btrfs-progs: tests: Enahance INSTRUMENT coverage
v2 cover btrfs-progs: tests: Enahance INSTRUMENT coverage
This patchset will enhance INSTRUMENT coverage for all btrfs related commands.
Since now INSTRUMENT would also cover a lot of btrfs inspect-dump comands, fix some test cases where INSTRUMENT output could easily pollute them.
Now fsck/convert/misc/mkfs all pass with
INSTRUMENT="valgrind --leak-check=full".
Changelog:
v2:
- Pass INSTRUMENT as space split command
- Unset previous cmmd_array in expand_command()
Instead of unsetting it.
- Use local variable @i in expand_command()
To fix misc/004
- Add extra commands for INSTRUMENT coverage
This incldues btrfstune, btrfs-corrupt-block, btrfs-image, btrfs-select-super, btrfs-find-root.
- Fix test cases which uses run_check_stdout without filtering
Qu Wenruo (2):
btrfs-progs: tests: Don't use run_check_stdout without filtering
btrfs-progs: tests: Introduce expand_command() to inject aruguments
more accurately
tests/common | 179 +++++++++--------- tests/misc-tests/004-shrink-fs/test.sh | 2 +- .../009-subvolume-sync-must-wait/test.sh | 2 +- .../013-subvolume-sync-crash/test.sh | 2 +- .../024-inspect-internal-rootid/test.sh | 14 +- .../031-qgroup-parent-child-relation/test.sh | 4 +- 6 files changed, 102 insertions(+), 101 deletions(-)
Signature · 1 line
2.26.0
PATCH v2 1/2 btrfs-progs: tests: Don't use run_check_stdout without filtering
Since run_check_stdout() can insert INSTRUMENT, which could easily pollute the stdout, any caller of run_check_stdout() should filter its output before using it.
There are some callers which just grab the output without filtering it, most of them are simple inspect-dump commands.
To avoid false alert with INSTRUMENT set, just don't utilize run_check_stdout() for those call sites.
Since those call sites are pretty simple, shouldn't cause too many holes in the coverage.
--- tests/common | 3 +++ tests/misc-tests/004-shrink-fs/test.sh | 2 +- .../009-subvolume-sync-must-wait/test.sh | 2 +- tests/misc-tests/013-subvolume-sync-crash/test.sh | 2 +- .../misc-tests/024-inspect-internal-rootid/test.sh | 14 +++++++------- .../031-qgroup-parent-child-relation/test.sh | 4 ++-- 6 files changed, 15 insertions(+), 12 deletions(-)
tests/common +3 -0
@@ -169,6 +169,9 @@ run_check() # same as run_check but the stderr+stdout output is duplicated on stdout and # can be processed further +# +# NOTE: If this function is called on btrfs related command, caller should +# filter the output, as INSTRUMENT can easily pollute the output. run_check_stdout() { local spec
tests/misc-tests/004-shrink-fs/test.sh +1 -1
@@ -14,7 +14,7 @@ setup_root_helper # Optionally take id of the device to shrink shrink_test() { - min_size=$(run_check_stdout $SUDO_HELPER "$TOP/btrfs" inspect-internal min-dev-size ${1:+--id "$1"} "$TEST_MNT") + min_size=$($SUDO_HELPER "$TOP/btrfs" inspect-internal min-dev-size ${1:+--id "$1"} "$TEST_MNT") min_size=$(echo "$min_size" | cut -d ' ' -f 1) _log "min size = ${min_size}" if [ -z "$min_size" ]; then
tests/misc-tests/009-subvolume-sync-must-wait/test.sh +1 -1
@@ -31,7 +31,7 @@ done run_check $SUDO_HELPER "$TOP/btrfs" subvolume list . run_check $SUDO_HELPER "$TOP/btrfs" subvolume list -d . -idtodel=`run_check_stdout $SUDO_HELPER "$TOP/btrfs" inspect-internal rootid snap3` +idtodel=`$SUDO_HELPER "$TOP/btrfs" inspect-internal rootid snap3` # delete, sync after some time run_check $SUDO_HELPER "$TOP/btrfs" subvolume delete -c snap3
tests/misc-tests/013-subvolume-sync-crash/test.sh +1 -1
