncate failed" + check_stats ost1 "punch" 1 - rm -f ${testdir}/${tfile} || error "file remove failed" + rm -f $testdir/$tfile || error "file remove failed" wait_delete_completed - check_stats ost "destroy" 1 + check_stats ost1 "destroy" 1 - rm -rf $DIR/${tdir} + rm -rf $DIR/$tdir } run_test 133c "Verifying OST stats ========================================" @@ -9648,17 +10017,15 @@ test_133e() { } run_test 133e "Verifying OST {read,write}_bytes nid stats =================" -proc_dirs="" -for dir in /proc/fs/lustre/ /proc/sys/lnet/ /proc/sys/lustre/ \ - /sys/fs/lustre/ /sys/fs/lnet/ /sys/kernel/debug/lnet/ \ - /sys/kernel/debug/lustre/; do - [[ -d $dir ]] && proc_dirs+=" $dir" -done +proc_regexp="/{proc,sys}/{fs,sys,kernel/debug}/{lustre,lnet}/" test_133f() { remote_mds_nodsh && skip "remote MDS with nodsh" && return remote_ost_nodsh && skip "remote OST with nodsh" && return # First without trusting modes. + local proc_dirs=$(eval \ls -d $proc_regexp 2>/dev/null) + echo "proc_dirs='$proc_dirs'" + [ -n "$proc_dirs" ] || error "no proc_dirs on $HOSTNAME" find $proc_dirs -exec cat '{}' \; &> /dev/null # Second verifying readability. @@ -9667,12 +10034,16 @@ test_133f() { # eventually, this can also be replaced with "lctl get_param -R", # but not until that option is always available on the server local facet - for facet in $SINGLEMDS ost1; do - do_facet $facet find $proc_dirs \ + for facet in mds1 ost1; do + local facet_proc_dirs=$(do_facet $facet \ + \\\ls -d $proc_regexp 2>/dev/null) + echo "${facet}_proc_dirs='$facet_proc_dirs'" + [ -z "$facet_proc_dirs" ] && error "no proc_dirs on $facet" + do_facet $facet find $facet_proc_dirs \ ! -name req_history \ -exec cat '{}' \\\; &> /dev/null - do_facet $facet find $proc_dirs \ + do_facet $facet find $facet_proc_dirs \ ! -name req_history \ -type f \ -exec cat '{}' \\\; &> /dev/null || @@ -9685,6 +10056,9 @@ test_133g() { remote_mds_nodsh && skip "remote MDS with nodsh" && return remote_ost_nodsh && skip "remote OST with nodsh" && return # Second verifying writability. + local proc_dirs=$(eval \ls -d $proc_regexp 2>/dev/null) + echo "proc_dirs='$proc_dirs'" + [ -n "$proc_dirs" ] || error "no proc_dirs on $HOSTNAME" find $proc_dirs \ -type f \ -not -name force_lbug \ @@ -9692,21 +10066,20 @@ test_133g() { -exec badarea_io '{}' \; &> /dev/null || error "find $proc_dirs failed" - [ $(lustre_version_code $SINGLEMDS) -le $(version_code 2.5.54) ] && - skip "Too old lustre on MDS" && return - - [ $(lustre_version_code ost1) -le $(version_code 2.5.54) ] && - skip "Too old lustre on ost1" && return - local facet - for facet in $SINGLEMDS ost1; do - do_facet $facet find $proc_dirs \ + for facet in mds1 ost1; do + [ $(lustre_version_code $facet) -le $(version_code 2.5.54) ] && + skip "Too old lustre on $facet" && continue + local facet_proc_dirs=$(do_facet $facet \ + \\\ls -d $proc_regexp 2> /dev/null) + echo "${facet}_proc_dirs='$facet_proc_dirs'" + [ -z "$facet_proc_dirs" ] && error "no proc_dirs on $facet" + do_facet $facet find $facet_proc_dirs \ -type f \ -not -name force_lbug \ -not -name changelog_mask \ -exec badarea_io '{}' \\\; &> /dev/null || - error "$facet find $proc_dirs failed" - + error "$facet find $facet_proc_dirs failed" done # remount the FS in case writes/reads /proc break the FS @@ -9716,6 +10089,34 @@ test_133g() { } run_test 133g "Check for Oopses on bad io area writes/reads in /proc" +test_133h() { + remote_mds_nodsh && skip "remote MDS with nodsh" && return + remote_ost_nodsh && skip "remote OST with nodsh" && return + [[ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.9.54) ]] && + skip "Need MDS version at least 2.9.54" && return + + local facet + for facet in client mds1 ost1; do + local facet_proc_dirs=$(do_facet $facet \ + \\\ls -d $proc_regexp 2> /dev/null) + [ -z "$facet_proc_dirs" ] && error "no proc_dirs on $facet" + echo "${facet}_proc_dirs='$facet_proc_dirs'" + # Get the list of files that are missing the terminating newline + local missing=($(do_facet $facet \ + find ${facet_proc_dirs} -type f \| \ + while read F\; do \ + awk -v FS='\v' -v RS='\v\v' \ + "'END { if(NR>0 && \ + \\\$NF !~ /.*\\\n\$/) \ + print FILENAME}'" \ + '\$F'\; \ + done 2>/dev/null)) + [ ${#missing[*]} -eq 0 ] || + error "files do not end with newline: ${missing[*]}" + done +} +run_test 133h "Proc files should end with newlines" + test_134a() { remote_mds_nodsh && skip "remote MDS with nodsh" && return [[ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.7.54) ]] && @@ -9795,32 +10196,32 @@ run_test 134b "Server rejects lock request when reaching lock_limit_mb" test_140() { #bug-17379 [ $PARALLEL == "yes" ] && skip "skip parallel run" && return - test_mkdir -p $DIR/$tdir || error "Creating dir $DIR/$tdir" + test_mkdir $DIR/$tdir cd $DIR/$tdir || error "Changing to $DIR/$tdir" cp $(which stat) . || error "Copying stat to $DIR/$tdir" # VFS limits max symlink depth to 5(4KSTACK) or 7(8KSTACK) or 8 # For kernel > 3.5, bellow only tests consecutive symlink (MAX 40) local i=0 - while i=`expr $i + 1`; do - test_mkdir -p $i || error "Creating dir $i" - cd $i || error "Changing to $i" - ln -s ../stat stat || error "Creating stat symlink" - # Read the symlink until ELOOP present, - # not LBUGing the system is considered success, - # we didn't overrun the stack. - $OPENFILE -f O_RDONLY stat >/dev/null 2>&1; ret=$? - [ $ret -ne 0 ] && { - if [ $ret -eq 40 ]; then - break # -ELOOP - else - error "Open stat symlink" - return - fi - } - done - i=`expr $i - 1` - echo "The symlink depth = $i" + while i=$((i + 1)); do + test_mkdir $i + cd $i || error "Changing to $i" + ln -s ../stat stat || error "Creating stat symlink" + # Read the symlink until ELOOP present, + # not LBUGing the system is considered success, + # we didn't overrun the stack. + $OPENFILE -f O_RDONLY stat >/dev/null 2>&1; ret=$? + if [ $ret -ne 0 ]; then + if [ $ret -eq 40 ]; then + break # -ELOOP + else + error "Open stat symlink" + return + fi + fi + done + i=$((i - 1)) + echo "The symlink depth = $i" [ $i -eq 5 -o $i -eq 7 -o $i -eq 8 -o $i -eq 40 ] || error "Invalid symlink depth" @@ -9836,31 +10237,31 @@ test_150() { [ $PARALLEL == "yes" ] && skip "skip parallel run" && return local TF="$TMP/$tfile" - dd if=/dev/urandom of=$TF bs=6096 count=1 || error "dd failed" - cp $TF $DIR/$tfile - cancel_lru_locks osc - cmp $TF $DIR/$tfile || error "$TMP/$tfile $DIR/$tfile differ" - remount_client $MOUNT - df -P $MOUNT - cmp $TF $DIR/$tfile || error "$TF $DIR/$tfile differ (remount)" + dd if=/dev/urandom of=$TF bs=6096 count=1 || error "dd failed" + cp $TF $DIR/$tfile + cancel_lru_locks $OSC + cmp $TF $DIR/$tfile || error "$TMP/$tfile $DIR/$tfile differ" + remount_client $MOUNT + df -P $MOUNT + cmp $TF $DIR/$tfile || error "$TF $DIR/$tfile differ (remount)" - $TRUNCATE $TF 6000 - $TRUNCATE $DIR/$tfile 6000 - cancel_lru_locks osc - cmp $TF $DIR/$tfile || error "$TF $DIR/$tfile differ (truncate1)" + $TRUNCATE $TF 6000 + $TRUNCATE $DIR/$tfile 6000 + cancel_lru_locks $OSC + cmp $TF $DIR/$tfile || error "$TF $DIR/$tfile differ (truncate1)" - echo "12345" >>$TF - echo "12345" >>$DIR/$tfile - cancel_lru_locks osc - cmp $TF $DIR/$tfile || error "$TF $DIR/$tfile differ (append1)" + echo "12345" >>$TF + echo "12345" >>$DIR/$tfile + cancel_lru_locks $OSC + cmp $TF $DIR/$tfile || error "$TF $DIR/$tfile differ (append1)" - echo "12345" >>$TF - echo "12345" >>$DIR/$tfile - cancel_lru_locks osc - cmp $TF $DIR/$tfile || error "$TF $DIR/$tfile differ (append2)" + echo "12345" >>$TF + echo "12345" >>$DIR/$tfile + cancel_lru_locks $OSC + cmp $TF $DIR/$tfile || error "$TF $DIR/$tfile differ (append2)" - rm -f $TF - true + rm -f $TF + true } run_test 150 "truncate/append tests" @@ -9868,12 +10269,12 @@ run_test 150 "truncate/append tests" function roc_hit_init() { local list=$(comma_list $(osts_nodes)) local dir=$DIR/$tdir-check - local file=$dir/file + local file=$dir/$tfile local BEFORE local AFTER local idx - test_mkdir -p $dir + test_mkdir $dir #use setstripe to do a write to every ost for i in $(seq 0 $((OSTCOUNT-1))); do $SETSTRIPE -c 1 -i $i $dir || error "$SETSTRIPE $file failed" @@ -10391,7 +10792,8 @@ run_test 154f "get parent fids by reading link ea" test_154g() { - [[ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.6.92) ]] || + [[ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.6.92) && \ + $(lustre_version_code client) -gt $(version_code 2.6.99) ]] || { skip "Need MDS version at least 2.6.92"; return 0; } [ -n "$FILESET" ] && skip "SKIP due to FILESET set" && return @@ -10407,7 +10809,7 @@ test_155_small_load() { dd if=/dev/urandom of=$temp bs=6096 count=1 || \ error "dd of=$temp bs=6096 count=1 failed" cp $temp $file - cancel_lru_locks osc + cancel_lru_locks $OSC cmp $temp $file || error "$temp $file differ" $TRUNCATE $temp 6000 @@ -10949,9 +11351,41 @@ test_160d() { } run_test 160d "verify that changelog log catch the migrate event" +test_160e() { + remote_mds_nodsh && skip "remote MDS with nodsh" && return + + # Create a user + CL_USER=$(do_facet $SINGLEMDS $LCTL --device $MDT0 \ + changelog_register -n) + echo "Registered as changelog user $CL_USER" + trap cleanup_changelog EXIT + + # Delete a future user (expect fail) + do_facet $SINGLEMDS $LCTL --device $MDT0 changelog_deregister cl77 + local rc=$? + + if [ $rc -eq 0 ]; then + error "Deleted non-existant user cl77" + elif [ $rc -ne 2 ]; then + error "changelog_deregister failed with $rc, " \ + "expected 2 (ENOENT)" + fi + + # Clear to a bad index (1 billion should be safe) + $LFS changelog_clear $MDT0 $CL_USER 1000000000 + rc=$? + + if [ $rc -eq 0 ]; then + error "Successfully cleared to invalid CL index" + elif [ $rc -ne 22 ]; then + error "changelog_clear failed with $rc, expected 22 (EINVAL)" + fi +} +run_test 160e "changelog negative testing" + test_161a() { [ $PARALLEL == "yes" ] && skip "skip parallel run" && return - test_mkdir -p -c1 $DIR/$tdir + test_mkdir -c1 $DIR/$tdir cp /etc/hosts $DIR/$tdir/$tfile test_mkdir -c1 $DIR/$tdir/foo1 test_mkdir -c1 $DIR/$tdir/foo2 @@ -11124,6 +11558,55 @@ test_161c() { } run_test 161c "check CL_RENME[UNLINK] changelog record flags" +test_161d() { + local user + local pid + local fid + + # cleanup previous run + rm -rf $DIR/$tdir/$tfile + + user=$(do_facet $SINGLEMDS $LCTL --device $MDT0 \ + changelog_register -n) + [[ $? -eq 0 ]] || error "changelog_register failed" + + # work in a standalone dir to avoid locking on $DIR/$MOUNT to + # interfer with $MOUNT/.lustre/fid/ access + mkdir $DIR/$tdir + [[ $? -eq 0 ]] || error "mkdir failed" + + #define OBD_FAIL_LLITE_CREATE_NODE_PAUSE 0x140c | OBD_FAIL_ONCE + $LCTL set_param fail_loc=0x8000140c + # 5s pause + $LCTL set_param fail_val=5 + + # create file + echo foofoo > $DIR/$tdir/$tfile & + pid=$! + + # wait for create to be delayed + sleep 2 + + ps -p $pid + [[ $? -eq 0 ]] || error "create should be blocked" + + local tempfile=$(mktemp) + fid=$(changelog_extract_field $MDT0 "CREAT" "$tfile" "t=") + cat $MOUNT/.lustre/fid/$fid 2>/dev/null >$tempfile || error "cat failed" + # some delay may occur during ChangeLog publishing and file read just + # above, that could allow file write to happen finally + [[ -s $tempfile ]] && echo "file should be empty" + + $LCTL set_param fail_loc=0 + + wait $pid + [[ $? -eq 0 ]] || error "create failed" + + $LFS changelog_clear $MDT0 $user 0 + do_facet $SINGLEMDS $LCTL --device $MDT0 changelog_deregister $user +} +run_test 161d "create with concurrent .lustre/fid access" + check_path() { local expected=$1 shift @@ -11437,12 +11920,11 @@ test_180b() { local rc=0 local rmmod_remote=0 - do_facet ost1 "lsmod | grep -q obdecho || " \ - "{ insmod ${LUSTRE}/obdecho/obdecho.ko || " \ - "modprobe obdecho; }" && rmmod_remote=1 + do_rpc_nodes $(facet_active_host ost1) load_module obdecho/obdecho && + rmmod_remote=true || error "failed to load module obdecho" target=$(do_facet ost1 $LCTL dl | awk '/obdfilter/ {print $4;exit}') [[ -n $target ]] && { obdecho_test $target ost1 || rc=1; } - [ $rmmod_remote -eq 1 ] && do_facet ost1 "rmmod obdecho" + $rmmod_remote && do_facet ost1 "rmmod obdecho" return $rc } run_test 180b "test obdecho directly on obdfilter" @@ -11475,7 +11957,7 @@ test_180c() { # LU-2598 run_test 180c "test huge bulk I/O size on obdfilter, don't LASSERT" test_181() { # bug 22177 - test_mkdir -p $DIR/$tdir || error "creating dir $DIR/$tdir" + test_mkdir $DIR/$tdir # create enough files to index the directory createmany -o $DIR/$tdir/foobar 4000 # print attributes for debug purpose @@ -11552,7 +12034,7 @@ test_184a() { check_swap_layouts_support && return 0 dir0=$DIR/$tdir/$testnum - test_mkdir -p -c1 $dir0 || error "creating dir $dir0" + test_mkdir -p -c1 $dir0 ref1=/etc/passwd ref2=/etc/group file1=$dir0/f1 @@ -11673,14 +12155,16 @@ test_184d() { error "create $file2 failed" $OPENFILE -f O_CREAT:O_LOV_DELAY_CREATE $file3 || error "create $file3 failed" - lovea1=$($LFS getstripe $file1 | sed 1d) + lovea1=$(get_layout_param $file1) $LFS swap_layouts $file2 $file3 || error "swap $file2 $file3 layouts failed" $LFS swap_layouts $file1 $file2 || error "swap $file1 $file2 layouts failed" - lovea2=$($LFS getstripe $file2 | sed 1d) + lovea2=$(get_layout_param $file2) + echo "$lovea1" + echo "$lovea2" [ "$lovea1" == "$lovea2" ] || error "lovea $lovea1 != $lovea2" lovea1=$(getfattr -n trusted.lov $file1 | grep ^trusted) @@ -11817,8 +12301,12 @@ test_200() { local test_path=$POOL_ROOT/$POOL_DIR_NAME local file_dir=$POOL_ROOT/file_tst local subdir=$test_path/subdir - local rc=0 + + if ! combined_mgs_mds ; then + mount_mgs_client + fi + while : ; do # former test_200a test_200b pool_add $POOL || { rc=$? ; break; } @@ -11838,7 +12326,7 @@ test_200() { pool_create_files $POOL $file_dir $files "$ost_list" \ || { rc=$? ; break; } # former test_200g test_200h - pool_lfs_df $POOL || { rc=$? ; break; } + pool_lfs_df $POOL || { rc=$? ; break; } pool_file_rel_path $POOL $test_path || { rc=$? ; break; } # former test_201a test_201b test_201c @@ -11846,11 +12334,15 @@ test_200() { local f=$test_path/$tfile pool_remove_all_targets $POOL $f || { rc=$? ; break; } - pool_remove $POOL $f || { rc=$? ; break; } + pool_remove $POOL $f || { rc=$? ; break; } break done - cleanup_pools + destroy_test_pools + + if ! combined_mgs_mds ; then + umount_mgs_client + fi return $rc } run_test 200 "OST pools" @@ -11864,109 +12356,93 @@ default_attr() { check_default_stripe_attr() { ACTUAL=$($GETSTRIPE $* $DIR/$tdir) case $1 in - --stripe-count|--count) + --stripe-count|-c) [ -n "$2" ] && EXPECTED=0 || EXPECTED=$(default_attr count);; - --stripe-size|--size) + --stripe-size|-S) [ -n "$2" ] && EXPECTED=0 || EXPECTED=$(default_attr size);; - --stripe-index|--index) + --stripe-index|-i) EXPECTED=-1;; *) error "unknown getstripe attr '$1'" esac - [ $ACTUAL != $EXPECTED ] && + [ $ACTUAL == $EXPECTED ] || error "$DIR/$tdir has $1 '$ACTUAL', not '$EXPECTED'" } test_204a() { - test_mkdir -p $DIR/$tdir + test_mkdir $DIR/$tdir $SETSTRIPE --stripe-count 0 --stripe-size 0 --stripe-index -1 $DIR/$tdir check_default_stripe_attr --stripe-count check_default_stripe_attr --stripe-size check_default_stripe_attr --stripe-index - - return 0 } -run_test 204a "Print default stripe attributes =================" +run_test 204a "Print default stripe attributes" test_204b() { - test_mkdir -p $DIR/$tdir + test_mkdir $DIR/$tdir $SETSTRIPE --stripe-count 1 $DIR/$tdir check_default_stripe_attr --stripe-size check_default_stripe_attr --stripe-index - - return 0 } -run_test 204b "Print default stripe size and offset ===========" +run_test 204b "Print default stripe size and offset" test_204c() { - test_mkdir -p $DIR/$tdir + test_mkdir $DIR/$tdir $SETSTRIPE --stripe-size 65536 $DIR/$tdir check_default_stripe_attr --stripe-count check_default_stripe_attr --stripe-index - - return 0 } -run_test 204c "Print default stripe count and offset ===========" +run_test 204c "Print default stripe count and offset" test_204d() { - test_mkdir -p $DIR/$tdir + test_mkdir $DIR/$tdir $SETSTRIPE --stripe-index 0 $DIR/$tdir check_default_stripe_attr --stripe-count check_default_stripe_attr --stripe-size - - return 0 } -run_test 204d "Print default stripe count and size =============" +run_test 204d "Print default stripe count and size" test_204e() { - test_mkdir -p $DIR/$tdir + test_mkdir $DIR/$tdir $SETSTRIPE -d $DIR/$tdir check_default_stripe_attr --stripe-count --raw check_default_stripe_attr --stripe-size --raw check_default_stripe_attr --stripe-index --raw - - return 0 } -run_test 204e "Print raw stripe attributes =================" +run_test 204e "Print raw stripe attributes" test_204f() { - test_mkdir -p $DIR/$tdir + test_mkdir $DIR/$tdir $SETSTRIPE --stripe-count 1 $DIR/$tdir check_default_stripe_attr --stripe-size --raw check_default_stripe_attr --stripe-index --raw - - return 0 } -run_test 204f "Print raw stripe size and offset ===========" +run_test 204f "Print raw stripe size and offset" test_204g() { - test_mkdir -p $DIR/$tdir + test_mkdir $DIR/$tdir $SETSTRIPE --stripe-size 65536 $DIR/$tdir check_default_stripe_attr --stripe-count --raw check_default_stripe_attr --stripe-index --raw - - return 0 } -run_test 204g "Print raw stripe count and offset ===========" +run_test 204g "Print raw stripe count and offset" test_204h() { - test_mkdir -p $DIR/$tdir + test_mkdir $DIR/$tdir $SETSTRIPE --stripe-index 0 $DIR/$tdir check_default_stripe_attr --stripe-count --raw check_default_stripe_attr --stripe-size --raw - - return 0 } -run_test 204h "Print raw stripe count and size =============" +run_test 204h "Print raw stripe count and size" # Figure out which job scheduler is being used, if any, # or use a fake one @@ -12050,6 +12526,9 @@ cleanup_205() { } test_205() { # Job stats + [[ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.7.1) ]] || + { skip "Need MDS version with at least 2.7.1"; return 0; } + [ $PARALLEL == "yes" ] && skip "skip parallel run" && return remote_mgs_nodsh && skip "remote MGS with nodsh" && return remote_mds_nodsh && skip "remote MDS with nodsh" && return @@ -12304,16 +12783,12 @@ run_test 214 "hash-indexed directory test - bug 20133" # having "abc" as 1st arg, creates $TMP/lnet_abc.out and $TMP/lnet_abc.sys create_lnet_proc_files() { - lctl get_param -n $1 >$TMP/lnet_$1.out || error "cannot read lnet.$1" - sysctl lnet.$1 >$TMP/lnet_$1.sys_tmp || error "cannot read lnet.$1" - - sed "s/^lnet.$1\ =\ //g" "$TMP/lnet_$1.sys_tmp" >$TMP/lnet_$1.sys - rm -f "$TMP/lnet_$1.sys_tmp" + lctl get_param -n $1 >$TMP/lnet_$1.sys || error "cannot read lnet.$1" } # counterpart of create_lnet_proc_files remove_lnet_proc_files() { - rm -f $TMP/lnet_$1.out $TMP/lnet_$1.sys + rm -f $TMP/lnet_$1.sys } # uses 1st arg as trailing part of filename, 2nd arg as description for reports, @@ -12423,7 +12898,6 @@ test_215() { # for bugs 18102, 21079, 21517 # can we successfully write to lnet.stats? lctl set_param -n stats=0 || error "cannot write to lnet.stats" - sysctl -w lnet.stats=0 || error "cannot write to lnet.stats" } run_test 215 "lnet exists and has proper content - bugs 18102, 21079, 21517" @@ -12442,32 +12916,34 @@ test_216() { # bug 20317 "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 + # 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 - $DIRECTIO write $DIR/$tfile 0 10 4096 - $CHECKSTAT -s 40960 $DIR/$tfile + $DIRECTIO write $DIR/$tfile 0 10 4096 + $CHECKSTAT -s 40960 $DIR/$tfile - # disable lockless i/o - for node in $(osts_nodes); do - do_node $node 'lctl set_param -n ldlm.namespaces.filter-*.max_nolock_bytes 0; lctl set_param -n ldlm.namespaces.filter-*.contended_locks 32; lctl set_param -n ldlm.namespaces.filter-*.contention_seconds 0' - done - lctl set_param -n osc.