X-Git-Url: https://git.whamcloud.com/?p=fs%2Flustre-release.git;a=blobdiff_plain;f=lustre%2Ftests%2Fsanity.sh;h=88e73ff79ec6ec0203b288dff405d3612a08558b;hp=ac9f21c705c16fa9dcc23e5e8c6218697c1d9c84;hb=d38920afefd403ce5e367a2203df5ee196792c9f;hpb=f4d727d04dd22831d2c037271894036826230cbd diff --git a/lustre/tests/sanity.sh b/lustre/tests/sanity.sh index ac9f21c..88e73ff 100644 --- a/lustre/tests/sanity.sh +++ b/lustre/tests/sanity.sh @@ -10,7 +10,7 @@ set -e ONLY=${ONLY:-"$*"} # bug number for skipped test: 13297 2108 9789 3637 9789 3561 12622 12653 12653 5188 16260 19742 -ALWAYS_EXCEPT=" 27u 42a 42b 42c 42d 45 51d 65a 65e 68b 119d 130 $SANITY_EXCEPT" +ALWAYS_EXCEPT=" 27u 42a 42b 42c 42d 45 51d 65a 65e 68b $SANITY_EXCEPT" # bug number for skipped test: 2108 9789 3637 9789 3561 5188/5749 1443 #ALWAYS_EXCEPT=${ALWAYS_EXCEPT:-"27m 42a 42b 42c 42d 45 68 76"} # UPDATE THE COMMENT ABOVE WITH BUG NUMBERS WHEN CHANGING ALWAYS_EXCEPT! @@ -72,8 +72,9 @@ LUSTRE=${LUSTRE:-$(cd $(dirname $0)/..; echo $PWD)} . $LUSTRE/tests/test-framework.sh init_test_env $@ . ${CONFIG:=$LUSTRE/tests/cfg/${NAME}.sh} +init_logging -[ "$SLOW" = "no" ] && EXCEPT_SLOW="24o 24v 27m 36f 36g 51b 51c 60c 63 64b 68 71 73 77f 78 101 103 115 120g 124b" +[ "$SLOW" = "no" ] && EXCEPT_SLOW="24o 24v 27m 36f 36g 36h 51b 51c 60c 63 64b 68 71 73 77f 78 101 103 115 120g 124b" SANITYLOG=${TESTSUITELOG:-$TMP/$(basename $0 .sh).log} FAIL_ON_ERROR=false @@ -984,7 +985,7 @@ test_27m() { run_test 27m "create file while OST0 was full ==================" sleep_maxage() { - local DELAY=$(do_facet mds lctl get_param -n lov.*.qos_maxage | head -n 1 | awk '{print $1 + 5}') + local DELAY=$(do_facet mds lctl get_param -n lov.*.qos_maxage | head -n 1 | awk '{print $1 * 2}') sleep $DELAY } @@ -1020,7 +1021,7 @@ exhaust_precreations() { do_facet mds${MDSIDX} lctl get_param osc.*OST*-osc-${MDT_INDEX}.prealloc* - mkdir -p $DIR/d27/${OST} + mkdir -p $DIR/$tdir/${OST} $SETSTRIPE $DIR/$tdir/${OST} -i $OSTIDX -c 1 #define OBD_FAIL_OST_ENOSPC 0x215 do_facet ost$((OSTIDX + 1)) lctl set_param fail_val=$FAILIDX @@ -1200,15 +1201,17 @@ run_test 27v "skip object creation on slow OST =================" test_27w() { # bug 10997 mkdir -p $DIR/$tdir || error "mkdir failed" $LSTRIPE $DIR/$tdir/f0 -s 65536 || error "lstripe failed" - size=`$GETSTRIPE $DIR/$tdir/f0 -qs | head -n 1` + size=`$GETSTRIPE $DIR/$tdir/f0 -s` [ $size -ne 65536 ] && error "stripe size $size != 65536" || true + gsdir=$($LFS getstripe -d $DIR/$tdir) + [ $(echo $gsdir | grep -c stripe_count) -ne 1 ] && error "$LFS getstripe -d $DIR/$tdir failed" [ "$OSTCOUNT" -lt "2" ] && skip_env "skipping multiple stripe count/offset test" && return for i in `seq 1 $OSTCOUNT`; do offset=$(($i-1)) $LSTRIPE $DIR/$tdir/f$i -c $i -i $offset || error "lstripe -c $i -i $offset failed" - count=`$GETSTRIPE -qc $DIR/$tdir/f$i | head -n 1` - index=`$GETSTRIPE -qo $DIR/$tdir/f$i | head -n 1` + count=`$GETSTRIPE -c $DIR/$tdir/f$i` + index=`$GETSTRIPE -o $DIR/$tdir/f$i` [ $count -ne $i ] && error "stripe count $count != $i" || true [ $index -ne $offset ] && error "stripe offset $index != $offset" || true done @@ -1238,6 +1241,12 @@ test_27y() { [ "$OSTCOUNT" -lt "2" ] && skip_env "$OSTCOUNT < 2 OSTs -- skipping" && return remote_mds_nodsh && skip "remote MDS with nodsh" && return + local last_id=$(do_facet $SINGLEMDS lctl get_param -n osc.