perf test: Improve perf record tests (v2)
Namhyung Kim
maintainercover perf test: Improve perf record tests (v2)
Hello,
This patchset improves the perf record tests to check more cases so that it can find problems early. The motivation was a failure per-thread mmap with multi-threaded targets which Adrian is working on the fix.
Changes in v2)
* fix shellcheck issues
* drop unsupported --per-thread and --threads combination
* do not use initial delay (-D option); instead it runs the target and wait for it separately using the recent waiting.sh library
* add Adrian's Reviewed-by tags
I added a custom test program and more combinations like system-wide and command line workload (in per-process mode) testing with multi-threaded recording mode.
Currently it succeeds every tests when running as root!
$ sudo ./perf test -v 91 91: perf record tests : --- start --- test child forked, pid 108975 Build a test program Basic --per-thread mode test Basic --per-thread mode test [Success] Register capture test Register capture test [Success] Basic --system-wide mode test Basic --system-wide mode test [Success] Basic target workload test Basic target workload test [Success] test child finished with 0 ---- end ---- perf record tests: Ok
You can find it in 'perf/record-test-v2' branch in
git://git.kernel.org/pub/scm/linux/kernel/git/namhyung/linux-perf.git
Thanks,
Namhyung
Namhyung Kim (8):
perf test: Do not use instructions:u explicitly
perf test: Fix shellcheck issues in the record test perf test: Use a test program in perf record tests perf test: Wait for a new thread when testing --per-thread record perf test: Add system-wide mode in perf record tests perf test: Add target workload test in perf record tests perf test: Test record with --threads option
perf test: Do not set TEST_SKIP for record subtests
tools/perf/tests/shell/record.sh | 180 +++++++++++++++++++++++++++---- 1 file changed, 158 insertions(+), 22 deletions(-)
base-commit: a3a365655a28f12f07eddf4f3fd596987b175e1d
Signature · 1 line
2.38.0.135.g90850a2211-goog
Namhyung Kim
maintainerPATCH 1/8 perf test: Do not use instructions:u explicitly
I think it's to support non-root user tests. But perf record can handle the case and fall back to a software event (cpu-clock). Practically this would affect when it's run on a VM, but it seems no reason to prevent running the test in the guest.
--- tools/perf/tests/shell/record.sh | 10 +++++----- 1 file changed, 5 insertions(+), 5 deletions(-)
tools/perf/tests/shell/record.sh +5 -5
@@ -21,18 +21,18 @@ trap trap_cleanup exit term int test_per_thread() { echo "Basic --per-thread mode test" - if ! perf record -e instructions:u -o ${perfdata} --quiet true 2> /dev/null + if ! perf record -o /dev/null --quiet true 2> /dev/null then - echo "Per-thread record [Skipped instructions:u not supported]" + echo "Per-thread record [Skipped event not supported]" if [ $err -ne 1 ] then err=2 fi return fi - if ! perf record -e instructions:u --per-thread -o ${perfdata} true 2> /dev/null + if ! perf record --per-thread -o ${perfdata} true 2> /dev/null then - echo "Per-thread record of instructions:u [Failed]" + echo "Per-thread record [Failed record]" err=1 return fi
@@ -49,7 +49,7 @@ test_register_capture() { echo "Register capture test" if ! perf list | egrep -q 'br_inst_retired.near_call' then - echo "Register capture test [Skipped missing instruction]" + echo "Register capture test [Skipped missing event]" if [ $err -ne 1 ] then err=2
Signature · 1 line
2.38.0.135.g90850a2211-goog
Re: [PATCH 1/8] perf test: Do not use instructions:u explicitly
Quoted 8 lines · Namhyung Kim
I think it's to support non-root user tests. But perf record can handle the case and fall back to a software event (cpu-clock). Practically this would affect when it's run on a VM, but it seems no reason to prevent running the test in the guest.
Reviewed-by Adrian Hunter <[email protected]>Signed-off-by Namhyung Kim <[email protected]>
Thanks,
Ian
Quoted 43 lines
--- tools/perf/tests/shell/record.sh | 10 +++++----- 1 file changed, 5 insertions(+), 5 deletions(-)tools/perf/tests/shell/record.sh +5
-6@@ -21,18 +21,18 @@ trap trap_cleanup exit term int test_per_thread() { echo "Basic --per-thread mode test" - if ! perf record -e instructions:u -o ${perfdata} --quiet true 2> /dev/null + if ! perf record -o /dev/null --quiet true 2> /dev/null then - echo "Per-thread record [Skipped instructions:u not supported]" + echo "Per-thread record [Skipped event not supported]" if [ $err -ne 1 ] then err=2 fi return fi - if ! perf record -e instructions:u --per-thread -o ${perfdata} true 2> /dev/null + if ! perf record --per-thread -o ${perfdata} true 2> /dev/null then - echo "Per-thread record of instructions:u [Failed]" + echo "Per-thread record [Failed record]" err=1 return fi@@ -49,7 +49,7 @@ test_register_capture() { echo "Register capture test" if ! perf list | egrep -q 'br_inst_retired.near_call' then - echo "Register capture test [Skipped missing instruction]" + echo "Register capture test [Skipped missing event]" if [ $err -ne 1 ] then err=2 --2.38.0.135.g90850a2211-goog
Namhyung Kim
maintainerPATCH 2/8 perf test: Fix shellcheck issues in the record test
In reply to Namhyung KimBasically there are 3 issues:
1. quote shell expansion
2. do not use egrep
3. use upper case letters for signal names
--- tools/perf/tests/shell/record.sh | 18 +++++++++--------- 1 file changed, 9 insertions(+), 9 deletions(-)
tools/perf/tests/shell/record.sh +9 -9
@@ -8,16 +8,16 @@ err=0 perfdata=$(mktemp /tmp/__perf_test.perf.data.XXXXX) cleanup() { - rm -f ${perfdata} - rm -f ${perfdata}.old - trap - exit term int + rm -f "${perfdata}" + rm -f "${perfdata}".old + trap - EXIT TERM INT } trap_cleanup() { cleanup exit 1 } -trap trap_cleanup exit term int +trap trap_cleanup EXIT TERM INT test_per_thread() { echo "Basic --per-thread mode test"
@@ -30,13 +30,13 @@ test_per_thread() { fi return fi - if ! perf record --per-thread -o ${perfdata} true 2> /dev/null + if ! perf record --per-thread -o "${perfdata}" true 2> /dev/null then echo "Per-thread record [Failed record]" err=1 return fi - if ! perf report -i ${perfdata} -q | egrep -q true + if ! perf report -i "${perfdata}" -q | grep -q true then echo "Per-thread record [Failed missing output]" err=1