@@ -32,7 +32,7 @@ done run_check $SUDO_HELPER "$TOP/btrfs" subvolume list . run_check $SUDO_HELPER "$TOP/btrfs" subvolume list -d . -idtodel=`run_check_stdout $SUDO_HELPER "$TOP/btrfs" inspect-internal rootid snap3` +idtodel=`$SUDO_HELPER "$TOP/btrfs" inspect-internal rootid snap3` # delete, sync after some time run_check $SUDO_HELPER "$TOP/btrfs" subvolume delete -c snap*
tests/misc-tests/024-inspect-internal-rootid/test.sh +7 -7
@@ -21,19 +21,19 @@ run_check touch file1 run_check touch dir/file2 run_check touch sub/file3 -id1=$(run_check_stdout "$TOP/btrfs" inspect-internal rootid .) \ +id1=$("$TOP/btrfs" inspect-internal rootid .) \ || { echo $id1; exit 1; } -id2=$(run_check_stdout "$TOP/btrfs" inspect-internal rootid sub) \ +id2=$("$TOP/btrfs" inspect-internal rootid sub) \ || { echo $id2; exit 1; } -id3=$(run_check_stdout "$TOP/btrfs" inspect-internal rootid sub/subsub) \ +id3=$("$TOP/btrfs" inspect-internal rootid sub/subsub) \ || { echo $id3; exit 1; } -id4=$(run_check_stdout "$TOP/btrfs" inspect-internal rootid dir) \ +id4=$("$TOP/btrfs" inspect-internal rootid dir) \ || { echo $id4; exit 1; } -id5=$(run_check_stdout "$TOP/btrfs" inspect-internal rootid file1) \ +id5=$("$TOP/btrfs" inspect-internal rootid file1) \ || { echo $id5; exit 1; } -id6=$(run_check_stdout "$TOP/btrfs" inspect-internal rootid dir/file2) \ +id6=$("$TOP/btrfs" inspect-internal rootid dir/file2) \ || { echo $id6; exit 1; } -id7=$(run_check_stdout "$TOP/btrfs" inspect-internal rootid sub/file3) \ +id7=$("$TOP/btrfs" inspect-internal rootid sub/file3) \ || { echo $id7; exit 1; } if ! ([ "$id1" -ne "$id2" ] && [ "$id1" -ne "$id3" ] && [ "$id2" -ne "$id3" ]); then
tests/misc-tests/031-qgroup-parent-child-relation/test.sh +2 -2
@@ -18,10 +18,10 @@ run_check $SUDO_HELPER "$TOP/btrfs" qgroup assign 0/5 1/0 "$TEST_MNT" run_check $SUDO_HELPER "$TOP/btrfs" quota rescan -w "$TEST_MNT" run_check_stdout $SUDO_HELPER "$TOP/btrfs" qgroup show --sort=-qgroupid \ - -p "$TEST_MNT" | tail -n 1 | grep -q "1/0" \ + -p "$TEST_MNT" | grep -q "1/0" \ || _fail "parent qgroup check failed, please check the log" run_check_stdout $SUDO_HELPER "$TOP/btrfs" qgroup show --sort=qgroupid \ - -c "$TEST_MNT" | tail -n 1 | grep -q "0/5" \ + -c "$TEST_MNT" | grep -q "0/5" \ || _fail "child qgroup check failed, please check the log" run_check_umount_test_dev "$TEST_MNT"
Signature · 1 line
2.26.0
Re: [PATCH v2 1/2] btrfs-progs: tests: Don't use run_check_stdout without filtering
Quoted 11 lines · Qu Wenruo
Since
run_check_stdout()can insert INSTRUMENT, which could easily pollute the stdout, any caller ofrun_check_stdout()should filter its output before using it.There are some callers which just grab the output without filtering it, most of them are simple inspect-dump commands.
To avoid false alert with INSTRUMENT set, just don't utilize
run_check_stdout()for those call sites.
Since those call sites are pretty simple, shouldn't cause too many holes in the coverage.
I don't like this, it removes a lot of the coverage. The instrument tools should provide a way to avoid stdout/stderr pollution, eg. valgrind has --log-fd or --log-file. We can add a shortcut that will provide some defaults so we can conveniently use 'INSTRUMENT=valgrind'.
Re: [PATCH v2 1/2] btrfs-progs: tests: Don't use run_check_stdout without filtering
Quoted 13 lines · David Sterba
Quoted 11 lines · Qu Wenruo
Since
run_check_stdout()can insert INSTRUMENT, which could easily pollute the stdout, any caller ofrun_check_stdout()should filter its output before using it.There are some callers which just grab the output without filtering it, most of them are simple inspect-dump commands.
To avoid false alert with INSTRUMENT set, just don't utilize
run_check_stdout()for those call sites.
Since those call sites are pretty simple, shouldn't cause too many holes in the coverage.