*0000-osc-MDT0000.prealloc_last_id) + local next_id=$(do_facet $SINGLEMDS lctl get_param -n osc.*0000-osc-MDT0000.prealloc_next_id) + local fcount=$((last_id - next_id)) + [ $fcount -eq 0 ] && skip "not enough space on OST0" && return + [ $fcount -gt $OSTCOUNT ] && fcount=$OSTCOUNT + MDS_OSCS=`do_facet mds lctl dl | awk '/[oO][sS][cC].*md[ts]/ { print $4 }'` OFFSET=$(($OSTCOUNT-1)) OST=-1 @@ -1256,7 +1265,7 @@ test_27y() { do_facet ost$OSTIDX lctl set_param -n obdfilter.$OST.degraded 1 sleep_maxage - createmany -o $DIR/$tdir/$tfile $OSTCOUNT + createmany -o $DIR/$tdir/$tfile $fcount do_facet ost$OSTIDX lctl set_param -n obdfilter.$OST.degraded 0 for i in `seq 0 $OFFSET`; do @@ -1778,6 +1787,23 @@ test_34f() { # bug 6242, 6243 } run_test 34f "read from a file with no objects until EOF =======" +test_34g() { + dd if=/dev/zero of=$DIR/$tfile bs=1 count=100 seek=$TEST_34_SIZE || error + $TRUNCATE $DIR/$tfile $((TEST_34_SIZE / 2))|| error + $CHECKSTAT -s $((TEST_34_SIZE / 2)) $DIR/$tfile || error "truncate failed" + cancel_lru_locks osc + $CHECKSTAT -s $((TEST_34_SIZE / 2)) $DIR/$tfile || \ + error "wrong size after lock cancel" + + $TRUNCATE $DIR/$tfile $TEST_34_SIZE || error + $CHECKSTAT -s $TEST_34_SIZE $DIR/$tfile || \ + error "expanding truncate failed" + cancel_lru_locks osc + $CHECKSTAT -s $TEST_34_SIZE $DIR/$tfile || \ + error "wrong expanded size after lock cancel" +} +run_test 34g "truncate long file ===============================" + test_35a() { cp /bin/sh $DIR/f35a chmod 444 $DIR/f35a @@ -1823,15 +1849,15 @@ test_36e() { } run_test 36e "utime on non-owned file (should return error) ====" -test_36f() { +subr_36fh() { + local fl="$1" local LANG_SAVE=$LANG local LC_LANG_SAVE=$LC_LANG export LANG=C LC_LANG=C # for date language DATESTR="Dec 20 2000" mkdir -p $DIR/$tdir - #define OBD_FAIL_OST_BRW_PAUSE_BULK 0x214 - lctl set_param fail_loc=0x80000214 + lctl set_param fail_loc=$fl date; date +%s cp /etc/hosts $DIR/$tdir/$tfile sync & # write RPC generated with "current" inode timestamp, but delayed @@ -1849,6 +1875,11 @@ test_36f() { export LANG=$LANG_SAVE LC_LANG=$LC_LANG_SAVE } + +test_36f() { + #define OBD_FAIL_OST_BRW_PAUSE_BULK 0x214 + subr_36fh "0x80000214" +} run_test 36f "utime on file racing with OST BRW write ==========" test_36g() { @@ -1866,6 +1897,12 @@ test_36g() { } run_test 36g "filter mod data cache expiry =====================" +test_36h() { + #define OBD_FAIL_OST_BRW_PAUSE_BULK2 0x227 + subr_36fh "0x80000227" +} +run_test 36h "utime on file racing with OST BRW write ==========" + test_37() { mkdir -p $DIR/$tdir echo f > $DIR/$tdir/fbugfile @@ -2066,6 +2103,73 @@ test_42d() { } run_test 42d "test complete truncate of file with cached dirty data" +page_size() { + getconf PAGE_SIZE +} + +test_42e() { # bug22074 + local TDIR=$DIR/${tdir}e + local pagesz=$(page_size) + local pages=16 + local files=$((OSTCOUNT * 500)) # hopefully 500 files on each OST + local proc_osc0="osc.