@@ -47,7 +47,7 @@ test_per_thread() { test_register_capture() { echo "Register capture test" - if ! perf list | egrep -q 'br_inst_retired.near_call' + if ! perf list | grep -q 'br_inst_retired.near_call' then echo "Register capture test [Skipped missing event]" if [ $err -ne 1 ]
@@ -56,7 +56,7 @@ test_register_capture() { fi return fi - if ! perf record --intr-regs=\? 2>&1 | egrep -q 'available registers: AX BX CX DX SI DI BP SP IP FLAGS CS SS R8 R9 R10 R11 R12 R13 R14 R15' + if ! perf record --intr-regs=\? 2>&1 | grep -q 'available registers: AX BX CX DX SI DI BP SP IP FLAGS CS SS R8 R9 R10 R11 R12 R13 R14 R15' then echo "Register capture test [Skipped missing registers]" return
@@ -64,7 +64,7 @@ test_register_capture() { if ! perf record -o - --intr-regs=di,r8,dx,cx -e br_inst_retired.near_call:p \ -c 1000 --per-thread true 2> /dev/null \ | perf script -F ip,sym,iregs -i - 2> /dev/null \ - | egrep -q "DI:" + | grep -q "DI:" then echo "Register capture test [Failed missing output]" err=1
Signature · 1 line
2.38.0.135.g90850a2211-goog
Re: [PATCH 2/8] perf test: Fix shellcheck issues in the record test
Quoted 8 lines · Namhyung Kim
Basically there are 3 issues:
1. quote shell expansion
2. do not use egrep
3. use upper case letters for signal namesSigned-off-by Namhyung Kim <[email protected]>
Thanks,
Ian
Quoted 75 lines
--- tools/perf/tests/shell/record.sh | 18 +++++++++--------- 1 file changed, 9 insertions(+), 9 deletions(-)tools/perf/tests/shell/record.sh +9
-10@@ -8,16 +8,16 @@ err=0 perfdata=$(mktemp /tmp/__perf_test.perf.data.XXXXX) cleanup() { - rm -f ${perfdata} - rm -f ${perfdata}.old - trap - exit term int + rm -f "${perfdata}" + rm -f "${perfdata}".old + trap - EXIT TERM INT } trap_cleanup() { cleanup exit 1 } -trap trap_cleanup exit term int +trap trap_cleanup EXIT TERM INT test_per_thread() { echo "Basic --per-thread mode test"@@ -30,13 +30,13 @@ test_per_thread() { fi return fi - if ! perf record --per-thread -o ${perfdata} true 2> /dev/null + if ! perf record --per-thread -o "${perfdata}" true 2> /dev/null then echo "Per-thread record [Failed record]" err=1 return fi - if ! perf report -i ${perfdata} -q | egrep -q true + if ! perf report -i "${perfdata}" -q | grep -q true then echo "Per-thread record [Failed missing output]" err=1@@ -47,7 +47,7 @@ test_per_thread() { test_register_capture() { echo "Register capture test" - if ! perf list | egrep -q 'br_inst_retired.near_call' + if ! perf list | grep -q 'br_inst_retired.near_call' then echo "Register capture test [Skipped missing event]" if [ $err -ne 1 ]@@ -56,7 +56,7 @@ test_register_capture() { fi return fi - if ! perf record --intr-regs=\? 2>&1 | egrep -q 'available registers: AX BX CX DX SI DI BP SP IP FLAGS CS SS R8 R9 R10 R11 R12 R13 R14 R15' + if ! perf record --intr-regs=\? 2>&1 | grep -q 'available registers: AX BX CX DX SI DI BP SP IP FLAGS CS SS R8 R9 R10 R11 R12 R13 R14 R15' then echo "Register capture test [Skipped missing registers]" return@@ -64,7 +64,7 @@ test_register_capture() { if ! perf record -o - --intr-regs=di,r8,dx,cx -e br_inst_retired.near_call:p \ -c 1000 --per-thread true 2> /dev/null \ | perf script -F ip,sym,iregs -i - 2> /dev/null \ - | egrep -q "DI:" + | grep -q "DI:" then echo "Register capture test [Failed missing output]" err=1 --2.38.0.135.g90850a2211-goog
Re: [PATCH 2/8] perf test: Fix shellcheck issues in the record test
In reply to Namhyung KimQuoted 7 lines · Namhyung Kim
Basically there are 3 issues:
1. quote shell expansion
2. do not use egrep
3. use upper case letters for signal namesSigned-off-by Namhyung Kim <[email protected]>
Quoted 72 lines
--- tools/perf/tests/shell/record.sh | 18 +++++++++--------- 1 file changed, 9 insertions(+), 9 deletions(-)tools/perf/tests/shell/record.sh +9
-9@@ -8,16 +8,16 @@ err=0 perfdata=$(mktemp /tmp/__perf_test.perf.data.XXXXX) cleanup() { - rm -f ${perfdata} - rm -f ${perfdata}.old - trap - exit term int + rm -f "${perfdata}" + rm -f "${perfdata}".old + trap - EXIT TERM INT } trap_cleanup() { cleanup exit 1 } -trap trap_cleanup exit term int +trap trap_cleanup EXIT TERM INT test_per_thread() { echo "Basic --per-thread mode test"@@ -30,13 +30,13 @@ test_per_thread() { fi return fi - if ! perf record --per-thread -o ${perfdata} true 2> /dev/null + if ! perf record --per-thread -o "${perfdata}" true 2> /dev/null then echo "Per-thread record [Failed record]" err=1 return fi - if ! perf report -i ${perfdata} -q | egrep -q true + if ! perf report -i "${perfdata}" -q | grep -q true then echo "Per-thread record [Failed missing output]" err=1@@ -47,7 +47,7 @@ test_per_thread() { test_register_capture() { echo "Register capture test" - if ! perf list | egrep -q 'br_inst_retired.near_call' + if ! perf list | grep -q 'br_inst_retired.near_call' then echo "Register capture test [Skipped missing event]" if [ $err -ne 1 ]@@ -56,7 +56,7 @@ test_register_capture() { fi return fi - if ! perf record --intr-regs=\? 2>&1 | egrep -q 'available registers: AX BX CX DX SI DI BP SP IP FLAGS CS SS R8 R9 R10 R11 R12 R13 R14 R15' + if ! perf record --intr-regs=\? 2>&1 | grep -q 'available registers: AX BX CX DX SI DI BP SP IP FLAGS CS SS R8 R9 R10 R11 R12 R13 R14 R15' then echo "Register capture test [Skipped missing registers]" return@@ -64,7 +64,7 @@ test_register_capture() { if ! perf record -o - --intr-regs=di,r8,dx,cx -e br_inst_retired.near_call:p \ -c 1000 --per-thread true 2> /dev/null \ | perf script -F ip,sym,iregs -i - 2> /dev/null \ - | egrep -q "DI:" + | grep -q "DI:" then echo "Register capture test [Failed missing output]" err=1
Namhyung Kim
maintainerPATCH 3/8 perf test: Use a test program in perf record tests
In reply to Namhyung KimIf the system has cc it could build a test program with two threads and then use it for more detailed testing. Also it accepts an option to run a thread forever to ensure multi-thread runs.