Quoted 5 lines
I don't like this, it removes a lot of the coverage. The instrument tools should provide a way to avoid stdout/stderr pollution, eg. valgrind has --log-fd or --log-file. We can add a shortcut that will provide some defaults so we can conveniently use 'INSTRUMENT=valgrind'.
Then that's too INSTURMENT dependent.
What about adding filtering for the mentioned callers?
Thanks,
Qu
Re: [PATCH v2 1/2] btrfs-progs: tests: Don't use run_check_stdout without filtering
Quoted 14 lines · Qu Wenruo
Quoted 13 lines · David Sterba
Quoted 11 lines · Qu Wenruo
Since
run_check_stdout()can insert INSTRUMENT, which could easily pollute the stdout, any caller ofrun_check_stdout()should filter its output before using it.There are some callers which just grab the output without filtering it, most of them are simple inspect-dump commands.
To avoid false alert with INSTRUMENT set, just don't utilize
run_check_stdout()for those call sites.
Since those call sites are pretty simple, shouldn't cause too many holes in the coverage.
Quoted 6 lines
Quoted 5 lines
I don't like this, it removes a lot of the coverage. The instrument tools should provide a way to avoid stdout/stderr pollution, eg. valgrind has --log-fd or --log-file. We can add a shortcut that will provide some defaults so we can conveniently use 'INSTRUMENT=valgrind'.
Then that's too INSTURMENT dependent.
Yes, but how many do we use besides valgrind. I have tried gdb with some init script or batch mode to catch crashes, but I can't think of other tools.
Quoted 1 line
What about adding filtering for the mentioned callers?
I'm not sure what exactly do you mean, can you send an example?
Re: [PATCH v2 1/2] btrfs-progs: tests: Don't use run_check_stdout without filtering
Quoted 5 lines · David Sterba
Quoted 2 lines · Qu Wenruo
On 2020/4/7 上午12:56, David Sterba wrote:
What about adding filtering for the mentioned callers?I'm not sure what exactly do you mean, can you send an example?
Never mind, it's in the v3 patches.
PATCH v2 2/2 btrfs-progs: tests: Introduce expand_command() to inject aruguments more accurately
In reply to Qu Wenruo[PROBLEM]
We want to inject $INSTRUMENT (mostly valgrind) before btrfs command but after root_helper.
Currently we won't inject $INSTRUMENT at all if we are using root_helper.
This means the coverage is not good enough.
[FIX]
This patch introduce a new function, expand_command(), to handle all parameter/argument injection, including existing 'btrfs check' inject.
This function will:
- Detect where to inject $INSTRUMENT
If we have root_helper and the command is target command (btrfs/mkfs.btrfs/btrfs-convert), then we inject $INSTRUMENT after root_helper.
If we don't have root_helper, and the command is target command, we inject $INSTRUMENT before the command.
Or we don't inject $INSTRUMENT (it's not the target command).
- Use existing spec facility to inject extra arguments
- Use an array to restore to result
To avoid bash interpret the IFS inside path/commands.
Now we can make sure no matter if we use root_helper, $INSTRUMENT is always injected corrected.
--- tests/common | 176 +++++++++++++++++++++++++-------------------------- 1 file changed, 87 insertions(+), 89 deletions(-)
tests/common +87 -89
@@ -3,6 +3,9 @@ # Common routines for all tests # +# Temporary command array for building real command +declare -a cmd_array + # assert that argument is not empty and is an existing path (file or directory) _assert_path() {