*.contention_seconds 0 - clear_osc_stats + # disable lockless i/o + do_nodes $(comma_list $(osts_nodes)) \ + "lctl set_param -n ldlm.namespaces.filter-*.max_nolock_bytes=0 \ + ldlm.namespaces.filter-*.contended_locks=32 \ + ldlm.namespaces.filter-*.contention_seconds=0" + lctl set_param -n osc.*.contention_seconds=0 + clear_stats osc.*.osc_stats - dd if=/dev/zero of=$DIR/$tfile count=0 - $CHECKSTAT -s 0 $DIR/$tfile + dd if=/dev/zero of=$DIR/$tfile count=0 + $CHECKSTAT -s 0 $DIR/$tfile - restore_lustre_params <$p - rm -f $p - rm $DIR/$tfile + restore_lustre_params <$p + rm -f $p + rm $DIR/$tfile } -run_test 216 "check lockless direct write works and updates file size and kms correctly" +run_test 216 "check lockless direct write updates file size and kms correctly" test_217() { # bug 22430 [ $PARALLEL == "yes" ] && skip "skip parallel run" && return @@ -12477,8 +12953,8 @@ test_217() { # bug 22430 for node in $(nodes_list); do nid=$(host_nids_address $node $NETTYPE) if [[ $nid = *-* ]] ; then - echo "lctl ping $nid@$NETTYPE" - lctl ping $nid@$NETTYPE + echo "lctl ping $(h2nettype $nid)" + lctl ping $(h2nettype $nid) else echo "skipping $node (no hyphen detected)" fi @@ -12544,10 +13020,14 @@ test_220() { #LU-325 $LFS df -i + if ! combined_mgs_mds ; then + mount_mgs_client + fi + do_facet ost$((OSTIDX + 1)) lctl set_param fail_val=-1 #define OBD_FAIL_OST_ENOINO 0x229 do_facet ost$((OSTIDX + 1)) lctl set_param fail_loc=0x229 - do_facet mgs $LCTL pool_new $FSNAME.$TESTNAME || return 1 + create_pool $FSNAME.$TESTNAME || return 1 do_facet mgs $LCTL pool_add $FSNAME.$TESTNAME $OST || return 2 $SETSTRIPE $DIR/$tdir -i $OSTIDX -c 1 -p $FSNAME.$TESTNAME @@ -12574,10 +13054,16 @@ test_220() { #LU-325 do_facet ost$((OSTIDX + 1)) lctl set_param fail_val=0 do_facet ost$((OSTIDX + 1)) lctl set_param fail_loc=0 - do_facet mgs $LCTL pool_remove $FSNAME.$TESTNAME $OST || return 4 - do_facet mgs $LCTL pool_destroy $FSNAME.$TESTNAME || return 5 + do_facet mgs $LCTL pool_remove $FSNAME.$TESTNAME $OST || + error "$LCTL pool_remove $FSNAME.$TESTNAME $OST failed" + do_facet mgs $LCTL pool_destroy $FSNAME.$TESTNAME || + error "$LCTL pool_destroy $FSNAME.$TESTNAME failed" echo "unlink $MDSOBJS files @$next_id..." - unlinkmany $DIR/$tdir/f $MDSOBJS || return 6 + unlinkmany $DIR/$tdir/f $MDSOBJS || error "unlinkmany failed" + + if ! combined_mgs_mds ; then + umount_mgs_client + fi } run_test 220 "preallocated MDS objects still used if ENOSPC from OST" @@ -12596,25 +13082,25 @@ run_test 221 "make sure fault and truncate race to not cause OOM" test_222a () { [ $PARALLEL == "yes" ] && skip "skip parallel run" && return - rm -rf $DIR/$tdir - test_mkdir -p $DIR/$tdir - $SETSTRIPE -c 1 -i 0 $DIR/$tdir - createmany -o $DIR/$tdir/$tfile 10 - cancel_lru_locks mdc - cancel_lru_locks osc - #define OBD_FAIL_LDLM_AGL_DELAY 0x31a - $LCTL set_param fail_loc=0x31a - ls -l $DIR/$tdir > /dev/null || error "AGL for ls failed" - $LCTL set_param fail_loc=0 - rm -r $DIR/$tdir + rm -rf $DIR/$tdir + test_mkdir $DIR/$tdir + $LFS setstripe -c 1 -i 0 $DIR/$tdir + createmany -o $DIR/$tdir/$tfile 10 + cancel_lru_locks mdc + cancel_lru_locks osc + #define OBD_FAIL_LDLM_AGL_DELAY 0x31a + $LCTL set_param fail_loc=0x31a + ls -l $DIR/$tdir > /dev/null || error "AGL for ls failed" + $LCTL set_param fail_loc=0 + rm -r $DIR/$tdir } -run_test 222a "AGL for ls should not trigger CLIO lock failure ================" +run_test 222a "AGL for ls should not trigger CLIO lock failure" test_222b () { [ $PARALLEL == "yes" ] && skip "skip parallel run" && return rm -rf $DIR/$tdir - test_mkdir -p $DIR/$tdir - $SETSTRIPE -c 1 -i 0 $DIR/$tdir + test_mkdir $DIR/$tdir + $LFS setstripe -c 1 -i 0 $DIR/$tdir createmany -o $DIR/$tdir/$tfile 10 cancel_lru_locks mdc cancel_lru_locks osc @@ -12623,21 +13109,21 @@ test_222b () { rm -r $DIR/$tdir || error "AGL for rmdir failed" $LCTL set_param fail_loc=0 } -run_test 222b "AGL for rmdir should not trigger CLIO lock failure =============" +run_test 222b "AGL for rmdir should not trigger CLIO lock failure" test_223 () { [ $PARALLEL == "yes" ] && skip "skip parallel run" && return - rm -rf $DIR/$tdir - test_mkdir -p $DIR/$tdir - $SETSTRIPE -c 1 -i 0 $DIR/$tdir - createmany -o $DIR/$tdir/$tfile 10 - cancel_lru_locks mdc - cancel_lru_locks osc - #define OBD_FAIL_LDLM_AGL_NOLOCK 0x31b - $LCTL set_param fail_loc=0x31b - ls -l $DIR/$tdir > /dev/null || error "reenqueue failed" - $LCTL set_param fail_loc=0 - rm -r $DIR/$tdir + rm -rf $DIR/$tdir + test_mkdir $DIR/$tdir + $LFS setstripe -c 1 -i 0 $DIR/$tdir + createmany -o $DIR/$tdir/$tfile 10 + cancel_lru_locks mdc + cancel_lru_locks osc + #define OBD_FAIL_LDLM_AGL_NOLOCK 0x31b + $LCTL set_param fail_loc=0x31b + ls -l $DIR/$tdir > /dev/null || error "reenqueue failed" + $LCTL set_param fail_loc=0 + rm -r $DIR/$tdir } run_test 223 "osc reenqueue if without AGL lock granted =======================" @@ -12749,8 +13235,6 @@ test_225b () { skip_env "Need to mount OST to test" && return fi - [ $MDSCOUNT -ge 2 ] && - skip "skipping now for more than one MDT" && return local mds=$(facet_host $SINGLEMDS) local target=$(do_nodes $mds 'lctl dl' | \ awk "{if (\$2 == \"UP\" && \$3 == \"mdt\") {print \$4}}") @@ -12852,7 +13336,7 @@ test_228a() { [ $PARALLEL == "yes" ] && skip "skip parallel run" && return remote_mds_nodsh && skip "remote MDS with nodsh" && return [ "$(facet_fstype $SINGLEMDS)" != "ldiskfs" ] && - skip "non-ldiskfs backend" && return + skip "ldiskfs only test" && return local MDT_DEV=$(mdsdevname ${SINGLEMDS//mds/}) local myDIR=$DIR/$tdir @@ -12894,7 +13378,7 @@ test_228b() { [ $PARALLEL == "yes" ] && skip "skip parallel run" && return remote_mds_nodsh && skip "remote MDS with nodsh" && return [ "$(facet_fstype $SINGLEMDS)" != "ldiskfs" ] && - skip "non-ldiskfs backend" && return + skip "ldiskfs only test" && return local MDT_DEV=$(mdsdevname ${SINGLEMDS//mds/}) local myDIR=$DIR/$tdir @@ -12944,7 +13428,7 @@ test_228c() { [ $PARALLEL == "yes" ] && skip "skip parallel run" && return remote_mds_nodsh && skip "remote MDS with nodsh" && return [ "$(facet_fstype $SINGLEMDS)" != "ldiskfs" ] && - skip "non-ldiskfs backend" && return + skip "ldiskfs only test" && return local MDT_DEV=$(mdsdevname ${SINGLEMDS//mds/}) local myDIR=$DIR/$tdir @@ -13446,8 +13930,12 @@ test_231a() # is the same across all OSCs local max_pages=$($LCTL get_param -n osc.*.max_pages_per_rpc | head -n1) local bulk_size=$((max_pages * 4096)) + local brw_size=$(do_facet ost1 $LCTL get_param -n obdfilter.*.brw_size | + head -n 1) mkdir -p $DIR/$tdir + $LFS setstripe -S ${brw_size}M $DIR/$tdir || + error "failed to set stripe with -S ${brw_size}M option" # clear the OSC stats $LCTL set_param osc.*.stats=0 &>/dev/null @@ -13583,7 +14071,7 @@ run_test 235 "LU-1715: flock deadlock detection does not work properly" #LU-2935 test_236() { check_swap_layouts_support && return 0 - test_mkdir -p -c1 $DIR/$tdir || error "mkdir $tdir failed" + test_mkdir -c1 $DIR/$tdir local ref1=/etc/passwd local ref2=/etc/group @@ -13653,9 +14141,10 @@ test_239() { mkdir -p $DIR/$tdir createmany -o $DIR/$tdir/f- 5000 unlinkmany $DIR/$tdir/f- 5000 - do_nodes $list "lctl set_param -n osp*.*.sync_changes 1" - changes=$(do_nodes $list "lctl get_param -n osc.*MDT*.sync_changes \ - osc.*MDT*.sync_in_flight" | calc_sum) + [ $(lustre_version_code $SINGLEMDS) -gt $(version_code 2.10.53) ] && + do_nodes $list "lctl set_param -n osp.*.force_sync=1" + changes=$(do_nodes $list "lctl get_param -n osp.*MDT*.sync_changes \ + osp.*MDT*.sync_in_flight" | calc_sum) [ "$changes" -eq 0 ] || error "$changes not synced" } run_test 239 "osp_sync test" @@ -13708,7 +14197,7 @@ run_test 240 "race between ldlm enqueue and the connection RPC (no ASSERT)" test_241_bio() { for LOOP in $(seq $1); do dd if=$DIR/$tfile of=/dev/null bs=40960 count=1 2>/dev/null - cancel_lru_locks osc || true + cancel_lru_locks $OSC || true done } @@ -13722,7 +14211,7 @@ test_241_dio() { test_241a() { # was test_241 dd if=/dev/zero of=$DIR/$tfile count=1 bs=40960 ls -la $DIR/$tfile - cancel_lru_locks osc + cancel_lru_locks $OSC test_241_bio 1000 & PID=$! test_241_dio 1000 @@ -13756,14 +14245,14 @@ run_test 242 "mdt_readpage failure should not cause directory unreadable" test_243() { - test_mkdir -p $DIR/$tdir + test_mkdir $DIR/$tdir group_lock_test -d $DIR/$tdir || error "A group lock test failed" } run_test 243 "various group lock tests" test_244() { - test_mkdir -p $DIR/$tdir + test_mkdir $DIR/$tdir dd if=/dev/zero of=$DIR/$tdir/$tfile bs=1M count=35 sendfile_grouplock $DIR/$tdir/$tfile || \ error "sendfile+grouplock failed" @@ -14000,7 +14489,7 @@ test_252() { remote_ost_nodsh && skip "remote OST with nodsh" && return if [ "$(facet_fstype ost1)" != "ldiskfs" -o \ "$(facet_fstype mds1)" != "ldiskfs" ]; then - skip "can only run lr_reader on ldiskfs target" + skip "ldiskfs only test" return fi @@ -14076,8 +14565,8 @@ test_253() { remote_mds_nodsh && skip "remote MDS with nodsh" && return remote_mgs_nodsh && skip "remote MGS with nodsh" && return - local ost_name=$($LFS osts | grep ${ostidx}": " | \ - awk '{print $2}' | sed -e 's/_UUID$//') + local ost_name=$($LFS osts | + sed -n 's/^'$ostidx': \(.*\)_UUID .*/\1/p') # on the mdt's osc local mdtosc_proc1=$(get_mdtosc_proc_path $SINGLEMDS $ost_name) do_facet $SINGLEMDS $LCTL get_param -n \ @@ -14096,8 +14585,10 @@ test_253() { osp.