If cc is not found, it falls back to use the default value 'true'.
--- tools/perf/tests/shell/record.sh | 64 ++++++++++++++++++++++++++++++-- 1 file changed, 60 insertions(+), 4 deletions(-)
tools/perf/tests/shell/record.sh +60 -4
@@ -6,10 +6,17 @@ set -e err=0 perfdata=$(mktemp /tmp/__perf_test.perf.data.XXXXX) +testprog=$(mktemp /tmp/__perf_test.prog.XXXXXX) +testsym="test_loop" cleanup() { rm -f "${perfdata}" rm -f "${perfdata}".old + + if [ "${testprog}" != "true" ]; then + rm -f "${testprog}" + fi + trap - EXIT TERM INT }
@@ -19,9 +26,56 @@ trap_cleanup() { } trap trap_cleanup EXIT TERM INT +build_test_program() { + if ! [ -x "$(command -v cc)" ]; then + # No CC found. Fall back to 'true' + testprog=true + testsym=true + return + fi + + echo "Build a test program" + cat <<EOF | cc -o ${testprog} -xc - -pthread +#include <stdio.h> +#include <stdlib.h> +#include <pthread.h> + +void test_loop(void) { + volatile int count = 1000000; + + while (count--) + continue; +} + +void *thfunc(void *arg) { + int forever = *(int *)arg; + + do { + test_loop(); + } while (forever); + + return NULL; +} + +int main(int argc, char *argv[]) { + pthread_t th; + int forever = 0; + + if (argc > 1) + forever = atoi(argv[1]); + + pthread_create(&th, NULL, thfunc, &forever); + test_loop(); + pthread_join(th, NULL); + + return 0; +} +EOF +} + test_per_thread() { echo "Basic --per-thread mode test" - if ! perf record -o /dev/null --quiet true 2> /dev/null + if ! perf record -o /dev/null --quiet ${testprog} 2> /dev/null then echo "Per-thread record [Skipped event not supported]" if [ $err -ne 1 ]
@@ -30,13 +84,13 @@ test_per_thread() { fi return fi - if ! perf record --per-thread -o "${perfdata}" true 2> /dev/null + if ! perf record --per-thread -o "${perfdata}" ${testprog} 2> /dev/null then echo "Per-thread record [Failed record]" err=1 return fi - if ! perf report -i "${perfdata}" -q | grep -q true + if ! perf report -i "${perfdata}" -q | grep -q "${testsym}" then echo "Per-thread record [Failed missing output]" err=1
@@ -62,7 +116,7 @@ test_register_capture() { return fi if ! perf record -o - --intr-regs=di,r8,dx,cx -e br_inst_retired.near_call:p \ - -c 1000 --per-thread true 2> /dev/null \ + -c 1000 --per-thread ${testprog} 2> /dev/null \ | perf script -F ip,sym,iregs -i - 2> /dev/null \ | grep -q "DI:" then
@@ -73,6 +127,8 @@ test_register_capture() { echo "Register capture test [Success]" } +build_test_program + test_per_thread test_register_capture
Signature · 1 line
2.38.0.135.g90850a2211-goog
Re: [PATCH 3/8] perf test: Use a test program in perf record tests
Quoted 8 lines · Namhyung Kim
If the system has cc it could build a test program with two threads and then use it for more detailed testing. Also it accepts an option to run a thread forever to ensure multi-thread runs.
If cc is not found, it falls back to use the default value 'true'.
Signed-off-by Namhyung Kim <[email protected]>
I wonder if these utilities should just be built into perf to avoid the cc dependency. Perhaps we can have a hidden option built into perf test.