@@ -135,6 +138,60 @@ _cmd_spec() fi } +_is_target_command() +{ + [[ $1 =~ /btrfs$ ]] && return 0 + [[ $1 =~ /mkfs.btrfs$ ]] && return 0 + [[ $1 =~ /btrfs-convert$ ]] && return 0 + [[ $1 =~ /btrfs-corrupt-block$ ]] && return 0 + [[ $1 =~ /btrfs-image$ ]] && return 0 + [[ $1 =~ /btrfs-select-super$ ]] && return 0 + [[ $1 =~ /btrfs-find-root$ ]] && return 0 + [[ $1 =~ /btrfstune$ ]] && return 0 + return 1 +} + +# Expanding extra commands/options for current command string +# This function is responsible for inserting the following things: +# - @INSTRUMENT before btrfs commands +# To ensure that @INSTRUMENT is not executed for things like sudo/mount +# which normally has setuid/setgid which will not be traced. +# +# - Add extra arguments for 'btrfs check'/'mkfs.btrfs'/'btrfs-convert' +# This allows us to test certain function like lowmem mode for btrfs check +expand_command() +{ + local command_index + local ins + local spec + local i + + ins=$(_get_spec_ins "$@") + spec=$(($ins-1)) + spec=$(_cmd_spec "${@:$spec}") + cmd_array=() + + if [ "$1" = 'root_helper' ] && _is_target_command "$2" ; then + command_index=2 + elif _is_target_command "$1" ; then + command_index=1 + else + # Since we iterate from offset 1, this never get hit, + # thus we won't insert $INSTRUMENT + command_index=0 + fi + + for ((i = 1; i <= $#; i++ )); do + if [ $i -eq $command_index -a ! -z "$INSTRUMENT" ]; then + cmd_array+=($INSTRUMENT) + fi + if [ $i -eq $ins -a ! -z "$spec" ]; then + cmd_array+=("$spec") + fi + cmd_array+=("${!i}") + done +} + # Argument passing magic: # the command passed to run_* helpers is inspected, if there's 'btrfs command' # found and there are defined additional arguments, they're inserted just after
@@ -145,26 +202,11 @@ _cmd_spec() run_check() { - local spec - local ins + expand_command "$@" + echo "====== RUN CHECK ${cmd_array[@]}" >> "$RESULTS" 2>&1 + if [[ $TEST_LOG =~ tty ]]; then echo "CMD: ${cmd_array[@]}" > /dev/tty; fi - ins=$(_get_spec_ins "$@") - spec=$(($ins-1)) - spec=$(_cmd_spec "${@:$spec}") - set -- "${@:1:$(($ins-1))}" $spec "${@: $ins}" - echo "====== RUN CHECK $@" >> "$RESULTS" 2>&1 - if [[ $TEST_LOG =~ tty ]]; then echo "CMD: $@" > /dev/tty; fi - - # If the command is `root_helper` or mount/umount, don't call INSTRUMENT - # as most profiling tool like valgrind can't handle setuid/setgid/setcap - # which mount normally has. - # And since mount/umount is only called with run_check(), we don't need - # to do the same check on other run_*() functions. - if [ "$1" = 'root_helper' -o "$1" = 'mount' -o "$1" = 'umount' ]; then - "$@" >> "$RESULTS" 2>&1 || _fail "failed: $@" - else - $INSTRUMENT "$@" >> "$RESULTS" 2>&1 || _fail "failed: $@" - fi + "${cmd_array[@]}" >> "$RESULTS" 2>&1 || _fail "failed: ${cmd_array[@]}" } # same as run_check but the stderr+stdout output is duplicated on stdout and
@@ -174,20 +216,11 @@ run_check() # filter the output, as INSTRUMENT can easily pollute the output. run_check_stdout() { - local spec - local ins - - ins=$(_get_spec_ins "$@") - spec=$(($ins-1)) - spec=$(_cmd_spec "${@:$spec}") - set -- "${@:1:$(($ins-1))}" $spec "${@: $ins}" - echo "====== RUN CHECK $@" >> "$RESULTS" 2>&1 - if [[ $TEST_LOG =~ tty ]]; then echo "CMD(stdout): $@" > /dev/tty; fi - if [ "$1" = 'root_helper' ]; then - "$@" 2>&1 | tee -a "$RESULTS" - else - $INSTRUMENT "$@" 2>&1 | tee -a "$RESULTS" - fi + expand_command "$@" + echo "====== RUN CHECK ${cmd_array[@]}" >> "$RESULTS" 2>&1 + if [[ $TEST_LOG =~ tty ]]; then echo "CMD(stdout): ${cmd_array[@]}" \ + > /dev/tty; fi + "${cmd_array[@]}" 2>&1 | tee -a "$RESULTS" if [ ${PIPESTATUS[0]} -ne 0 ]; then _fail "failed: $@" fi