$mdtosc_proc1.reserved_mb_low) echo "prev high watermark $last_wm_h, prev low watermark $last_wm_l" - do_facet mgs $LCTL pool_new $FSNAME.$TESTNAME || - error "Pool creation failed" + if ! combined_mgs_mds ; then + mount_mgs_client + fi + create_pool $FSNAME.$TESTNAME || error "Pool creation failed" do_facet mgs $LCTL pool_add $FSNAME.$TESTNAME $ost_name || error "Adding $ost_name to pool failed" @@ -14161,6 +14652,10 @@ test_253() { error "Remove $ost_name from pool failed" do_facet mgs $LCTL pool_destroy $FSNAME.$TESTNAME || error "Pool destroy fialed" + + if ! combined_mgs_mds ; then + umount_mgs_client + fi } run_test 253 "Check object allocation limit" @@ -14235,68 +14730,94 @@ ladvise_no_ioctl() return 1 } -ladvise_willread_performance() -{ - local repeat=10 - local average_cache=0 - local average_ladvise=0 +percent() { + bc <<<"scale=2; ($1 - $2) * 100 / $2" +} - for ((i = 1; i <= $repeat; i++)); do +# run a random read IO workload +# usage: random_read_iops +random_read_iops() { + local file=$1 + local fsize=$2 + local iosize=${3:-4096} + + $READS -f $file -s $fsize -b $iosize -n $((fsize / iosize)) -t 60 | + sed -e '/^$/d' -e 's#.*s, ##' -e 's#MB/s##' +} + +drop_file_oss_cache() { + local file="$1" + local nodes="$2" + + $LFS ladvise -a dontneed $file 2>/dev/null || + do_nodes $nodes "echo 3 > /proc/sys/vm/drop_caches" +} + +ladvise_willread_performance() +{ + local repeat=10 + local average_origin=0 + local average_cache=0 + local average_ladvise=0 + + for ((i = 1; i <= $repeat; i++)); do echo "Iter $i/$repeat: reading without willread hint" cancel_lru_locks osc - do_nodes $(comma_list $(osts_nodes)) \ - "echo 3 > /proc/sys/vm/drop_caches" - local speed_origin=$($READS -f $DIR/$tfile -s $size \ - -b 4096 -n $((size / 4096)) -t 60 | - sed -e '/^$/d' -e 's#.*s, ##' -e 's#MB/s##') + drop_file_oss_cache $DIR/$tfile $(comma_list $(osts_nodes)) + local speed_origin=$(random_read_iops $DIR/$tfile $size) + echo "Iter $i/$repeat: uncached speed: $speed_origin" + average_origin=$(bc <<<"$average_origin + $speed_origin") - echo "Iter $i/$repeat: Reading again without willread hint" cancel_lru_locks osc - local speed_cache=$($READS -f $DIR/$tfile -s $size \ - -b 4096 -n $((size / 4096)) -t 60 | - sed -e '/^$/d' -e 's#.*s, ##' -e 's#MB/s##') + local speed_cache=$(random_read_iops $DIR/$tfile $size) + echo "Iter $i/$repeat: OSS cache speed: $speed_cache" + average_cache=$(bc <<<"$average_cache + $speed_cache") - echo "Iter $i/$repeat: reading with willread hint" cancel_lru_locks osc - do_nodes $(comma_list $(osts_nodes)) \ - "echo 3 > /proc/sys/vm/drop_caches" - lfs ladvise -a willread $DIR/$tfile || - error "Ladvise failed" - local speed_ladvise=$($READS -f $DIR/$tfile -s $size \ - -b 4096 -n $((size / 4096)) -t 60 | - sed -e '/^$/d' -e 's#.*s, ##' -e 's#MB/s##') - - local cache_speedup=$(echo "scale=2; \ - ($speed_cache-$speed_origin)/$speed_origin*100" | bc) - cache_speedup=$(echo ${cache_speedup%.*}) - echo "Iter $i/$repeat: cache speedup: $cache_speedup%" - average_cache=$((average_cache + cache_speedup)) - - local ladvise_speedup=$(echo "scale=2; \ - ($speed_ladvise-$speed_origin)/$speed_origin*100" | bc) - ladvise_speedup=$(echo ${ladvise_speedup%.*}) - echo "Iter $i/$repeat: ladvise speedup: $ladvise_speedup%" - average_ladvise=$((average_ladvise + ladvise_speedup)) - done - average_cache=$((average_cache / repeat)) - average_ladvise=$((average_ladvise / repeat)) - - if [ $average_cache -lt 20 ]; then - echo "Speedup with cache is less than 20% ($average_cache%),"\ - "skipping check of speedup with willread:"\ - "$average_ladvise%" + drop_file_oss_cache $DIR/$tfile $(comma_list $(osts_nodes)) + $LFS ladvise -a willread $DIR/$tfile || error "ladvise failed" + local speed_ladvise=$(random_read_iops $DIR/$tfile $size) + echo "Iter $i/$repeat: ladvise speed: $speed_ladvise" + average_ladvise=$(bc <<<"$average_ladvise + $speed_ladvise") + done + average_origin=$(bc <<<"scale=2; $average_origin / $repeat") + average_cache=$(bc <<<"scale=2; $average_cache / $repeat") + average_ladvise=$(bc <<<"scale=2; $average_ladvise / $repeat") + + speedup_cache=$(percent $average_cache $average_origin) + speedup_ladvise=$(percent $average_ladvise $average_origin) + + echo "Average uncached read: $average_origin" + echo "Average speedup with OSS cached read: " \ + "$average_cache = +$speedup_cache%" + echo "Average speedup with ladvise willread: " \ + "$average_ladvise = +$speedup_ladvise%" + + local lowest_speedup=20 + if [ ${average_cache%.*} -lt $lowest_speedup ]; then + echo "Speedup with OSS cached read less than $lowest_speedup%, " + "got $average_cache%. Skipping ladvise willread check." return 0 fi - local lowest_speedup=$((average_cache / 2)) - [ $average_ladvise -gt $lowest_speedup ] || + # the test won't work on ZFS until it supports 'ladvise dontneed', but + # it is still good to run until then to exercise 'ladvise willread' + ! $LFS ladvise -a dontneed $DIR/$tfile && + [ "$(facet_fstype ost1)" = "zfs" ] && + echo "osd-zfs does not support dontneed or drop_caches" && + return 0 + + lowest_speedup=$(bc <<<"scale=2; $average_cache / 2") + [ ${average_ladvise%.*} -gt $lowest_speedup ] || error_not_in_vm "Speedup with willread is less than " \ - "$lowest_speedup%, got $average_ladvise%" - echo "Speedup with willread ladvise: $average_ladvise%" - echo "Speedup with cache: $average_cache%" + "$lowest_speedup%, got $average_ladvise%" } test_255a() { + [ $(lustre_version_code ost1) -lt $(version_code 2.8.54) ] && + skip "lustre < 2.8.54 does not support ladvise " && return + remote_ost_nodsh && skip "remote OST with nodsh" && return + lfs setstripe -c -1 -i 0 $DIR/$tfile || error "$tfile failed" ladvise_no_type willread $DIR/$tfile && @@ -14305,9 +14826,6 @@ test_255a() { ladvise_no_ioctl $DIR/$tfile && skip "ladvise ioctl is not supported" && return - [ $(lustre_version_code ost1) -lt $(version_code 2.8.54) ] && - skip "lustre < 2.8.54 does not support ladvise " && return - local size_mb=100 local size=$((size_mb * 1048576)) dd if=/dev/zero of=$DIR/$tfile bs=1048576 count=$size_mb || @@ -14383,6 +14901,10 @@ facet_meminfo() { } test_255b() { + remote_ost_nodsh && skip "remote OST with nodsh" && return + + lfs setstripe -c 1 -i 0 $DIR/$tfile + ladvise_no_type dontneed $DIR/$tfile && skip "dontneed ladvise is not supported" && return @@ -14392,10 +14914,9 @@ test_255b() { [ $(lustre_version_code ost1) -lt $(version_code 2.8.54) ] && skip "lustre < 2.8.54 does not support ladvise" && return - [ "$(facet_fstype ost1)" = "zfs" ] && - skip "zfs-osd does not support dontneed advice" && return - - lfs setstripe -c 1 -i 0 $DIR/$tfile + ! $LFS ladvise -a dontneed $DIR/$tfile && + [ "$(facet_fstype ost1)" = "zfs" ] && + skip "zfs-osd does not support 'ladvise dontneed'" && return local size_mb=100 local size=$((size_mb * 1048576)) @@ -14440,6 +14961,87 @@ test_255b() { } run_test 255b "check 'lfs ladvise -a dontneed'" +test_255c() { + local count + local new_count + local difference + local i + local rc + test_mkdir -p $DIR/$tdir + $SETSTRIPE -i 0 $DIR/$tdir + + #test 10 returns only success/failure + i=10 + lockahead_test -d $DIR/$tdir -t $i + rc=$? + if [ $rc -eq 255 ]; then + error "Ladvise test${i} failed, ${rc}" + fi + + #test 11 counts lock enqueue requests, all others count new locks + i=11 + count=$(do_facet ost1 \ + $LCTL get_param -n ost.OSS.ost.stats) + count=$(echo "$count" | grep ldlm_extent_enqueue | awk '{ print $2 }') + + lockahead_test -d $DIR/$tdir -t $i + rc=$? + if [ $rc -eq 255 ]; then + error "Ladvise test${i} failed, ${rc}" + fi + + new_count=$(do_facet ost1 \ + $LCTL get_param -n ost.OSS.ost.stats) + new_count=$(echo "$new_count" | grep ldlm_extent_enqueue | \ + awk '{ print $2 }') + + difference="$((new_count - count))" + if [ $difference -ne $rc ]; then + error "Ladvise test${i}, bad enqueue count, returned " \ + "${rc}, actual ${difference}" + fi + + for i in $(seq 12 21); do + # If we do not do this, we run the risk of having too many + # locks and starting lock cancellation while we are checking + # lock counts. + cancel_lru_locks osc + + count=$($LCTL get_param -n \ + ldlm.namespaces.$FSNAME-OST0000*osc-f*.lock_unused_count) + + lockahead_test -d $DIR/$tdir -t $i + rc=$? + if [ $rc -eq 255 ]; then + error "Ladvise test ${i} failed, ${rc}" + fi + + new_count=$($LCTL get_param -n \ + ldlm.namespaces.