${FSNAME}-OST0000-osc-[^MDT]*" + local max_dirty_mb + local warmup_files + + mkdir -p $TDIR + $LFS setstripe -c 1 $TDIR + createmany -o $TDIR/f $files + + max_dirty_mb=$($LCTL get_param -n $proc_osc0/max_dirty_mb) + + # we assume that with $OSTCOUNT files, at least one of them will + # be allocated on OST0. + warmup_files=$((OSTCOUNT * max_dirty_mb)) + createmany -o $TDIR/w $warmup_files + + # write a large amount of data into one file and sync, to get good + # avail_grant number from OST. + for ((i=0; i<$warmup_files; i++)); do + idx=$($LFS getstripe -i $TDIR/w$i) + [ $idx -ne 0 ] && continue + dd if=/dev/zero of=$TDIR/w$i bs="$max_dirty_mb"M count=1 + break + done + [ $i -gt $warmup_files ] && error "OST0 is still cold" + sync + $LCTL get_param $proc_osc0/cur_dirty_bytes + $LCTL get_param $proc_osc0/cur_grant_bytes + + # create as much dirty pages as we can while not to trigger the actual + # RPCs directly. but depends on the env, VFS may trigger flush during this + # period, hopefully we are good. + for ((i=0; i<$warmup_files; i++)); do + idx=$($LFS getstripe -i $TDIR/w$i) + [ $idx -ne 0 ] && continue + dd if=/dev/zero of=$TDIR/w$i bs=1M count=1 2>/dev/null + done + $LCTL get_param $proc_osc0/cur_dirty_bytes + $LCTL get_param $proc_osc0/cur_grant_bytes + + # perform the real test + $LCTL set_param $proc_osc0/rpc_stats 0 + for ((;i<$files; i++)); do + [ $($LFS getstripe -i $TDIR/f$i) -eq 0 ] || continue + dd if=/dev/zero of=$TDIR/f$i bs=$pagesz count=$pages 2>/dev/null + done + sync + $LCTL get_param $proc_osc0/rpc_stats + + $LCTL get_param $proc_osc0/rpc_stats | + while read PPR RRPC RPCT RCUM BAR WRPC WPCT WCUM; do + [ "$PPR" != "16:" ] && continue + [ $WPCT -lt 85 ] && error "$pages-page write RPCs only $WPCT% < 85%" + break # we only want the "pages per rpc" stat + done + rm -rf $TDIR +} +run_test 42e "verify sub-RPC writes are not done synchronously" + test_43() { mkdir -p $DIR/$tdir cp -p /bin/ls $DIR/$tdir/$tfile @@ -2191,10 +2295,6 @@ test_45() { } run_test 45 "osc io page accounting ============================" -page_size() { - getconf PAGE_SIZE -} - # in a 2 stripe file (lov.sh), page 1023 maps to page 511 in its object. this # test tickles a bug where re-dirtying a page was failing to be mapped to the # objects offset and an assert hit when an rpc was built with 1023's mapped @@ -2464,7 +2564,7 @@ test_52a() { link $DIR/d52a/foo $DIR/d52a/foo_link 2>/dev/null && error "link worked" echo foo >> $DIR/d52a/foo || error "append foo failed" mrename $DIR/d52a/foo $DIR/d52a/foo_ren && error "rename worked" - lsattr $DIR/d52a/foo | egrep -q "^-+[ae]-+ $DIR/d52a/foo" || error "lsattr" + lsattr $DIR/d52a/foo | egrep -q "^-+a[-e]+ $DIR/d52a/foo" || error "lsattr" chattr -a $DIR/d52a/foo || error "chattr -a failed" cp -r $DIR/d52a /tmp/ rm -fr $DIR/d52a || error "cleanup rm failed" @@ -2484,7 +2584,7 @@ test_52b() { mrename $DIR/d52b/foo $DIR/d52b/foo_ren && error "rename worked" [ -f $DIR/d52b/foo ] || error [ -f $DIR/d52b/foo_ren ] && error - lsattr $DIR/d52b/foo | egrep -q "^-+[ie]-+ $DIR/d52b/foo" || error "lsattr" + lsattr $DIR/d52b/foo | egrep -q "^-+i[-e]+ $DIR/d52b/foo" || error "lsattr" chattr -i $DIR/d52b/foo || error "chattr failed" rm -fr $DIR/d52b || error @@ -2651,12 +2751,14 @@ test_56a() { # was test_56 [ "$OSTCOUNT" -lt 2 ] && \ skip_env "skipping other lfs getstripe --obd test" && return - FILENUM=`$GETSTRIPE --recursive $DIR/d56 | sed -n '/^[ ]*1[ ]/p' | wc -l` - OBDUUID=`$GETSTRIPE --recursive $DIR/d56 | sed -n '/^[ ]*1:/p' | awk '{print $2}'` - FOUND=`$GETSTRIPE -r --obd $OBDUUID $DIR/d56 | wc -l` + OSTIDX=1 + OBDUUID=$(lfs osts | grep ${OSTIDX}": " | awk '{print $2}') + FILENUM=`$GETSTRIPE -ir $DIR/d56 | grep -x $OSTIDX | wc -l` + FOUND=`$GETSTRIPE -r --obd $OBDUUID $DIR/d56 | grep obdidx | wc -l` [ $FOUND -eq $FILENUM ] || \ error "lfs getstripe --obd wrong: found $FOUND, expected $FILENUM" - [ `$GETSTRIPE -r -v --obd $OBDUUID $DIR/d56 | sed '/^[ ]*1[ ]/d' |\ + [ `$GETSTRIPE -r -v --obd $OBDUUID $DIR/d56 | \ + sed '/^[ ]*'${OSTIDX}'[ ]/d' |\ sed -n '/^[ ]*[0-9][0-9]*[ ]/p' | wc -l` -eq 0 ] || \ error "lfs getstripe --obd wrong: should not show file on other obd" echo "lfs getstripe --obd passed." @@ -2736,7 +2838,7 @@ run_test 56h "check lfs find ! -name =============================" test_56i() { tdir=${tdir}i mkdir -p $DIR/$tdir - UUID=`$GETSTRIPE $DIR/$tdir | awk '/0: / { print $2 }'` + UUID=`$LFS osts | awk '/0: / { print $2 }'` OUT="`$LFIND -ost $UUID $DIR/$tdir`" [ "$OUT" ] && error "$LFIND returned directory '$OUT'" || true } @@ -3638,7 +3740,9 @@ rm -f $F77_TMP unset F77_TMP test_78() { # bug 10901 - NSEQ=5 + remote_ost || { skip_env "local OST" && return; } + + NSEQ=5 F78SIZE=$(($(awk '/MemFree:/ { print $2 }' /proc/meminfo) / 1024)) echo "MemFree: $F78SIZE, Max file size: $MAXFREE" MEMTOTAL=$(($(awk '/MemTotal:/ { print $2 }' /proc/meminfo) / 1024)) @@ -4328,7 +4432,7 @@ test_103 () { } run_test 103 "acl test =========================================" -test_104() { +test_104a() { touch $DIR/$tfile lfs df || error "lfs df failed" lfs df -ih || error "lfs df -ih failed" @@ -4344,7 +4448,18 @@ test_104() { lfs df || error "lfs df with reactivated OSC failed" rm -f $DIR/$tfile } -run_test 104 "lfs df [-ih] [path] test =========================" +run_test 104a "lfs df [-ih] [path] test =========================" + +test_104b() { + [ $RUNAS_ID -eq $UID ] && skip_env "RUNAS_ID = UID = $UID -- skipping" && return + chmod 666 /dev/obd + denied_cnt=$((`$RUNAS $LFS check servers 2>&1 | grep "Permission denied" | wc -l`)) + if [ $denied_cnt -ne 0 ]; + then + error "lfs check servers test failed" + fi +} +run_test 104b "$RUNAS lfs check servers test ====================" test_105a() { # doesn't work on 2.4 kernels @@ -5926,9 +6041,12 @@ test_132() { #1028, SOM stat $DIR/$tfile >/dev/null gl2=$(get_ost_param "ldlm_glimpse_enqueue") echo "====> SOM is "$som1", "$((gl2 - gl1))" glimpse RPC occured" - cancel_lru_locks osc + rm $DIR/$tfile som_mode_switch $som1 $gl1 $gl2 + dd if=/dev/zero of=$DIR/$tfile count=1 2>/dev/null + cancel_lru_locks osc + som2=$(do_facet $mymds "$LCTL get_param mdt.