@@ -198,21 +231,11 @@ run_check_stdout() # output is logged run_mayfail() { - local spec - local ins - local ret - - ins=$(_get_spec_ins "$@") - spec=$(($ins-1)) - spec=$(_cmd_spec "${@:$spec}") - set -- "${@:1:$(($ins-1))}" $spec "${@: $ins}" - echo "====== RUN MAYFAIL $@" >> "$RESULTS" 2>&1 - if [[ $TEST_LOG =~ tty ]]; then echo "CMD(mayfail): $@" > /dev/tty; fi - if [ "$1" = 'root_helper' ]; then - "$@" >> "$RESULTS" 2>&1 - else - $INSTRUMENT "$@" >> "$RESULTS" 2>&1 - fi + expand_command "$@" + echo "====== RUN MAYFAIL ${cmd_array[@]}" >> "$RESULTS" 2>&1 + if [[ $TEST_LOG =~ tty ]]; then echo "CMD(mayfail): ${cmd_array[@]}" \ + > /dev/tty; fi + "${cmd_array[@]}" >> "$RESULTS" 2>&1 ret=$? if [ $ret != 0 ]; then echo "failed (ignored, ret=$ret): $@" >> "$RESULTS"
@@ -234,23 +257,13 @@ run_mayfail() # store the output to a variable for further processing. run_mayfail_stdout() { - local spec - local ins - local ret - tmp_output=$(mktemp --tmpdir btrfs-progs-test--mayfail-stdtout.XXXXXX) - ins=$(_get_spec_ins "$@") - spec=$(($ins-1)) - spec=$(_cmd_spec "${@:$spec}") - set -- "${@:1:$(($ins-1))}" $spec "${@: $ins}" - echo "====== RUN MAYFAIL $@" >> "$RESULTS" 2>&1 - if [[ $TEST_LOG =~ tty ]]; then echo "CMD(mayfail): $@" > /dev/tty; fi - if [ "$1" = 'root_helper' ]; then - "$@" 2>&1 > "$tmp_output" - else - $INSTRUMENT "$@" 2>&1 > "$tmp_output" - fi + expand_command "$@" + echo "====== RUN MAYFAIL ${cmd_array[@]}" >> "$RESULTS" 2>&1 + if [[ $TEST_LOG =~ tty ]]; then echo "CMD(mayfail): ${cmd_array[@]}" \ + > /dev/tty; fi + "${cmd_array[@]}" 2>&1 > "$tmp_output" ret=$? cat "$tmp_output" >> "$RESULTS"
@@ -275,8 +288,6 @@ run_mayfail_stdout() # same as run_check but expects the command to fail, output is logged run_mustfail() { - local spec - local ins local msg msg="$1"
@@ -287,17 +298,12 @@ run_mustfail() exit 1 fi - ins=$(_get_spec_ins "$@") - spec=$(($ins-1)) - spec=$(_cmd_spec "${@:$spec}") - set -- "${@:1:$(($ins-1))}" $spec "${@: $ins}" - echo "====== RUN MUSTFAIL $@" >> "$RESULTS" 2>&1 - if [[ $TEST_LOG =~ tty ]]; then echo "CMD(mustfail): $@" > /dev/tty; fi - if [ "$1" = 'root_helper' ]; then - "$@" >> "$RESULTS" 2>&1 - else - $INSTRUMENT "$@" >> "$RESULTS" 2>&1 - fi + + expand_command "$@" + echo "====== RUN MUSTFAIL ${cmd_array[@]}" >> "$RESULTS" 2>&1 + if [[ $TEST_LOG =~ tty ]]; then echo "CMD(mustfail): ${cmd_array[@]}" \ + > /dev/tty; fi + "${cmd_array[@]}" >> "$RESULTS" 2>&1 if [ $? != 0 ]; then echo "failed (expected): $@" >> "$RESULTS" return 0
@@ -315,8 +321,6 @@ run_mustfail() # So it doesn't support pipeline in the @cmd run_mustfail_stdout() { - local spec - local ins local msg local ret local tmp_output
@@ -331,17 +335,11 @@ run_mustfail_stdout() exit 1 fi - ins=$(_get_spec_ins "$@") - spec=$(($ins-1)) - spec=$(_cmd_spec "${@:$spec}") - set -- "${@:1:$(($ins-1))}" $spec "${@: $ins}" - echo "====== RUN MUSTFAIL $@" >> "$RESULTS" 2>&1 - if [[ $TEST_LOG =~ tty ]]; then echo "CMD(mustfail): $@" > /dev/tty; fi - if [ "$1" = 'root_helper' ]; then - "$@" 2>&1 > "$tmp_output" - else - $INSTRUMENT "$@" 2>&1 > "$tmp_output" - fi + expand_command "$@" + echo "====== RUN MUSTFAIL ${cmd_array[@]}" >> "$RESULTS" 2>&1 + if [[ $TEST_LOG =~ tty ]]; then echo "CMD(mustfail): ${cmd_array[@]}" \ + > /dev/tty; fi + "${cmd_array[@]}" 2>&1 > "$tmp_output" ret=$? cat "$tmp_output" >> "$RESULTS"
Signature · 1 line
2.26.0