$FSNAME-OST0000*osc-f*.lock_unused_count) + difference="$((new_count - count))" + + # Test 15 output is divided by 1000 to map down to valid return + if [ $i -eq 15 ]; then + rc="$((rc * 1000))" + fi + + if [ $difference -ne $rc ]; then + error "Ladvise test ${i}, bad lock count, returned " \ + "${rc}, actual ${difference}" + fi + done + + #test 22 returns only success/failure + i=22 + lockahead_test -d $DIR/$tdir -t $i + rc=$? + if [ $rc -eq 255 ]; then + error "Ladvise test${i} failed, ${rc}" + fi + +} +run_test 255c "suite of ladvise lockahead tests" + test_256() { local cl_user local cat_sl @@ -14448,7 +15050,7 @@ test_256() { [ $PARALLEL == "yes" ] && skip "skip parallel run" && return remote_mds_nodsh && skip "remote MDS with nodsh" && return [ "$(facet_fstype mds1)" != "ldiskfs" ] && - skip "non-ldiskfs backend" && return + skip "ldiskfs only test" && return mdt_dev=$(mdsdevname 1) echo $mdt_dev @@ -14478,9 +15080,12 @@ test_256() { #after mount new plainllog is used touch $DIR/$tdir/{11..19} + do_facet mds1 sync + local TEMP256FILE=$(mktemp -u TEMP256XXXXXX) cat_sl=$(do_facet mds1 \ - "$DEBUGFS -R \\\"dump changelog_catalog cat.dmp\\\" $mdt_dev; \ - llog_reader cat.dmp | grep \\\"type=1064553b\\\" | wc -l") + "$DEBUGFS -R \\\"dump changelog_catalog $TEMP256FILE\\\" $mdt_dev; \ + llog_reader $TEMP256FILE | grep \\\"type=1064553b\\\" | wc -l") + do_facet mds1 rm $TEMP256FILE if (( cat_sl != 2 )); then do_facet mds1 $LCTL --device $MDT0 changelog_deregister $cl_user @@ -14489,9 +15094,12 @@ test_256() { $LFS changelog_clear $MDT0 $cl_user 0 + do_facet mds1 sync + TEMP256FILE=$(mktemp -u TEMP256XXXXXX) cat_sl=$(do_facet mds1 \ - "$DEBUGFS -R \\\"dump changelog_catalog cat.dmp\\\" $mdt_dev; \ - llog_reader cat.dmp | grep \\\"type=1064553b\\\" | wc -l") + "$DEBUGFS -R \\\"dump changelog_catalog $TEMP256FILE\\\" $mdt_dev; \ + llog_reader $TEMP256FILE | grep \\\"type=1064553b\\\" | wc -l") + do_facet mds1 rm $TEMP256FILE do_facet mds1 $LCTL --device $MDT0 changelog_deregister $cl_user @@ -14509,7 +15117,7 @@ test_257() { [[ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.8.55) ]] && skip "Need MDS version at least 2.8.55" && return - test_mkdir -p $DIR/$tdir + test_mkdir $DIR/$tdir setfattr -n trusted.name1 -v value1 $DIR/$tdir || error "setfattr -n trusted.name1=value1 $DIR/$tdir failed" @@ -14527,6 +15135,39 @@ test_257() { } run_test 257 "xattr locks are not lost" +# Verify we take the i_mutex when security requires it +test_258a() { +#define OBD_FAIL_IMUTEX_SEC 0x141c + $LCTL set_param fail_loc=0x141c + touch $DIR/$tfile + chmod u+s $DIR/$tfile + chmod a+rwx $DIR/$tfile + $RUNAS dd if=/dev/zero of=$DIR/$tfile bs=4k count=1 oflag=append + RC=$? + if [ $RC -ne 0 ]; then + error "error, failed to take i_mutex, rc=$?" + fi + rm -f $DIR/$tfile +} +run_test 258a + +# Verify we do NOT take the i_mutex in the normal case +test_258b() { +#define OBD_FAIL_IMUTEX_NOSEC 0x141d + $LCTL set_param fail_loc=0x141d + touch $DIR/$tfile + chmod a+rwx $DIR + chmod a+rw $DIR/$tfile + $RUNAS dd if=/dev/zero of=$DIR/$tfile bs=4k count=1 oflag=append + RC=$? + if [ $RC -ne 0 ]; then + error "error, took i_mutex unnecessarily, rc=$?" + fi + rm -f $DIR/$tfile + +} +run_test 258b "verify i_mutex security behavior" + test_260() { #define OBD_FAIL_MDC_CLOSE 0x806 $LCTL set_param fail_loc=0x80000806 @@ -14535,6 +15176,370 @@ test_260() { } run_test 260 "Check mdc_close fail" +### Data-on-MDT sanity tests ### +test_270a() { + # create DoM file + local dom=$DIR/$tdir/dom_file + local tmp=$DIR/$tdir/tmp_file + + mkdir -p $DIR/$tdir + + # basic checks for DoM component creation + $LFS setstripe -E 1024K -E 1024K -L mdt $dom 2>/dev/null && + error "Can set MDT layout to non-first entry" + + $LFS setstripe -E 1024K -L mdt -E 1024K -L mdt $dom 2>/dev/null && + error "Can define multiple entries as MDT layout" + + $LFS setstripe -E 1M -L mdt $dom || + error "Can't create DoM layout" + + [ $($LFS getstripe -L $dom) == 100 ] || error "bad pattern" + [ $($LFS getstripe -c $dom) == 0 ] || error "bad stripe count" + [ $($LFS getstripe -S $dom) == 1048576 ] || error "bad stripe size" + + local mdtidx=$($GETSTRIPE -M $dom) + local mdtname=MDT$(printf %04x $mdtidx) + local facet=mds$((mdtidx + 1)) + local space_check=1 + + # Skip free space checks with ZFS + if [ "$(facet_fstype $facet)" == "zfs" ]; then + space_check=0 + fi + + # write + sync + local mdtfree1=$(do_facet $facet \ + lctl get_param -n osd*.*$mdtname.kbytesfree) + dd if=/dev/urandom of=$tmp bs=1024 count=100 + # check also direct IO along write + dd if=$tmp of=$dom bs=102400 count=1 oflag=direct + sync + cmp $tmp $dom || error "file data is different" + [ $(stat -c%s $dom) == 102400 ] || error "bad size after write" + if [ $space_check == 1 ]; then + local mdtfree2=$(do_facet $facet \ + lctl get_param -n osd*.*$mdtname.kbytesfree) + [ $(($mdtfree1 - $mdtfree2)) -ge 102 ] || + error "MDT free space is wrong after write" + fi + + # truncate + $TRUNCATE $dom 10000 + [ $(stat -c%s $dom) == 10000 ] || error "bad size after truncate" + if [ $space_check == 1 ]; then + mdtfree1=$(do_facet $facet \ + lctl get_param -n osd*.*$mdtname.kbytesfree) + [ $(($mdtfree1 - $mdtfree2)) -ge 92 ] || + error "MDT free space is wrong after truncate" + fi + + # append + cat $tmp >> $dom + sync + [ $(stat -c%s $dom) == 112400 ] || error "bad size after append" + if [ $space_check == 1 ]; then + mdtfree2=$(do_facet $facet \ + lctl get_param -n osd*.*$mdtname.kbytesfree) + [ $(($mdtfree1 - $mdtfree2)) -ge 102 ] || + error "MDT free space is wrong after append" + fi + + # delete + rm $dom + if [ $space_check == 1 ]; then + mdtfree1=$(do_facet $facet \ + lctl get_param -n osd*.*$mdtname.kbytesfree) + [ $(($mdtfree1 - $mdtfree2)) -ge 112 ] || + error "MDT free space is wrong after removal" + fi + + # combined striping + $LFS setstripe -E 1024K -L mdt -E EOF $dom || + error "Can't create DoM + OST striping" + + dd if=/dev/urandom of=$tmp bs=1024 count=2000 + # check also direct IO along write + dd if=$tmp of=$dom bs=102400 count=20 oflag=direct + sync + cmp $tmp $dom || error "file data is different" + [ $(stat -c%s $dom) == 2048000 ] || error "bad size after write" + rm $dom + rm $tmp + + return 0 +} +run_test 270a "DoM: basic functionality tests" + +test_270b() { + local dom=$DIR/$tdir/dom_file + local max_size=1048576 + + mkdir -p $DIR/$tdir + $LFS setstripe -E $max_size -L mdt $dom + + # truncate over the limit + $TRUNCATE $dom $(($max_size + 1)) && + error "successful truncate over the maximum size" + # write over the limit + dd if=/dev/zero of=$dom bs=$max_size seek=1 count=1 && + error "successful write over the maximum size" + # append over the limit + dd if=/dev/zero of=$dom bs=$(($max_size - 3)) count=1 + echo "12345" >> $dom && error "successful append over the maximum size" + rm $dom + + return 0 +} +run_test 270b "DoM: maximum size overflow checks for DoM-only file" + +test_270c() { + mkdir -p $DIR/$tdir + $LFS setstripe -E 1024K -L mdt $DIR/$tdir + + # check files inherit DoM EA + touch $DIR/$tdir/first + [ $($GETSTRIPE -L $DIR/$tdir/first) == 100 ] || + error "bad pattern" + [ $($LFS getstripe -c $DIR/$tdir/first) == 0 ] || + error "bad stripe count" + [ $($LFS getstripe -S $DIR/$tdir/first) == 1048576 ] || + error "bad stripe size" + + # check directory inherits DoM EA and uses it as default + mkdir $DIR/$tdir/subdir + touch $DIR/$tdir/subdir/second + [ $($LFS getstripe -L $DIR/$tdir/subdir/second) == 100 ] || + error "bad pattern in sub-directory" + [ $($LFS getstripe -c $DIR/$tdir/subdir/second) == 0 ] || + error "bad stripe count in sub-directory" + [ $($LFS getstripe -S $DIR/$tdir/subdir/second) == 1048576 ] || + error "bad stripe size in sub-directory" + return 0 +} +run_test 270c "DoM: DoM EA inheritance tests" + +test_270d() { + mkdir -p $DIR/$tdir + $LFS setstripe -E 1024K -L mdt $DIR/$tdir + + # inherit default DoM striping + mkdir $DIR/$tdir/subdir + touch $DIR/$tdir/subdir/f1 + + # change default directory striping + $LFS setstripe -c 1 $DIR/$tdir/subdir + touch $DIR/$tdir/subdir/f2 + [ $($LFS getstripe -c $DIR/$tdir/subdir/f2) == 1 ] || + error "wrong default striping in file 2" + [ $($LFS getstripe -L $DIR/$tdir/subdir/f2) == 1 ] || + error "bad pattern in file 2" + return 0 +} +run_test 270d "DoM: change striping from DoM to RAID0" + +test_270e() { + mkdir -p $DIR/$tdir/dom + mkdir -p $DIR/$tdir/norm + DOMFILES=20 + NORMFILES=10 + $LFS setstripe -E 1M -L mdt $DIR/$tdir/dom + $LFS setstripe -i 0 -S 2M $DIR/$tdir/norm + + createmany -o $DIR/$tdir/dom/dom- $DOMFILES + createmany -o $DIR/$tdir/norm/norm- $NORMFILES + + # find DoM files by layout + NUM=$($LFIND -L mdt -type f $DIR/$tdir 2>/dev/null | wc -l) + [ $NUM -eq $DOMFILES ] || + error "lfs find -L: found $NUM, expected $DOMFILES" + echo "Test 1: lfs find 20 DOM files by layout: OK" + + # there should be 1 dir with default DOM striping + NUM=$($LFIND -L mdt -type d $DIR/$tdir 2>/dev/null | wc -l) + [ $NUM -eq 1 ] || + error "lfs find -L: found $NUM, expected 1 dir" + echo "Test 2: lfs find 1 DOM dir by layout: OK" + + # find DoM files by stripe size + NUM=$($LFIND -S -1200K -type f $DIR/$tdir 2>/dev/null | wc -l) + [ $NUM -eq $DOMFILES ] || + error "lfs find -S: found $NUM, expected $DOMFILES" + echo "Test 4: lfs find 20 DOM files by stripe size: OK" + + # find files by stripe offset except DoM files + NUM=$($LFIND -i 0 -type f $DIR/$tdir 2>/dev/null | wc -l) + [ $NUM -eq $NORMFILES ] || + error "lfs find -i: found $NUM, expected $NORMFILES" + echo "Test 5: lfs find no DOM files by stripe index: OK" + return 0 +} +run_test 270e "DoM: lfs find with DoM files test" + +test_270f() { + local mdtname=${FSNAME}-MDT0000-mdtlov + local dom=$DIR/$tdir/dom_file + local dom_limit_saved=$(do_facet mds1 $LCTL get_param -n \ + lod.$mdtname.dom_stripesize) + local dom_limit=131072 + + do_facet mds1 $LCTL set_param -n lod.$mdtname.dom_stripesize=$dom_limit + local dom_current=$(do_facet mds1 $LCTL get_param -n \ + lod.$mdtname.dom_stripesize) + [ ${dom_limit} -eq ${dom_current} ] || + error "Cannot change per-MDT DoM stripe limit to $dom_limit" + + $LFS mkdir -i 0 -c 1 $DIR/$tdir + $LFS setstripe -d $DIR/$tdir + $LFS setstripe -E $dom_limit -L mdt $DIR/$tdir || + error "Can't set directory default striping" + + # exceed maximum stripe size + $LFS setstripe -E $(($dom_limit * 2)) -L mdt $dom && + error "Able to create DoM component size more than LOD limit" + + do_facet mds1 $LCTL set_param -n lod.$mdtname.dom_stripesize=0 + dom_current=$(do_facet mds1 $LCTL get_param -n \ + lod.$mdtname.dom_stripesize) + [ 0 -eq ${dom_current} ] || + error "Can't set zero DoM stripe limit" + + # too low values to be aligned with smallest stripe size 64K + do_facet mds1 $LCTL set_param -n lod.$mdtname.dom_stripesize=30000 + dom_current=$(do_facet mds1 $LCTL get_param -n \ + lod.