*.som" | awk -F= ' {print $2}' | head -n 1) if [ $som1 == $som2 ]; then error "som is still "$som2 @@ -6506,15 +6624,16 @@ run_test 162 "path lookup sanity" test_163() { remote_mds_nodsh && skip "remote MDS with nodsh" && return - copytool --test || { skip "copytool not runnable: $?" && return; } - copytool & + copytool --test $FSNAME || { skip "copytool not runnable: $?" && return; } + copytool $FSNAME & sleep 1 local uuid=$($LCTL get_param -n mdc.${FSNAME}-MDT0000-mdc-*.uuid) # this proc file is temporary and linux-only - do_facet mds lctl set_param mdt.${FSNAME}-MDT0000.mdccomm=$uuid || error "lnl send failed" - kill $! + do_facet mds lctl set_param mdt.${FSNAME}-MDT0000.mdccomm=$uuid ||\ + error "kernel->userspace send failed" + kill -INT $! } -run_test 163 "LustreNetLink kernelcomms" +run_test 163 "kernel <-> userspace comms" test_169() { # do directio so as not to populate the page cache @@ -6857,100 +6976,134 @@ test_214() { # for bug 20133 } run_test 214 "hash-indexed directory test - bug 20133" -test_215() { # for bug 18102 - # /proc/sys/lnet/stats should look as 11 space-separated numbers - cat /proc/sys/lnet/stats >$TMP/lnet_stats.out - sysctl lnet.stats |sed 's/^lnet.stats\ =\ //g' >$TMP/lnet_stats.sys - STATS_LINES_OUT=$(cat $TMP/lnet_stats.out|wc -l) - [ "$STATS_LINES_OUT" = 1 ] || error "/proc/sys/lnet/stats has more than 1 line: $STATS" - STATS_LINES_SYS=$(cat $TMP/lnet_stats.sys|wc -l) - [ "$STATS_LINES_SYS" = 1 ] || error "lnet.stats has more than 1 line: $STATS" - STATS_REG='^[0-9]\+ [0-9]\+ [0-9]\+ [0-9]\+ [0-9]\+ [0-9]\+ [0-9]\+ [0-9]\+ [0-9]\+ [0-9]\+ [0-9]\+$' - grep "$STATS_REG" $TMP/lnet_stats.out || (cat $TMP/lnet_stats.out && - error "/proc/sys/lnet/stats misformatted") - grep "$STATS_REG" $TMP/lnet_stats.sys || (cat $TMP/lnet_stats.sys && - error "lnet.stats misformatted") - rm -f $TMP/lnet_stats.out $TMP/lnet_stats.sys - - # /proc/sys/lnet/routes should look exactly as expected - cat /proc/sys/lnet/routes >$TMP/lnet_routes.out - sysctl lnet.routes |sed 's/^lnet.routes\ =\ //g' >$TMP/lnet_routes.sys - echo "Routing disabled" >$TMP/lnet_routes.expected - echo "net hops state router" >>$TMP/lnet_routes.expected - diff $TMP/lnet_routes.expected $TMP/lnet_routes.out || - error "/proc/sys/lnet/routes does not look as expected" - diff $TMP/lnet_routes.expected $TMP/lnet_routes.sys || - error "lnet.routes does not look as expected" - rm -f $TMP/lnet_routes.expected $TMP/lnet_routes.out $TMP/lnet_routes.sys - - # /proc/sys/lnet/routers should look exactly as expected - cat /proc/sys/lnet/routers >$TMP/lnet_routers.out - sysctl lnet.routers |sed 's/^lnet.routers\ =\ //g' >$TMP/lnet_routers.sys - echo "ref rtr_ref alive_cnt state last_ping router" >$TMP/lnet_routers.expected - diff $TMP/lnet_routers.expected $TMP/lnet_routers.out || - error "/proc/sys/lnet/routers does not look as expected" - diff $TMP/lnet_routers.expected $TMP/lnet_routers.sys || - error "lnet.routers does not look as expected" - rm -f $TMP/lnet_routers.expected $TMP/lnet_routers.out $TMP/lnet_routers.sys - - # fisrt line of /proc/sys/lnet/peers should look exactly as expected - cat /proc/sys/lnet/peers >$TMP/lnet_peers.out - sysctl lnet.peers |sed 's/^lnet.peers\ =\ //g' >$TMP/lnet_peers.sys - head -1 $TMP/lnet_peers.out > $TMP/lnet_peers1.out - echo "nid refs state max rtr min tx min queue" >$TMP/lnet_peers1.expected - diff $TMP/lnet_peers1.expected $TMP/lnet_peers1.out || - error "first line of /proc/sys/lnet/peers does not look as expected" - rm -f $TMP/lnet_peers1.expected $TMP/lnet_peers1.out - # other lines should look as a nid followed by 1 number, a word, 6 numbers, e.g.: - # 0@lo 1 NA 0 0 0 0 0 0 - TOTAL_LINES=$(cat $TMP/lnet_peers.out |wc -l) - OTHER_LINES=$(($TOTAL_LINES - 1)) - MATCHED_LINES=$(cat $TMP/lnet_peers.out |tail -$TOTAL_LINES | - grep -c "^[0-9.]\+@[a-z0-9]\+ *[0-9]\+ *[a-zA-Z]\+ *[0-9]\+ *[0-9]\+ *-\?[0-9]\+ *[0-9]\+ *-\?[0-9]\+ *[0-9]\+$") - [ "$MATCHED_LINES" = "$OTHER_LINES" ] || (cat $TMP/lnet_peers.out && - error "/proc/sys/lnet/peers misformatted") - diff $TMP/lnet_peers.out $TMP/lnet_peers.sys || - error "lnet.peers does not look as expected" - rm -f $TMP/lnet_peers.out $TMP/lnet_peers.sys - - # /proc/sys/lnet/buffers should look exactly as expected - cat /proc/sys/lnet/buffers >$TMP/lnet_buffers.out - sysctl lnet.buffers |sed 's/^lnet.buffers\ =\ //g' >$TMP/lnet_buffers.sys - echo "pages count credits min" >$TMP/lnet_buffers.expected - echo " 0 0 0 0" >>$TMP/lnet_buffers.expected - echo " 1 0 0 0" >>$TMP/lnet_buffers.expected - echo " 256 0 0 0" >>$TMP/lnet_buffers.expected - diff $TMP/lnet_buffers.expected $TMP/lnet_buffers.out || - error "/proc/sys/lnet/buffers does not look as expected" - diff $TMP/lnet_buffers.expected $TMP/lnet_buffers.sys || - error "lnet.buffers does not look as expected" - rm -f $TMP/lnet_buffers.expected $TMP/lnet_buffers.out $TMP/lnet_buffers.sys - - # fisrt line of /proc/sys/lnet/nis should look exactly as expected - cat /proc/sys/lnet/nis >$TMP/lnet_nis.out - sysctl lnet.nis |sed 's/^lnet.nis\ =\ //g' >$TMP/lnet_nis.sys - head -1 $TMP/lnet_nis.out > $TMP/lnet_nis1.out - echo "nid refs peer rtr max tx min" >$TMP/lnet_nis1.expected - diff $TMP/lnet_nis1.expected $TMP/lnet_nis1.out || - error "first line of /proc/sys/lnet/nis does not look as expected" - rm -f $TMP/lnet_nis1.expected $TMP/lnet_nis1.out - # other lines should look as a nid followed by 6 numbers, e.g.: - # 0@lo 3 0 0 0 0 0 - TOTAL_LINES=$(cat $TMP/lnet_nis.out |wc -l) - OTHER_LINES=$(($TOTAL_LINES - 1)) - MATCHED_LINES=$(cat $TMP/lnet_nis.out |tail -$TOTAL_LINES | - grep -c "^[0-9.]