Thanks,
Ian
Quoted 121 lines
--- tools/perf/tests/shell/record.sh | 64 ++++++++++++++++++++++++++++++-- 1 file changed, 60 insertions(+), 4 deletions(-)tools/perf/tests/shell/record.sh +60
-5@@ -6,10 +6,17 @@ set -e err=0 perfdata=$(mktemp /tmp/__perf_test.perf.data.XXXXX) +testprog=$(mktemp /tmp/__perf_test.prog.XXXXXX) +testsym="test_loop" cleanup() { rm -f "${perfdata}" rm -f "${perfdata}".old + + if [ "${testprog}" != "true" ]; then + rm -f "${testprog}" + fi + trap - EXIT TERM INT }@@ -19,9 +26,56 @@ trap_cleanup() { } trap trap_cleanup EXIT TERM INT +build_test_program() { + if ! [ -x "$(command -v cc)" ]; then + # No CC found. Fall back to 'true' + testprog=true + testsym=true + return + fi + + echo "Build a test program" + cat <<EOF | cc -o ${testprog} -xc - -pthread +#include <stdio.h> +#include <stdlib.h> +#include <pthread.h> + +void test_loop(void) { + volatile int count = 1000000; + + while (count--) + continue; +} + +void *thfunc(void *arg) { + int forever = *(int *)arg; + + do { + test_loop(); + } while (forever); + + return NULL; +} + +int main(int argc, char *argv[]) { + pthread_t th; + int forever = 0; + + if (argc > 1) + forever = atoi(argv[1]); + + pthread_create(&th, NULL, thfunc, &forever); + test_loop(); + pthread_join(th, NULL); + + return 0; +} +EOF +} + test_per_thread() { echo "Basic --per-thread mode test" - if ! perf record -o /dev/null --quiet true 2> /dev/null + if ! perf record -o /dev/null --quiet ${testprog} 2> /dev/null then echo "Per-thread record [Skipped event not supported]" if [ $err -ne 1 ]@@ -30,13 +84,13 @@ test_per_thread() { fi return fi - if ! perf record --per-thread -o "${perfdata}" true 2> /dev/null + if ! perf record --per-thread -o "${perfdata}" ${testprog} 2> /dev/null then echo "Per-thread record [Failed record]" err=1 return fi - if ! perf report -i "${perfdata}" -q | grep -q true + if ! perf report -i "${perfdata}" -q | grep -q "${testsym}" then echo "Per-thread record [Failed missing output]" err=1@@ -62,7 +116,7 @@ test_register_capture() { return fi if ! perf record -o - --intr-regs=di,r8,dx,cx -e br_inst_retired.near_call:p \ - -c 1000 --per-thread true 2> /dev/null \ + -c 1000 --per-thread ${testprog} 2> /dev/null \ | perf script -F ip,sym,iregs -i - 2> /dev/null \ | grep -q "DI:" then@@ -73,6 +127,8 @@ test_register_capture() { echo "Register capture test [Success]" } +build_test_program + test_per_thread test_register_capture --2.38.0.135.g90850a2211-goog
Arnaldo Carvalho de Melo
maintainerRe: [PATCH 3/8] perf test: Use a test program in perf record tests
Em Thu, Oct 20, 2022 at 04:52:14PM -0700, Ian Rogers escreveu:
Quoted 8 lines · Namhyung Kim
Quoted 7 lines
If the system has cc it could build a test program with two threads and then use it for more detailed testing. Also it accepts an option to run a thread forever to ensure multi-thread runs.
If cc is not found, it falls back to use the default value 'true'.
Quoted 7 lines
Quoted 1 line
Signed-off-by Namhyung Kim <[email protected]>Acked-by Ian Rogers <[email protected]>I wonder if these utilities should just be built into perf to avoid the cc dependency. Perhaps we can have a hidden option built into perf test.
Agreed, not depending on a compiler makes 'perf test' usable in more systems, particularly production ones where we may want to check if perf is passing all tests applicable to that system.
- Arnaldo
Quoted 124 lines
Thanks,
IanQuoted 121 lines
--- tools/perf/tests/shell/record.sh | 64 ++++++++++++++++++++++++++++++-- 1 file changed, 60 insertions(+), 4 deletions(-)tools/perf/tests/shell/record.sh +60
-5@@ -6,10 +6,17 @@ set -e err=0 perfdata=$(mktemp /tmp/__perf_test.perf.data.XXXXX) +testprog=$(mktemp /tmp/__perf_test.prog.XXXXXX) +testsym="test_loop" cleanup() { rm -f "${perfdata}" rm -f "${perfdata}".old + + if [ "${testprog}" != "true" ]; then + rm -f "${testprog}" + fi + trap - EXIT TERM INT }@@ -19,9 +26,56 @@ trap_cleanup() { } trap trap_cleanup EXIT TERM INT +build_test_program() { + if ! [ -x "$(command -v cc)" ]; then + # No CC found. Fall back to 'true' + testprog=true + testsym=true + return + fi + + echo "Build a test program" + cat <<EOF | cc -o ${testprog} -xc - -pthread +#include <stdio.h> +#include <stdlib.h> +#include <pthread.h> + +void test_loop(void) { + volatile int count = 1000000; + + while (count--) + continue; +} + +void *thfunc(void *arg) { + int forever = *(int *)arg; + + do { + test_loop(); + } while (forever); + + return NULL; +} + +int main(int argc, char *argv[]) { + pthread_t th; + int forever = 0; + + if (argc > 1) + forever = atoi(argv[1]); + + pthread_create(&th, NULL, thfunc, &forever); + test_loop(); + pthread_join(th, NULL); + + return 0; +} +EOF +} + test_per_thread() { echo "Basic --per-thread mode test" - if ! perf record -o /dev/null --quiet true 2> /dev/null + if ! perf record -o /dev/null --quiet ${testprog} 2> /dev/null then echo "Per-thread record [Skipped event not supported]" if [ $err -ne 1 ]@@ -30,13 +84,13 @@ test_per_thread() { fi return fi - if ! perf record --per-thread -o "${perfdata}" true 2> /dev/null + if ! perf record --per-thread -o "${perfdata}" ${testprog} 2> /dev/null then echo "Per-thread record [Failed record]" err=1 return fi - if ! perf report -i "${perfdata}" -q | grep -q true + if ! perf report -i "${perfdata}" -q | grep -q "${testsym}" then echo "Per-thread record [Failed missing output]" err=1@@ -62,7 +116,7 @@ test_register_capture() { return fi if ! perf record -o - --intr-regs=di,r8,dx,cx -e br_inst_retired.near_call:p \ - -c 1000 --per-thread true 2> /dev/null \ + -c 1000 --per-thread ${testprog} 2> /dev/null \ | perf script -F ip,sym,iregs -i - 2> /dev/null \ | grep -q "DI:" then@@ -73,6 +127,8 @@ test_register_capture() { echo "Register capture test [Success]" } +build_test_program + test_per_thread test_register_capture --2.38.0.135.g90850a2211-goog
Signature · 1 line
- Arnaldo
Adrian Hunter
reviewerRe: [PATCH 3/8] perf test: Use a test program in perf record tests
Quoted 13 lines · Arnaldo Carvalho de Melo
Em Thu, Oct 20, 2022 at 04:52:14PM -0700, Ian Rogers escreveu:
Quoted 12 lines · Namhyung Kim
Quoted 8 lines
If the system has cc it could build a test program with two threads and then use it for more detailed testing. Also it accepts an option to run a thread forever to ensure multi-thread runs.