$mdtname.dom_stripesize) + [ 30000 -eq ${dom_current} ] && + error "Can set too small DoM stripe limit" + + do_facet mds1 $LCTL set_param -n lod.$mdtname.dom_stripesize=2147483648 + dom_current=$(do_facet mds1 $LCTL get_param -n \ + lod.$mdtname.dom_stripesize) + echo $dom_current + [ 2147483648 -eq ${dom_current} ] && + error "Can set too large DoM stripe limit" + + do_facet mds1 $LCTL set_param -n \ + lod.$mdtname.dom_stripesize=$((dom_limit * 2)) + $LFS setstripe -E $((dom_limit * 2)) -L mdt $dom || + error "Can't create DoM component size after limit change" + do_facet mds1 $LCTL set_param -n \ + lod.$mdtname.dom_stripesize=$((dom_limit / 2)) + $LFS setstripe -E $dom_limit -L mdt ${dom}_big && + error "Can create big DoM component after limit decrease" + touch ${dom}_def || + error "Can't create file with old default layout" + + do_facet mds1 $LCTL set_param -n lod.*.dom_stripesize=$dom_limit_saved + return 0 +} +run_test 270f "DoM: maximum DoM stripe size checks" + +test_271a() { + local dom=$DIR/$tdir/dom + + mkdir -p $DIR/$tdir + + $LFS setstripe -E 1024K -L mdt $dom + + lctl set_param -n mdc.*.stats=clear + dd if=/dev/zero of=$dom bs=4096 count=1 || return 1 + cat $dom > /dev/null + local reads=$(lctl get_param -n mdc.*.stats | + awk '/ost_read/ {print $2}') + [ -z $reads ] || error "Unexpected $reads READ RPCs" + ls $dom + rm -f $dom +} +run_test 271a "DoM: data is cached for read after write" + +test_271b() { + local dom=$DIR/$tdir/dom + + mkdir -p $DIR/$tdir + + $LFS setstripe -E 1024K -L mdt -E EOF $dom + + lctl set_param -n mdc.*.stats=clear + dd if=/dev/zero of=$dom bs=4096 count=1 || return 1 + cancel_lru_locks mdc + $CHECKSTAT -t file -s 4096 $dom || error "stat #1 fails" + # second stat to check size is cached on client + $CHECKSTAT -t file -s 4096 $dom || error "stat #2 fails" + local gls=$(lctl get_param -n mdc.*.stats | + awk '/ldlm_glimpse/ {print $2}') + [ -z $gls ] || error "Unexpected $gls glimpse RPCs" + rm -f $dom +} +run_test 271b "DoM: no glimpse RPC for stat (DoM only file)" + +test_271ba() { + local dom=$DIR/$tdir/dom + + mkdir -p $DIR/$tdir + + $LFS setstripe -E 1024K -L mdt -E EOF $dom + + lctl set_param -n mdc.*.stats=clear + lctl set_param -n osc.*.stats=clear + dd if=/dev/zero of=$dom bs=2048K count=1 || return 1 + cancel_lru_locks mdc + $CHECKSTAT -t file -s 2097152 $dom || error "stat" + # second stat to check size is cached on client + $CHECKSTAT -t file -s 2097152 $dom || error "stat" + local gls=$(lctl get_param -n mdc.*.stats | + awk '/ldlm_glimpse/ {print $2}') + [ -z $gls ] || error "Unexpected $gls glimpse RPCs" + local gls=$(lctl get_param -n osc.*.stats | + awk '/ldlm_glimpse/ {print $2}') + [ -z $gls ] || error "Unexpected $gls OSC glimpse RPCs" + rm -f $dom +} +run_test 271ba "DoM: no glimpse RPC for stat (combined file)" + +test_271c() { + # test to be enabled with lock_convert + skip "skipped until lock convert will be implemented" && return + + local dom=$DIR/$tdir/dom + + mkdir -p $DIR/$tdir + + $LFS setstripe -E 1024K -L mdt $DIR/$tdir + + local mdtidx=$($LFS getstripe -M $DIR/$tdir) + local facet=mds$((mdtidx + 1)) + + cancel_lru_locks mdc + do_facet $facet lctl set_param -n mdt.*.dom_lock=0 + createmany -o $dom 1000 + lctl set_param -n mdc.*.stats=clear + smalliomany -w $dom 1000 200 + lctl get_param -n mdc.*.stats + local enq=$(lctl get_param -n mdc.*.stats | + awk '/ldlm_ibits_enqueue/ {print $2}') + # Each file has 1 open, 1 IO enqueues, total 2000 + # but now we have also +1 getxattr for security.capability, total 3000 + [ $enq -ge 2000 ] || error "Too few enqueues $enq, expected > 2000" + unlinkmany $dom 1000 + + cancel_lru_locks mdc + do_facet $facet lctl set_param -n mdt.*.dom_lock=1 + createmany -o $dom 1000 + lctl set_param -n mdc.*.stats=clear + smalliomany -w $dom 1000 200 + lctl get_param -n mdc.*.stats + local enq_2=$(lctl get_param -n mdc.*.stats | + awk '/ldlm_ibits_enqueue/ {print $2}') + # Expect to see reduced amount of RPCs by 1000 due to single enqueue + # for OPEN and IO lock. + [ $((enq - enq_2)) -ge 1000 ] || + error "Too many enqueues $enq_2, expected about $((enq - 1000))" + unlinkmany $dom 1000 + return 0 +} +run_test 271c "DoM: IO lock at open saves enqueue RPCs" + cleanup_test_300() { trap 0 umask $SAVE_UMASK @@ -14560,7 +15565,11 @@ test_striped_dir() { error "getdirstripe failed" stripe_count=$($LFS getdirstripe -c $DIR/$tdir/striped_dir) if [ "$stripe_count" != "2" ]; then - error "stripe_count is $stripe_count, expect 2" + error "1:stripe_count is $stripe_count, expect 2" + fi + stripe_count=$($LFS getdirstripe -T $DIR/$tdir/striped_dir) + if [ "$stripe_count" != "2" ]; then + error "2:stripe_count is $stripe_count, expect 2" fi stripe_index=$($LFS getdirstripe -i $DIR/$tdir/striped_dir) @@ -14624,7 +15633,7 @@ test_300b() { local mtime2 local mtime3 - test_mkdir $DIR/$tdir || error "mkdir fail" + test_mkdir $DIR/$tdir $LFS setdirstripe -i 0 -c 2 -t all_char $DIR/$tdir/striped_dir || error "set striped dir error" for ((i=0; i<10; i++)); do @@ -14726,13 +15735,13 @@ test_300e() { mkdir $DIR/$tdir/striped_dir/dir_c $LFS setdirstripe -i 0 -c 2 -t all_char $DIR/$tdir/striped_dir/stp_a || - error "set striped dir under striped dir error" + error "set striped adir under striped dir error" - $LFS setdirstripe -i 0 -c 2 -t all_char $DIR/$tdir/striped_dir/stp_b || - error "set striped dir under striped dir error" + $LFS setdirstripe -i 0 -c 2 -H all_char $DIR/$tdir/striped_dir/stp_b || + error "set striped bdir under striped dir error" $LFS setdirstripe -i 0 -c 2 -t all_char $DIR/$tdir/striped_dir/stp_c || - error "set striped dir under striped dir error" + error "set striped cdir under striped dir error" mrename $DIR/$tdir/striped_dir/dir_a $DIR/$tdir/striped_dir/dir_b || error "rename dir under striped dir fails" @@ -14850,6 +15859,24 @@ test_300g() { mkdir $DIR/$tdir mkdir $DIR/$tdir/normal_dir + #Checking when client cache stripe index + $LFS setdirstripe -c$MDSCOUNT $DIR/$tdir/striped_dir + $LFS setdirstripe -D -i1 $DIR/$tdir/striped_dir || + error "create striped_dir failed" + + $LFS setdirstripe -i0 $DIR/$tdir/striped_dir/dir0 || + error "create dir0 fails" + stripe_index=$($LFS getdirstripe -i $DIR/$tdir/striped_dir/dir0) + [ $stripe_index -eq 0 ] || + error "dir0 expect index 0 got $stripe_index" + + mkdir $DIR/$tdir/striped_dir/dir1 || + error "create dir1 fails" + stripe_index=$($LFS getdirstripe -i $DIR/$tdir/striped_dir/dir1) + [ $stripe_index -eq 1 ] || + error "dir1 expect index 1 got $stripe_index" + + #check default stripe count/stripe index test_300_check_default_striped_dir normal_dir $MDSCOUNT 1 test_300_check_default_striped_dir normal_dir 1 0 test_300_check_default_striped_dir normal_dir 2 1 @@ -14918,11 +15945,23 @@ test_300i() { createmany -o $DIR/$tdir/striped_dir/f- 10 || error "create files under striped dir failed" + $LFS setdirstripe -i0 -c$MDSCOUNT -H all_char $DIR/$tdir/hashdir || + error "set striped hashdir error" + + $LFS setdirstripe -i0 -c$MDSCOUNT -H all_char $DIR/$tdir/hashdir/d0 || + error "create dir0 under hash dir failed" + $LFS setdirstripe -i0 -c$MDSCOUNT -H fnv_1a_64 $DIR/$tdir/hashdir/d1 || + error "create dir1 under hash dir failed" + # unfortunately, we need to umount to clear dir layout cache for now # once we fully implement dir layout, we can drop this umount_client $MOUNT || error "umount failed" mount_client $MOUNT || error "mount failed" + $LFS find -H fnv_1a_64 $DIR/$tdir/hashdir + local dircnt=$($LFS find -H fnv_1a_64 $DIR/$tdir/hashdir | wc -l) + [ $dircnt -eq 1 ] || error "lfs find striped dir got:$dircnt,except:1" + #set the stripe to be unknown hash type #define OBD_FAIL_UNKNOWN_LMV_STRIPE 0x1901 $LCTL set_param fail_loc=0x1901 @@ -15065,6 +16104,8 @@ test_300n() { skip "Need MDS version at least 2.7.55" && return [ $PARALLEL == "yes" ] && skip "skip parallel run" && return [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return + remote_mds_nodsh && skip "remote MDS with nodsh" && return + local stripe_index local list=$(comma_list $(mdts_nodes)) @@ -15257,6 +16298,7 @@ test_311() { [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.8.54) ] && skip "lustre < 2.8.54 does not contain LU-4825 fix" && return [ $PARALLEL == "yes" ] && skip "skip parallel run" && return + remote_mds_nodsh && skip "remote MDS with nodsh" && return local old_iused=$($LFS df -i | grep OST0000 | awk '{ print $3 }') @@ -15292,12 +16334,14 @@ test_311() { local new_iused for i in $(seq 120); do new_iused=$($LFS df -i | grep OST0000 | awk '{ print $3 }') - [ $((old_iused - new_iused)) -gt 800 ] && break + # system may be too busy to destroy all objs in time, use + # a somewhat small value to not fail autotest + [ $((old_iused - new_iused)) -gt 400 ] && break sleep 1 done echo "waited $i sec, old Iused $old_iused, new Iused $new_iused" - [ $((old_iused - new_iused)) -gt 800 ] || + [ $((old_iused - new_iused)) -gt 400 ] || error "objs not destroyed after unlink" } run_test 311 "disable OSP precreate, and unlink should destroy objs" @@ -15308,7 +16352,7 @@ zfs_oid_to_objid() local objid=$2 local vdevdir=$(dirname $(facet_vdevice $ost)) - local cmd="$ZDB -e -p $vdevdir -dddd $(facet_device $ost)" + local cmd="$ZDB -e -p $vdevdir -ddddd $(facet_device $ost)" local zfs_zapid=$(do_facet $ost $cmd | grep -w "/O/0/d$((objid%32))" -C 5 | awk '/Object/{getline; print $1}') @@ -15337,12 +16381,15 @@ zfs_object_blksz() { } test_312() { # LU-4856 + remote_ost_nodsh && skip "remote OST with nodsh" && return + [ $(facet_fstype ost1) = "zfs" ] || { skip "the test only applies to zfs" && return; } local max_blksz=$(do_facet ost1 \ $ZFS get -p recordsize $(facet_device ost1) | awk '!