\+@[a-z0-9]\+ *[0-9]\+ *[0-9]\+ *[0-9]\+ *[0-9]\+ *[0-9]\+ *[0-9]\+$") - [ "$MATCHED_LINES" = "$OTHER_LINES" ] || (cat $TMP/lnet_nis.out && - error "/proc/sys/lnet/nis misformatted") - diff $TMP/lnet_nis.out $TMP/lnet_nis.sys || - error "lnet.nis does not look as expected" - rm -f $TMP/lnet_nis.out $TMP/lnet_nis.sys +# having "abc" as 1st arg, creates $TMP/lnet_abc.out and $TMP/lnet_abc.sys +create_lnet_proc_files() { + cat /proc/sys/lnet/$1 >$TMP/lnet_$1.out || error "cannot read /proc/sys/lnet/$1" + sysctl lnet.$1 >$TMP/lnet_$1.sys_tmp || error "cannot read lnet.$1" + + sed "s/^lnet.$1\ =\ //g" "$TMP/lnet_$1.sys_tmp" >$TMP/lnet_$1.sys + rm -f "$TMP/lnet_$1.sys_tmp" +} + +# counterpart of create_lnet_proc_files +remove_lnet_proc_files() { + rm -f $TMP/lnet_$1.out $TMP/lnet_$1.sys +} + +# uses 1st arg as trailing part of filename, 2nd arg as description for reports, +# 3rd arg as regexp for body +check_lnet_proc_stats() { + local l=$(cat "$TMP/lnet_$1" |wc -l) + [ $l = 1 ] || (cat "$TMP/lnet_$1" && error "$2 is not of 1 line: $l") + + grep -E "$3" "$TMP/lnet_$1" || (cat "$TMP/lnet_$1" && error "$2 misformatted") +} + +# uses 1st arg as trailing part of filename, 2nd arg as description for reports, +# 3rd arg as regexp for body, 4th arg as regexp for 1st line, 5th arg is +# optional and can be regexp for 2nd line (lnet.routes case) +check_lnet_proc_entry() { + local blp=2 # blp stands for 'position of 1st line of body' + [ "$5" = "" ] || blp=3 # lnet.routes case + + local l=$(cat "$TMP/lnet_$1" |wc -l) + # subtracting one from $blp because the body can be empty + [ "$l" -ge "$(($blp - 1))" ] || (cat "$TMP/lnet_$1" && error "$2 is too short: $l") + + sed -n '1 p' "$TMP/lnet_$1" |grep -E "$4" >/dev/null || + (cat "$TMP/lnet_$1" && error "1st line of $2 misformatted") + + [ "$5" = "" ] || sed -n '2 p' "$TMP/lnet_$1" |grep -E "$5" >/dev/null || + (cat "$TMP/lnet_$1" && error "2nd line of $2 misformatted") + + # bail out if any unexpected line happened + sed -n "$blp~1 p" "$TMP/lnet_$1" |grep -Ev "$3" + [ "$?" != 0 ] || error "$2 misformatted" +} + +test_215() { # for bugs 18102, 21079, 21517 + local N='(0|[1-9][0-9]*)' # non-negative numeric + local P='[1-9][0-9]*' # positive numeric + local I='(0|-?[1-9][0-9]*)' # any numeric (0 | >0 | <0) + local NET='[a-z][a-z0-9]*' # LNET net like o2ib2 + local ADDR='[0-9.]+' # LNET addr like 10.0.0.1 + local NID="$ADDR@$NET" # LNET nid like 10.0.0.1@o2ib2 + + local L1 # regexp for 1st line + local L2 # regexp for 2nd line (optional) + local BR # regexp for the rest (body) + + # /proc/sys/lnet/stats should look as 11 space-separated non-negative numerics + BR="^$N $N $N $N $N $N $N $N $N $N $N$" + create_lnet_proc_files "stats" + check_lnet_proc_stats "stats.out" "/proc/sys/lnet/stats" "$BR" + check_lnet_proc_stats "stats.sys" "lnet.stats" "$BR" + remove_lnet_proc_files "stats" + + # /proc/sys/lnet/routes should look like this: + # Routing disabled/enabled + # net hops state router + # where net is a string like tcp0, hops >= 0, state is up/down, + # router