/VALUE/{print $3}') + local min_blksz=$(getconf PAGE_SIZE) # to make life a little bit easier $LFS mkdir -c 1 -i 0 $DIR/$tdir @@ -15357,7 +16404,7 @@ test_312() { # LU-4856 # block size change by sequential over write local blksz - for ((bs=4096; bs <= max_blksz; bs <<= 2)); do + for ((bs=$min_blksz; bs <= max_blksz; bs <<= 2)); do dd if=/dev/zero of=$tf bs=$bs count=1 oflag=sync conv=notrunc blksz=$(zfs_object_blksz ost1 $zfs_objid) @@ -15366,18 +16413,18 @@ test_312() { # LU-4856 rm -f $tf # block size change by sequential append write - dd if=/dev/zero of=$tf bs=4K count=1 oflag=sync conv=notrunc + dd if=/dev/zero of=$tf bs=$min_blksz count=1 oflag=sync conv=notrunc oid=$($LFS getstripe $tf | awk '/obdidx/{getline; print $2}') zfs_objid=$(zfs_oid_to_objid ost1 $oid) - for ((count = 1; count < $((max_blksz / 4096)); count *= 2)); do - dd if=/dev/zero of=$tf bs=4K count=$count seek=$count \ + for ((count = 1; count < $((max_blksz / min_blksz)); count *= 2)); do + dd if=/dev/zero of=$tf bs=$min_blksz count=$count seek=$count \ oflag=sync conv=notrunc blksz=$(zfs_object_blksz ost1 $zfs_objid) - blksz=$((blksz / 8192)) # in 2*4K unit - [ $blksz -eq $count ] || - error "blksz error(in 8k): $blksz, expected: $count" + [ $blksz -eq $((2 * count * min_blksz)) ] || + error "blksz error, actual $blksz, " \ + "expected: 2 * $count * $min_blksz" done rm -f $tf @@ -15386,9 +16433,10 @@ test_312() { # LU-4856 oid=$($LFS getstripe $tf | awk '/obdidx/{getline; print $2}') zfs_objid=$(zfs_oid_to_objid ost1 $oid) - dd if=/dev/zero of=$tf bs=8K count=1 oflag=sync conv=notrunc + dd if=/dev/zero of=$tf bs=1K count=1 oflag=sync conv=notrunc blksz=$(zfs_object_blksz ost1 $zfs_objid) - [ $blksz -eq 8192 ] || error "blksz error: $blksz, expected: 8k" + [ $blksz -eq $min_blksz ] || + error "blksz error: $blksz, expected: $min_blksz" dd if=/dev/zero of=$tf bs=64K count=1 oflag=sync conv=notrunc seek=128 blksz=$(zfs_object_blksz ost1 $zfs_objid) @@ -15400,7 +16448,32 @@ test_312() { # LU-4856 } run_test 312 "make sure ZFS adjusts its block size by write pattern" -test_399() { # LU-7655 for OST fake write +test_313() { + remote_ost_nodsh && skip "remote OST with nodsh" && return + + local file=$DIR/$tfile + rm -f $file + $SETSTRIPE -c 1 -i 0 $file || error "setstripe failed" + + # define OBD_FAIL_TGT_RCVD_EIO 0x720 + do_facet ost1 "$LCTL set_param fail_loc=0x720" + dd if=/dev/zero of=$file bs=4096 oflag=direct count=1 && + error "write should failed" + do_facet ost1 "$LCTL set_param fail_loc=0" + rm -f $file +} +run_test 313 "io should fail after last_rcvd update fail" + +test_fake_rw() { + local read_write=$1 + if [ "$read_write" = "write" ]; then + local dd_cmd="dd if=/dev/zero of=$DIR/$tfile" + elif [ "$read_write" = "read" ]; then + local dd_cmd="dd of=/dev/null if=$DIR/$tfile" + else + error "argument error" + fi + # turn off debug for performance testing local saved_debug=$($LCTL get_param -n debug) $LCTL set_param debug=0 @@ -15412,35 +16485,59 @@ test_399() { # LU-7655 for OST fake write local blocks=$((ost1_avail_size/2/1024)) # half avail space by megabytes [ $blocks -gt 1000 ] && blocks=1000 # 1G in maximum + if [ "$read_write" = "read" ]; then + truncate -s $(expr 1048576 \* $blocks) $DIR/$tfile + fi + local start_time=$(date +%s.%N) - dd if=/dev/zero of=$DIR/$tfile bs=1M count=$blocks oflag=sync || - error "real dd writing error" + $dd_cmd bs=1M count=$blocks oflag=sync || + error "real dd $read_write error" local duration=$(bc <<< "$(date +%s.%N) - $start_time") - rm -f $DIR/$tfile - # define OBD_FAIL_OST_FAKE_WRITE 0x238 + if [ "$read_write" = "write" ]; then + rm -f $DIR/$tfile + fi + + # define OBD_FAIL_OST_FAKE_RW 0x238 do_facet ost1 $LCTL set_param fail_loc=0x238 local start_time=$(date +%s.%N) - dd if=/dev/zero of=$DIR/$tfile bs=1M count=$blocks oflag=sync || - error "fake dd writing error" + $dd_cmd bs=1M count=$blocks oflag=sync || + error "fake dd $read_write error" local duration_fake=$(bc <<< "$(date +%s.%N) - $start_time") - # verify file size - cancel_lru_locks osc - $CHECKSTAT -t file -s $((blocks * 1024 * 1024)) $DIR/$tfile || - error "$tfile size not $blocks MB" - + if [ "$read_write" = "write" ]; then + # verify file size + cancel_lru_locks osc + $CHECKSTAT -t file -s $((blocks * 1024 * 1024)) $DIR/$tfile || + error "$tfile size not $blocks MB" + fi do_facet ost1 $LCTL set_param fail_loc=0 - echo "fake write $duration_fake vs. normal write $duration in seconds" + echo "fake $read_write $duration_fake vs. normal $read_write" \ + "$duration in seconds" [ $(bc <<< "$duration_fake < $duration") -eq 1 ] || error_not_in_vm "fake write is slower" $LCTL set_param -n debug="$saved_debug" rm -f $DIR/$tfile } -run_test 399 "fake write should not be slower than normal write" +test_399a() { # LU-7655 for OST fake write + remote_ost_nodsh && skip "remote OST with nodsh" && return + + test_fake_rw write +} +run_test 399a "fake write should not be slower than normal write" + +test_399b() { # LU-8726 for OST fake read + remote_ost_nodsh && skip "remote OST with nodsh" && return + + if [ "$(facet_fstype ost1)" != "ldiskfs" ]; then + skip "ldiskfs only test" && return 0 + fi + test_fake_rw read +} +run_test 399b "fake read should not be slower than normal read" test_400a() { # LU-1606, was conf-sanity test_74 local extra_flags='' @@ -15455,7 +16552,6 @@ test_400a() { # LU-1606, was conf-sanity test_74 if ! [[ -d $prefix ]]; then # Assume we're running in tree and fixup the include path. - extra_flags+=" -I$LUSTRE/../libcfs/include" extra_flags+=" -I$LUSTRE/include" extra_flags+=" -L$LUSTRE/utils" fi @@ -15471,7 +16567,7 @@ run_test 400a "Lustre client api program can compile and link" test_400b() { # LU-1606, LU-5011 local header local out=$TMP/$tfile - local prefix=/usr/include/lustre + local prefix=/usr/include/linux/lustre # We use a hard coded prefix so that this test will not fail # when run in tree. There are headers in lustre/include/lustre/ @@ -15490,8 +16586,8 @@ test_400b() { # LU-1606, LU-5011 continue fi - if [[ "$(basename $header)" == liblustreapi.h ]]; then - continue # liblustreapi.h is deprecated. + if [[ "$(basename $header)" == lustre_ioctl.h ]]; then + continue # lustre_ioctl.h is internal header fi $CC -Wall -Werror -include $header -c -x c /dev/null -o $out || @@ -15508,6 +16604,9 @@ test_401a() { #LU-7437 #count the number of parameters by "list_param -R" local params=$($LCTL list_param -R '*' 2>/dev/null | wc -l) #count the number of parameters by listing proc files + local proc_dirs=$(eval \ls -d $proc_regexp 2>/dev/null) + echo "proc_dirs='$proc_dirs'" + [ -n "$proc_dirs" ] || error "no proc_dirs on $HOSTNAME" local procs=$(find -L $proc_dirs -mindepth 1 -printf '%P\n' 2>/dev/null| sort -u | wc -l) @@ -15594,6 +16693,14 @@ test_401d() { run_test 401d "Verify 'lctl set_param' accepts values containing '='" test_402() { + local server_version=$(lustre_version_code $SINGLEMDS) + [[ $server_version -ge $(version_code 2.7.66) ]] || + [[ $server_version -ge $(version_code 2.7.18.4) && + $server_version -lt $(version_code 2.7.50) ]] || + [[ $server_version -ge $(version_code 2.7.2) && + $server_version -lt $(version_code 2.7.11) ]] || + { skip "Need MDS version 2.7.2+ or 2.7.18.4+ or 2.7.66+"; + return; } remote_mds_nodsh && skip "remote MDS with nodsh" && return $LFS setdirstripe -i 0 $DIR/$tdir || error "setdirstripe -i 0 failed" #define OBD_FAIL_MDS_FLD_LOOKUP 0x15c @@ -15633,6 +16740,10 @@ test_403() { run_test 403 "i_nlink should not drop to zero due to aliasing" test_404() { # LU-6601 + local server_version=$(lustre_version_code $SINGLEMDS) + [[ $server_version -ge $(version_code 2.8.53) ]] || + { skip "Need server version newer than 2.8.52"; return 0; } + remote_mds_nodsh && skip "remote MDS with nodsh" && return local mosps=$(do_facet $SINGLEMDS $LCTL dl | awk '/osp .*-osc-MDT/ { print $4}') @@ -15660,12 +16771,13 @@ test_404() { # LU-6601 run_test 404 "validate manual {de}activated works properly for OSPs" test_405() { - [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.6.92) ] && + [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.6.92) -o \ + [ $(lustre_version_code client) -lt $(version_code 2.6.99) ] && skip "Layout swap lock is not supported" && return check_swap_layouts_support && return 0 - test_mkdir -p $DIR/$tdir + test_mkdir $DIR/$tdir swap_lock_test -d $DIR/$tdir || error "One layout swap locked test failed" } @@ -15679,12 +16791,15 @@ test_406() { [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.8.50) ] && skip "Need MDS version at least 2.8.50" && return - local def_stripenr=$($GETSTRIPE -c $MOUNT) + local def_stripe_count=$($GETSTRIPE -c $MOUNT) local def_stripe_size=$($GETSTRIPE -S $MOUNT) local def_stripe_offset=$($GETSTRIPE -i $MOUNT) local def_pool=$($GETSTRIPE -p $MOUNT) - local test_pool=$TESTNAME + + if ! combined_mgs_mds ; then + mount_mgs_client + fi pool_add $test_pool || error "pool_add failed" pool_add_targets $test_pool 0 $(($OSTCOUNT - 1)) 1 || error "pool_add_targets failed" @@ -15734,26 +16849,30 @@ test_406() { # restore FS default striping if [ -z $def_pool ]; then - $SETSTRIPE -c $def_stripenr -S $def_stripe_size \ + $SETSTRIPE -c $def_stripe_count -S $def_stripe_size \ -i $def_stripe_offset $MOUNT || error "restore default striping failed" else - $SETSTRIPE -c $def_stripenr -S $def_stripe_size -p $def_pool \ - -i $def_stripe_offset $MOUNT || + $SETSTRIPE -c $def_stripe_count -S $def_stripe_size \ + -i $def_stripe_offset -p $def_pool $MOUNT || error "restore default striping with $def_pool failed" fi local f=$DIR/$tdir/$tfile pool_remove_all_targets $test_pool $f pool_remove $test_pool $f + + if ! combined_mgs_mds ; then + umount_mgs_client + fi } run_test 406 "DNE support fs default striping" test_407() { [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return - [[ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.8.55) ]] && skip "Need MDS version at least 2.8.55" && return + remote_mds_nodsh && skip "remote MDS with nodsh" && return $LFS mkdir -i 0 -c 1 $DIR/$tdir.0 || error "$LFS mkdir -i 0 -c 1 $tdir.0 failed" @@ -15824,6 +16943,330 @@ test_409() } run_test 409 "Large amount of cross-MDTs hard links on the same file" +test_410() +{ + [[ $(lustre_version_code client) -lt $(version_code 2.9.59) ]] && + skip "Need client version at least 2.9.59" && return + + # Create a file, and stat it from the kernel + local testfile=$DIR/$tfile + touch $testfile + + local run_id=$RANDOM + local my_ino=$(stat --format "%i" $testfile) + + # Try to insert the module. This will always fail as the + # module is designed to not be inserted. + insmod $LUSTRE/tests/kernel/kinode.ko run_id=$run_id fname=$testfile \ + &> /dev/null + + # Anything but success is a test failure + dmesg | grep -q \ + "lustre_kinode_$run_id: inode numbers are identical: $my_ino" || + error "no inode match" +} +run_test 410 "Test inode number returned from kernel thread" + +cleanup_test411_cgroup() { + trap 0 + rmdir "$1" +} + +test_411() { + local cg_basedir=/sys/fs/cgroup/memory + # LU-9966 + test -f "$cg_basedir/memory.kmem.limit_in_bytes" || + { skip "no setup for cgroup"; return; } + + dd if=/dev/zero of=$DIR/$tfile bs=1M count=100 conv=fsync || + error "test file creation failed" + cancel_lru_locks osc + + # Create a very small memory cgroup to force a slab allocation error + local cgdir=$cg_basedir/osc_slab_alloc + mkdir $cgdir || error "cgroup mkdir '$cgdir' failed" + trap "cleanup_test411_cgroup $cgdir" EXIT + echo 2M > $cgdir/memory.kmem.limit_in_bytes + echo 1M > $cgdir/memory.limit_in_bytes + + # Should not LBUG, just be killed by oom-killer + sh -c "echo \$$ > $cgdir/tasks && dd if=$DIR/$tfile of=/dev/null" && + error "fail to trigger a memory allocation error" + + cleanup_test411_cgroup $cgdir + + return 0 +} +run_test 411 "Slab allocation error with cgroup does not LBUG" + +prep_801() { + [[ $(lustre_version_code mds1) -lt $(version_code 2.9.55) ]] || + [[ $(lustre_version_code ost1) -lt $(version_code 2.9.55) ]] && + skip "Need server version at least 2.9.55" & exit 0 + start_full_debug_logging +} + +post_801() { + stop_full_debug_logging +} + +barrier_stat() { + if [ $(lustre_version_code mgs) -le $(version_code 2.10.0) ]; then + local st=$(do_facet mgs $LCTL barrier_stat $FSNAME | + awk '/The barrier for/ { print $7 }') + echo $st + else + local st=$(do_facet mgs $LCTL barrier_stat -s $FSNAME) + echo \'$st\' + fi +} + +barrier_expired() { + local expired + + if [ $(lustre_version_code mgs) -le $(version_code 2.10.0) ]; then + expired=$(do_facet mgs $LCTL barrier_stat $FSNAME | + awk '/will be expired/ { print $7 }') + else + expired=$(do_facet mgs $LCTL barrier_stat -t $FSNAME) + fi + + echo $expired +} + +test_801a() { + prep_801 + + echo "Start barrier_freeze at: $(date)" + #define OBD_FAIL_BARRIER_DELAY 0x2202 + do_facet mgs $LCTL set_param fail_val=5 fail_loc=0x2202 + do_facet mgs $LCTL barrier_freeze $FSNAME 10 & + + sleep 2 + local b_status=$(barrier_stat) + echo "Got barrier status at: $(date)" + [ "$b_status" = "'freezing_p1'" ] || + error "(1) unexpected barrier status $b_status" + + do_facet mgs $LCTL set_param fail_val=0 fail_loc=0 + wait + b_status=$(barrier_stat) + [ "$b_status" = "'frozen'" ] || + error "(2) unexpected barrier status $b_status" + + local expired=$(barrier_expired) + echo "sleep $((expired + 3)) seconds, then the barrier will be expired" + sleep $((expired + 3)) + + b_status=$(barrier_stat) + [ "$b_status" = "'expired'" ] || + error "(3) unexpected barrier status $b_status" + + do_facet mgs $LCTL barrier_freeze $FSNAME 10 || + error "(4) fail to freeze barrier" + + b_status=$(barrier_stat) + [ "$b_status" = "'frozen'" ] || + error "(5) unexpected barrier status $b_status" + + echo "Start barrier_thaw at: $(date)" + #define OBD_FAIL_BARRIER_DELAY 0x2202 + do_facet mgs $LCTL set_param fail_val=5 fail_loc=0x2202 + do_facet mgs $LCTL barrier_thaw $FSNAME & + + sleep 2 + b_status=$(barrier_stat) + echo "Got barrier status at: $(date)" + [ "$b_status" = "'thawing'" ] || + error "(6) unexpected barrier status $b_status" + + do_facet mgs $LCTL set_param fail_val=0 fail_loc=0 + wait + b_status=$(barrier_stat) + [ "$b_status" = "'thawed'" ] || + error "(7) unexpected barrier status $b_status" + + #define OBD_FAIL_BARRIER_FAILURE 0x2203 + do_facet $SINGLEMDS $LCTL set_param fail_loc=0x2203 + do_facet mgs $LCTL barrier_freeze $FSNAME + + b_status=$(barrier_stat) + [ "$b_status" = "'failed'" ] || + error "(8) unexpected barrier status $b_status" + + do_facet $SINGLEMDS $LCTL set_param fail_loc=0 + do_facet mgs $LCTL barrier_thaw $FSNAME + + post_801 +} +run_test 801a "write barrier user interfaces and stat machine" + +test_801b() { + prep_801 + + mkdir $DIR/$tdir || error "(1) fail to mkdir" + createmany -d $DIR/$tdir/d 6 || "(2) fail to mkdir" + touch $DIR/$tdir/d2/f10 || error "(3) fail to touch" + touch $DIR/$tdir/d3/f11 || error "(4) fail to touch" + touch $DIR/$tdir/d4/f12 || error "(5) fail to touch" + + cancel_lru_locks mdc + + # 180 seconds should be long enough + do_facet mgs $LCTL barrier_freeze $FSNAME 180 + + local b_status=$(barrier_stat) + [ "$b_status" = "'frozen'" ] || + error "(6) unexpected barrier status $b_status" + + mkdir $DIR/$tdir/d0/d10 & + mkdir_pid=$! + + touch $DIR/$tdir/d1/f13 & + touch_pid=$! + + ln $DIR/$tdir/d2/f10 $DIR/$tdir/d2/f14 & + ln_pid=$! + + mv $DIR/$tdir/d3/f11 $DIR/$tdir/d3/f15 & + mv_pid=$! + + rm -f $DIR/$tdir/d4/f12 & + rm_pid=$! + + stat $DIR/$tdir/d5 || error "(7) stat should succeed" + + # To guarantee taht the 'stat' is not blocked + b_status=$(barrier_stat) + [ "$b_status" = "'frozen'" ] || + error "(8) unexpected barrier status $b_status" + + # let above commands to run at background + sleep 5 + + ps -p $mkdir_pid || error "(9) mkdir should be blocked" + ps -p $touch_pid || error "(10) touch should be blocked" + ps -p $ln_pid || error "(11) link should be blocked" + ps -p $mv_pid || error "(12) rename should be blocked" + ps -p $rm_pid || error "(13) unlink should be blocked" + + b_status=$(barrier_stat) + [ "$b_status" = "'frozen'" ] || + error "(14) unexpected barrier status $b_status" + + do_facet mgs $LCTL barrier_thaw $FSNAME + b_status=$(barrier_stat) + [ "$b_status" = "'thawed'" ] || + error "(15) unexpected barrier status $b_status" + + wait $mkdir_pid || error "(16) mkdir should succeed" + wait $touch_pid || error "(17) touch should succeed" + wait $ln_pid || error "(18) link should succeed" + wait $mv_pid || error "(19) rename should succeed" + wait $rm_pid || error "(20) unlink should succeed" + + post_801 +} +run_test 801b "modification will be blocked by write barrier" + +test_801c() { + [[ $MDSCOUNT -lt 2 ]] && skip "needs >= 2 MDTs" && return + + prep_801 + + stop mds2 || error "(1) Fail to stop mds2" + + do_facet mgs $LCTL barrier_freeze $FSNAME 30 + + local b_status=$(barrier_stat) + [ "$b_status" = "'expired'" -o "$b_status" = "'failed'" ] || { + do_facet mgs $LCTL barrier_thaw $FSNAME + error "(2) unexpected barrier status $b_status" + } + + do_facet mgs $LCTL barrier_rescan $FSNAME || + error "(3) Fail to rescan barrier bitmap" + + do_facet mgs $LCTL barrier_freeze $FSNAME 10 + + b_status=$(barrier_stat) + [ "$b_status" = "'frozen'" ] || + error "(4) unexpected barrier status $b_status" + + do_facet mgs $LCTL barrier_thaw $FSNAME + b_status=$(barrier_stat) + [ "$b_status" = "'thawed'" ] || + error "(5) unexpected barrier status $b_status" + + local devname=$(mdsdevname 2) + + start mds2 $devname $MDS_MOUNT_OPTS || error "(6) Fail to start mds2" + + do_facet mgs $LCTL barrier_rescan $FSNAME || + error "(7) Fail to rescan barrier bitmap" + + post_801 +} +run_test 801c "rescan barrier bitmap" + +saved_MGS_MOUNT_OPTS=$MGS_MOUNT_OPTS +saved_MDS_MOUNT_OPTS=$MDS_MOUNT_OPTS +saved_OST_MOUNT_OPTS=$OST_MOUNT_OPTS + +cleanup_802() { + trap 0 + + stopall + MGS_MOUNT_OPTS=$saved_MGS_MOUNT_OPTS + MDS_MOUNT_OPTS=$saved_MDS_MOUNT_OPTS + OST_MOUNT_OPTS=$saved_OST_MOUNT_OPTS + setupall +} + +test_802() { + + [[ $(lustre_version_code mds1) -lt $(version_code 2.9.55) ]] || + [[ $(lustre_version_code ost1) -lt $(version_code 2.9.55) ]] && + skip "Need server version at least 2.9.55" & exit 0 + + mkdir $DIR/$tdir || error "(1) fail to mkdir" + + cp $LUSTRE/tests/test-framework.sh $DIR/$tdir/ || + error "(2) Fail to copy" + + trap cleanup_802 EXIT + + # sync by force before remount as readonly + sync; sync_all_data; sleep 3; sync_all_data + + stopall + + MGS_MOUNT_OPTS=$(csa_add "$MGS_MOUNT_OPTS" -o rdonly_dev) + MDS_MOUNT_OPTS=$(csa_add "$MDS_MOUNT_OPTS" -o rdonly_dev) + OST_MOUNT_OPTS=$(csa_add "$OST_MOUNT_OPTS" -o rdonly_dev) + + echo "Mount the server as read only" + setupall server_only || error "(3) Fail to start servers" + + echo "Mount client without ro should fail" + mount_client $MOUNT && + error "(4) Mount client without 'ro' should fail" + + echo "Mount client with ro should succeed" + mount_client $MOUNT ro || + error "(5) Mount client with 'ro' should succeed" + + echo "Modify should be refused" + touch $DIR/$tdir/guard && error "(6) Touch should fail under ro mode" + + echo "Read should be allowed" + diff $LUSTRE/tests/test-framework.sh $DIR/$tdir/test-framework.sh || + error "(7) Read should succeed under ro mode" + + cleanup_802 +} +run_test 802 "simulate readonly device" + # # tests that do cleanup/setup should be run at the end #