is a string like 192.168.1.1@tcp2 + L1="^Routing (disabled|enabled)$" + L2="^net +hops +state +router$" + BR="^$NET +$N +(up|down) +$NID$" + create_lnet_proc_files "routes" + check_lnet_proc_entry "routes.out" "/proc/sys/lnet/routes" "$BR" "$L1" "$L2" + check_lnet_proc_entry "routes.sys" "lnet.routes" "$BR" "$L1" "$L2" + remove_lnet_proc_files "routes" + + # /proc/sys/lnet/routers should look like this: + # ref rtr_ref alive_cnt state last_ping ping_sent deadline down_ni router + # where ref > 0, rtr_ref > 0, alive_cnt >= 0, state is up/down, + # last_ping >= 0, ping_sent is boolean (0/1), deadline and down_ni are + # numeric (0 or >0 or <0), router is a string like 192.168.1.1@tcp2 + L1="^ref +rtr_ref +alive_cnt +state +last_ping +ping_sent +deadline +down_ni +router$" + BR="^$P +$P +$N +(up|down) +$N +(0|1) +$I +$I +$NID$" + create_lnet_proc_files "routers" + check_lnet_proc_entry "routers.out" "/proc/sys/lnet/routers" "$BR" "$L1" + check_lnet_proc_entry "routers.sys" "lnet.routers" "$BR" "$L1" + remove_lnet_proc_files "routers" + + # /proc/sys/lnet/peers should look like this: + # nid refs state max rtr min tx min queue + # where nid is a string like 192.168.1.1@tcp2, refs > 0, + # state is up/down/NA, max >= 0. rtr, min, tx, min are + # numeric (0 or >0 or <0), queue >= 0. + L1="^nid +refs +state +max +rtr +min +tx +min +queue$" + BR="^$NID +$P +(up|down|NA) +$N +$I +$I +$I +$I +$N$" + create_lnet_proc_files "peers" + check_lnet_proc_entry "peers.out" "/proc/sys/lnet/peers" "$BR" "$L1" + check_lnet_proc_entry "peers.sys" "lnet.peers" "$BR" "$L1" + remove_lnet_proc_files "peers" + + # /proc/sys/lnet/buffers should look like this: + # pages count credits min + # where pages >=0, count >=0, credits and min are numeric (0 or >0 or <0) + L1="^pages +count +credits +min$" + BR="^ +$N +$N +$I +$I$" + create_lnet_proc_files "buffers" + check_lnet_proc_entry "buffers.out" "/proc/sys/lnet/buffers" "$BR" "$L1" + check_lnet_proc_entry "buffers.sys" "lnet.buffers" "$BR" "$L1" + remove_lnet_proc_files "buffers" + + # /proc/sys/lnet/nis should look like this: + # nid status alive refs peer rtr max tx min + # where nid is a string like 192.168.1.1@tcp2, status is up/down, + # alive is numeric (0 or >0 or <0), refs > 0, peer >= 0, + # rtr >= 0, max >=0, tx and min are numeric (0 or >0 or <0). + L1="^nid +status +alive +refs +peer +rtr +max +tx +min$" + BR="^$NID +(up|down) +$I +$P +$N +$N +$N +$I +$I$" + create_lnet_proc_files "nis" + check_lnet_proc_entry "nis.out" "/proc/sys/lnet/nis" "$BR" "$L1" + check_lnet_proc_entry "nis.sys" "lnet.nis" "$BR" "$L1" + remove_lnet_proc_files "nis" # can we successfully write to /proc/sys/lnet/stats? echo "0" >/proc/sys/lnet/stats || error "cannot write to /proc/sys/lnet/stats" sysctl -w lnet.stats=0 || error "cannot write to lnet.stats" } -run_test 215 "/proc/sys/lnet exists and has proper content - bug 18102" +run_test 215 "/proc/sys/lnet exists and has proper content - bugs 18102, 21079, 21517" test_216() { # bug 20317 local node