If cc is not found, it falls back to use the default value 'true'.
Signed-off-by Namhyung Kim <[email protected]>Acked-by Ian Rogers <[email protected]>
Quoted 7 lines
Quoted 3 lines
I wonder if these utilities should just be built into perf to avoid the cc dependency. Perhaps we can have a hidden option built into perf test.
Agreed, not depending on a compiler makes 'perf test' usable in more systems, particularly production ones where we may want to check if perf is passing all tests applicable to that system.
I haven't seen anyone package anything except the perf executable, so I presume the only people running these tests install the source, and so need a compiler anyway.
Arnaldo Carvalho de Melo
maintainerRe: [PATCH 3/8] perf test: Use a test program in perf record tests
Em Mon, Oct 24, 2022 at 05:00:14PM +0300, Adrian Hunter escreveu:
Quoted 6 lines · Arnaldo Carvalho de Melo
Quoted 4 lines
Em Thu, Oct 20, 2022 at 04:52:14PM -0700, Ian Rogers escreveu:
Quoted 3 lines
I wonder if these utilities should just be built into perf to avoid the cc dependency. Perhaps we can have a hidden option built into perf test.
Quoted 7 lines
Quoted 3 lines
Agreed, not depending on a compiler makes 'perf test' usable in more systems, particularly production ones where we may want to check if perf is passing all tests applicable to that system.
I haven't seen anyone package anything except the perf executable, so I presume the only people running these tests install the source, and so need a compiler anyway.
Humm?
[root@quaco ~]# head -3 /etc/os-release
NAME="Fedora Linux"
VERSION="36 (Workstation Edition)"
ID=fedora
[root@quaco ~]#G
[root@quaco ~]# rpm -qi perf
Name : perf
Version : 5.19.4
Release : 200.fc36
Architecture: x86_64
Install Date: Mon 24 Oct 2022 12:57:34 PM -03
Group : Unspecified
Size : 12663136
License : GPLv2
Signature : RSA/SHA256, Thu 25 Aug 2022 07:16:04 PM -03, Key ID 999f7cbf38ab71f4
Source RPM : kernel-tools-5.19.4-200.fc36.src.rpm
Build Date : Thu 25 Aug 2022 06:30:42 PM -03
Build Host : bkernel02.iad2.fedoraproject.org
Packager : Fedora Project
Vendor : Fedora Project
URL : http://www.kernel.org/
Bug URL : https://bugz.fedoraproject.org/kernel-tools
Summary : Performance monitoring for the Linux kernel
Description :
This package contains the perf tool, which enables performance monitoring of the Linux kernel.
[root@quaco ~]#
[root@quaco ~]# rpm -ql perf/etc/bash_completion.d/perf/usr/bin/perf/usr/lib/.build-id/usr/lib/.build-id/0c /usr/lib/.build-id/0c/54d587cab1b533b9ab777717ddd256ff84f241 /usr/lib/.build-id/15 /usr/lib/.build-id/15/6552556be0d4ca8cffb8e002d8a67add55aed5 /usr/lib/.build-id/15/c4dfaa3861f530ef60def5506385d4b0b8924f /usr/lib/.build-id/1f /usr/lib/.build-id/1f/799c084e326516a24a647cfd5f484cd054e0a2 /usr/lib/.build-id/23 /usr/lib/.build-id/23/311fe148dd3f3d487ca5ace54ff2faa72ea8da /usr/lib/.build-id/37 /usr/lib/.build-id/37/c9ee70f321fae7fc1fc00fca91514c63f9f052 /usr/lib/.build-id/54 /usr/lib/.build-id/54/43b6bc332ad188b907fdde69ac02c9a69d158f /usr/lib/.build-id/7e /usr/lib/.build-id/7e/55e7a8e4df03137ec23b4135e205bc8eb77e05 /usr/lib/.build-id/84 /usr/lib/.build-id/84/3c3863a123d129c98fa7cb99a58d620c9e5edc /usr/lib/.build-id/86 /usr/lib/.build-id/86/d063d2f9833a447d96504e5f9e472048d12c49 /usr/lib/.build-id/98 /usr/lib/.build-id/98/a28589fde6ce95d9f3d305f7c9853333a9415b /usr/lib/.build-id/9e /usr/lib/.build-id/9e/567f55ca9c2b4867eb95d4182da07d4843976d /usr/lib/.build-id/b6 /usr/lib/.build-id/b6/f8c9a6d6e1990f8577a9fbac875f812a316a20 /usr/lib/.build-id/c0 /usr/lib/.build-id/c0/8b255f02ce8f3e90df2b6156d2d38b1221995d /usr/lib/.build-id/d5/e964c3413f52492e6355852d25b2ab23b03e38 /usr/lib/.build-id/e8 /usr/lib/.build-id/e8/2de614bf77f501a4511cb70151b98392669c77 /usr/lib64/libperf-jvmti.so/usr/libexec/perf-core/usr/libexec/perf-core/dlfilters /usr/libexec/perf-core/dlfilters/dlfilter-show-cycles.so /usr/libexec/perf-core/dlfilters/dlfilter-test-api-v0.so /usr/libexec/perf-core/perf-archive /usr/libexec/perf-core/perf-iostat/usr/libexec/perf-core/scripts/usr/libexec/perf-core/scripts/perl /usr/libexec/perf-core/scripts/perl/Perf-Trace-Util /usr/libexec/perf-core/scripts/perl/Perf-Trace-Util/lib /usr/libexec/perf-core/scripts/perl/Perf-Trace-Util/lib/Perf /usr/libexec/perf-core/scripts/perl/Perf-Trace-Util/lib/Perf/Trace /usr/libexec/perf-core/scripts/perl/Perf-Trace-Util/lib/Perf/Trace/Context.pm /usr/libexec/perf-core/scripts/perl/Perf-Trace-Util/lib/Perf/Trace/Core.pm /usr/libexec/perf-core/scripts/perl/Perf-Trace-Util/lib/Perf/Trace/Util.pm /usr/libexec/perf-core/scripts/perl/bin /usr/libexec/perf-core/scripts/perl/bin/check-perf-trace-record /usr/libexec/perf-core/scripts/perl/bin/failed-syscalls-record /usr/libexec/perf-core/scripts/perl/bin/failed-syscalls-report /usr/libexec/perf-core/scripts/perl/bin/rw-by-file-record /usr/libexec/perf-core/scripts/perl/bin/rw-by-file-report /usr/libexec/perf-core/scripts/perl/bin/rw-by-pid-record /usr/libexec/perf-core/scripts/perl/bin/rw-by-pid-report /usr/libexec/perf-core/scripts/perl/bin/rwtop-record /usr/libexec/perf-core/scripts/perl/bin/rwtop-report /usr/libexec/perf-core/scripts/perl/bin/wakeup-latency-record /usr/libexec/perf-core/scripts/perl/bin/wakeup-latency-report /usr/libexec/perf-core/scripts/perl/check-perf-trace.pl /usr/libexec/perf-core/scripts/perl/failed-syscalls.pl /usr/libexec/perf-core/scripts/perl/rw-by-file.pl /usr/libexec/perf-core/scripts/perl/rw-by-pid.pl /usr/libexec/perf-core/scripts/perl/rwtop.pl /usr/libexec/perf-core/scripts/perl/wakeup-latency.pl /usr/libexec/perf-core/scripts/python /usr/libexec/perf-core/scripts/python/Perf-Trace-Util /usr/libexec/perf-core/scripts/python/Perf-Trace-Util/lib /usr/libexec/perf-core/scripts/python/Perf-Trace-Util/lib/Perf /usr/libexec/perf-core/scripts/python/Perf-Trace-Util/lib/Perf/Trace /usr/libexec/perf-core/scripts/python/Perf-Trace-Util/lib/Perf/Trace/Core.py /usr/libexec/perf-core/scripts/python/Perf-Trace-Util/lib/Perf/Trace/EventClass.py /usr/libexec/perf-core/scripts/python/Perf-Trace-Util/lib/Perf/Trace/SchedGui.py /usr/libexec/perf-core/scripts/python/Perf-Trace-Util/lib/Perf/Trace/Util.py /usr/libexec/perf-core/scripts/python/arm-cs-trace-disasm.py /usr/libexec/perf-core/scripts/python/bin /usr/libexec/perf-core/scripts/python/bin/compaction-times-record /usr/libexec/perf-core/scripts/python/bin/compaction-times-report /usr/libexec/perf-core/scripts/python/bin/event_analyzing_sample-record /usr/libexec/perf-core/scripts/python/bin/event_analyzing_sample-report /usr/libexec/perf-core/scripts/python/bin/export-to-postgresql-record /usr/libexec/perf-core/scripts/python/bin/export-to-postgresql-report /usr/libexec/perf-core/scripts/python/bin/export-to-sqlite-record /usr/libexec/perf-core/scripts/python/bin/export-to-sqlite-report /usr/libexec/perf-core/scripts/python/bin/failed-syscalls-by-pid-record /usr/libexec/perf-core/scripts/python/bin/failed-syscalls-by-pid-report /usr/libexec/perf-core/scripts/python/bin/flamegraph-record /usr/libexec/perf-core/scripts/python/bin/flamegraph-report /usr/libexec/perf-core/scripts/python/bin/futex-contention-record /usr/libexec/perf-core/scripts/python/bin/futex-contention-report /usr/libexec/perf-core/scripts/python/bin/intel-pt-events-record /usr/libexec/perf-core/scripts/python/bin/intel-pt-events-report /usr/libexec/perf-core/scripts/python/bin/mem-phys-addr-record /usr/libexec/perf-core/scripts/python/bin/mem-phys-addr-report /usr/libexec/perf-core/scripts/python/bin/net_dropmonitor-record /usr/libexec/perf-core/scripts/python/bin/net_dropmonitor-report /usr/libexec/perf-core/scripts/python/bin/netdev-times-record /usr/libexec/perf-core/scripts/python/bin/netdev-times-report /usr/libexec/perf-core/scripts/python/bin/powerpc-hcalls-record /usr/libexec/perf-core/scripts/python/bin/powerpc-hcalls-report /usr/libexec/perf-core/scripts/python/bin/sched-migration-record /usr/libexec/perf-core/scripts/python/bin/sched-migration-report /usr/libexec/perf-core/scripts/python/bin/sctop-record /usr/libexec/perf-core/scripts/python/bin/sctop-report /usr/libexec/perf-core/scripts/python/bin/stackcollapse-record /usr/libexec/perf-core/scripts/python/bin/stackcollapse-report /usr/libexec/perf-core/scripts/python/bin/syscall-counts-by-pid-record /usr/libexec/perf-core/scripts/python/bin/syscall-counts-by-pid-report /usr/libexec/perf-core/scripts/python/bin/syscall-counts-record /usr/libexec/perf-core/scripts/python/bin/syscall-counts-report /usr/libexec/perf-core/scripts/python/check-perf-trace.py /usr/libexec/perf-core/scripts/python/compaction-times.py /usr/libexec/perf-core/scripts/python/event_analyzing_sample.py /usr/libexec/perf-core/scripts/python/export-to-postgresql.py /usr/libexec/perf-core/scripts/python/export-to-sqlite.py /usr/libexec/perf-core/scripts/python/exported-sql-viewer.py /usr/libexec/perf-core/scripts/python/failed-syscalls-by-pid.py /usr/libexec/perf-core/scripts/python/flamegraph.py /usr/libexec/perf-core/scripts/python/futex-contention.py /usr/libexec/perf-core/scripts/python/intel-pt-events.py /usr/libexec/perf-core/scripts/python/libxed.py /usr/libexec/perf-core/scripts/python/mem-phys-addr.py /usr/libexec/perf-core/scripts/python/net_dropmonitor.py /usr/libexec/perf-core/scripts/python/netdev-times.py /usr/libexec/perf-core/scripts/python/powerpc-hcalls.py /usr/libexec/perf-core/scripts/python/sched-migration.py /usr/libexec/perf-core/scripts/python/sctop.py /usr/libexec/perf-core/scripts/python/stackcollapse.py /usr/libexec/perf-core/scripts/python/stat-cpi.py /usr/libexec/perf-core/scripts/python/syscall-counts-by-pid.py /usr/libexec/perf-core/scripts/python/syscall-counts.py /usr/libexec/perf-core/tests/usr/libexec/perf-core/tests/attr /usr/libexec/perf-core/tests/attr.py /usr/libexec/perf-core/tests/attr/README /usr/libexec/perf-core/tests/attr/base-record /usr/libexec/perf-core/tests/attr/base-record-spe /usr/libexec/perf-core/tests/attr/base-stat /usr/libexec/perf-core/tests/attr/system-wide-dummy /usr/libexec/perf-core/tests/attr/test-record-C0 /usr/libexec/perf-core/tests/attr/test-record-basic /usr/libexec/perf-core/tests/attr/test-record-branch-any /usr/libexec/perf-core/tests/attr/test-record-branch-filter-any /usr/libexec/perf-core/tests/attr/test-record-branch-filter-any_call /usr/libexec/perf-core/tests/attr/test-record-branch-filter-any_ret /usr/libexec/perf-core/tests/attr/test-record-branch-filter-hv /usr/libexec/perf-core/tests/attr/test-record-branch-filter-ind_call /usr/libexec/perf-core/tests/attr/test-record-branch-filter-k /usr/libexec/perf-core/tests/attr/test-record-branch-filter-u /usr/libexec/perf-core/tests/attr/test-record-count /usr/libexec/perf-core/tests/attr/test-record-data /usr/libexec/perf-core/tests/attr/test-record-freq /usr/libexec/perf-core/tests/attr/test-record-graph-default /usr/libexec/perf-core/tests/attr/test-record-graph-default-aarch64 /usr/libexec/perf-core/tests/attr/test-record-graph-dwarf /usr/libexec/perf-core/tests/attr/test-record-graph-fp /usr/libexec/perf-core/tests/attr/test-record-graph-fp-aarch64 /usr/libexec/perf-core/tests/attr/test-record-group /usr/libexec/perf-core/tests/attr/test-record-group-sampling /usr/libexec/perf-core/tests/attr/test-record-group1 /usr/libexec/perf-core/tests/attr/test-record-group2 /usr/libexec/perf-core/tests/attr/test-record-no-buffering /usr/libexec/perf-core/tests/attr/test-record-no-inherit /usr/libexec/perf-core/tests/attr/test-record-no-samples /usr/libexec/perf-core/tests/attr/test-record-period /usr/libexec/perf-core/tests/attr/test-record-pfm-period /usr/libexec/perf-core/tests/attr/test-record-raw /usr/libexec/perf-core/tests/attr/test-record-spe-period /usr/libexec/perf-core/tests/attr/test-record-spe-period-term /usr/libexec/perf-core/tests/attr/test-record-spe-physical-address /usr/libexec/perf-core/tests/attr/test-stat-C0 /usr/libexec/perf-core/tests/attr/test-stat-basic /usr/libexec/perf-core/tests/attr/test-stat-default /usr/libexec/perf-core/tests/attr/test-stat-detailed-1 /usr/libexec/perf-core/tests/attr/test-stat-detailed-2 /usr/libexec/perf-core/tests/attr/test-stat-detailed-3 /usr/libexec/perf-core/tests/attr/test-stat-group /usr/libexec/perf-core/tests/attr/test-stat-group1 /usr/libexec/perf-core/tests/attr/test-stat-no-inherit /usr/libexec/perf-core/tests/pe-file.exe /usr/libexec/perf-core/tests/pe-file.exe.debug /usr/libexec/perf-core/tests/shell /usr/libexec/perf-core/tests/shell/buildid.sh /usr/libexec/perf-core/tests/shell/daemon.sh /usr/libexec/perf-core/tests/shell/lib /usr/libexec/perf-core/tests/shell/lib/probe.sh /usr/libexec/perf-core/tests/shell/lib/probe_vfs_getname.sh /usr/libexec/perf-core/tests/shell/pipe_test.sh /usr/libexec/perf-core/tests/shell/probe_vfs_getname.sh /usr/libexec/perf-core/tests/shell/record+probe_libc_inet_pton.sh /usr/libexec/perf-core/tests/shell/record+script_probe_vfs_getname.sh /usr/libexec/perf-core/tests/shell/record+zstd_comp_decomp.sh /usr/libexec/perf-core/tests/shell/record.sh /usr/libexec/perf-core/tests/shell/record_offcpu.sh /usr/libexec/perf-core/tests/shell/stat+csv_output.sh /usr/libexec/perf-core/tests/shell/stat+csv_summary.sh /usr/libexec/perf-core/tests/shell/stat+shadow_stat.sh /usr/libexec/perf-core/tests/shell/stat.sh /usr/libexec/perf-core/tests/shell/stat_all_metricgroups.sh /usr/libexec/perf-core/tests/shell/stat_all_metrics.sh /usr/libexec/perf-core/tests/shell/stat_all_pmu.sh /usr/libexec/perf-core/tests/shell/stat_bpf_counters.sh /usr/libexec/perf-core/tests/shell/test_arm_callgraph_fp.sh /usr/libexec/perf-core/tests/shell/test_arm_coresight.sh /usr/libexec/perf-core/tests/shell/test_arm_spe.sh /usr/libexec/perf-core/tests/shell/test_arm_spe_fork.sh /usr/libexec/perf-core/tests/shell/test_intel_pt.sh /usr/libexec/perf-core/tests/shell/trace+probe_vfs_getname.sh /usr/share/doc/perf/usr/share/doc/perf-tip/tips.txt/usr/share/doc/perf/examples.txt/usr/share/licenses/perf/usr/share/licenses/perf/COPYING /usr/share/man/man1/perf-annotate.1.gz /usr/share/man/man1/perf-archive.1.gz /usr/share/man/man1/perf-arm-spe.1.gz /usr/share/man/man1/perf-bench.1.gz /usr/share/man/man1/perf-buildid-cache.1.gz /usr/share/man/man1/perf-buildid-list.1.gz /usr/share/man/man1/perf-c2c.1.gz /usr/share/man/man1/perf-config.1.gz /usr/share/man/man1/perf-daemon.1.gz /usr/share/man/man1/perf-data.1.gz/usr/share/man/man1/perf-diff.1.gz /usr/share/man/man1/perf-dlfilter.1.gz /usr/share/man/man1/perf-evlist.1.gz /usr/share/man/man1/perf-ftrace.1.gz /usr/share/man/man1/perf-help.1.gz /usr/share/man/man1/perf-inject.1.gz /usr/share/man/man1/perf-intel-pt.1.gz /usr/share/man/man1/perf-iostat.1.gz /usr/share/man/man1/perf-kallsyms.1.gz /usr/share/man/man1/perf-kmem.1.gz/usr/share/man/man1/perf-kvm.1.gz/usr/share/man/man1/perf-list.1.gz/usr/share/man/man1/perf-lock.1.gz/usr/share/man/man1/perf-mem.1.gz/usr/share/man/man1/perf-probe.1.gz /usr/share/man/man1/perf-record.1.gz /usr/share/man/man1/perf-report.1.gz /usr/share/man/man1/perf-sched.1.gz /usr/share/man/man1/perf-script-perl.1.gz /usr/share/man/man1/perf-script-python.1.gz /usr/share/man/man1/perf-script.1.gz /usr/share/man/man1/perf-stat.1.gz/usr/share/man/man1/perf-test.1.gz /usr/share/man/man1/perf-timechart.1.gz /usr/share/man/man1/perf-top.1.gz/usr/share/man/man1/perf-trace.1.gz /usr/share/man/man1/perf-version.1.gz/usr/share/man/man1/perf.1.gz/usr/share/perf-core/usr/share/perf-core/strace/usr/share/perf-core/strace/groups /usr/share/perf-core/strace/groups/file /usr/share/perf-core/strace/groups/string
Namhyung Kim
maintainerRe: [PATCH 3/8] perf test: Use a test program in perf record tests
Quoted 73 lines · Arnaldo Carvalho de Melo
Em Mon, Oct 24, 2022 at 05:00:14PM +0300, Adrian Hunter escreveu:
Quoted 6 lines · Arnaldo Carvalho de Melo
Quoted 4 lines
Em Thu, Oct 20, 2022 at 04:52:14PM -0700, Ian Rogers escreveu:
Quoted 3 lines
I wonder if these utilities should just be built into perf to avoid the cc dependency. Perhaps we can have a hidden option built into perf test.
Quoted 7 lines
Quoted 3 lines
Agreed, not depending on a compiler makes 'perf test' usable in more systems, particularly production ones where we may want to check if perf is passing all tests applicable to that system.
I haven't seen anyone package anything except the perf executable, so I presume the only people running these tests install the source, and so need a compiler anyway.
Humm?
[root@quaco ~]# head -3
/etc/os-release
NAME="Fedora Linux"
VERSION="36 (Workstation Edition)"
ID=fedora
[root@quaco ~]#G[root@quaco ~]# rpm -qi perf
Name : perf
Version : 5.19.4
Release : 200.fc36
Architecture:x86_64
Install Date: Mon 24 Oct 2022 12:57:34 PM -03
Group : Unspecified
Size : 12663136
License : GPLv2
Signature : RSA/SHA256, Thu 25 Aug 2022 07:16:04 PM -03, Key ID999f7cbf38ab71f4
Source RPM : kernel-tools-5.19.4-200.fc36.src.rpm
Build Date : Thu 25 Aug 2022 06:30:42 PM -03
Build Host : bkernel02.iad2.fedoraproject.org
Packager : Fedora Project
Vendor : Fedora Project
URL : http://www.kernel.org/
Bug URL : https://bugz.fedoraproject.org/kernel-tools
Summary : Performance monitoring for the Linux kernel
Description :
This package contains the perf tool, which enables performance monitoring of the Linux kernel.
[root@quaco ~]#[root@quaco ~]# rpm -ql perf
/etc/bash_completion.d/perf/usr/bin/perf/usr/lib/.build-id/usr/lib/.build-id/0c/usr/lib/.build-id/0c/54d587cab1b533b9ab777717ddd256ff84f241/usr/lib/.build-id/15/usr/lib/.build-id/15/6552556be0d4ca8cffb8e002d8a67add55aed5/usr/lib/.build-id/15/c4dfaa3861f530ef60def5506385d4b0b8924f/usr/lib/.build-id/1f/usr/lib/.build-id/1f/799c084e326516a24a647cfd5f484cd054e0a2/usr/lib/.build-id/23/usr/lib/.build-id/23/311fe148dd3f3d487ca5ace54ff2faa72ea8da/usr/lib/.build-id/37/usr/lib/.build-id/37/c9ee70f321fae7fc1fc00fca91514c63f9f052/usr/lib/.build-id/54/usr/lib/.build-id/54/43b6bc332ad188b907fdde69ac02c9a69d158f/usr/lib/.build-id/7e/usr/lib/.build-id/7e/55e7a8e4df03137ec23b4135e205bc8eb77e05/usr/lib/.build-id/84/usr/lib/.build-id/84/3c3863a123d129c98fa7cb99a58d620c9e5edc/usr/lib/.build-id/86/usr/lib/.build-id/86/d063d2f9833a447d96504e5f9e472048d12c49/usr/lib/.build-id/98/usr/lib/.build-id/98/a28589fde6ce95d9f3d305f7c9853333a9415b/usr/lib/.build-id/9e
Quoted 8 lines
/usr/lib/.build-id/9e/567f55ca9c2b4867eb95d4182da07d4843976d/usr/lib/.build-id/b6/usr/lib/.build-id/b6/f8c9a6d6e1990f8577a9fbac875f812a316a20/usr/lib/.build-id/c0/usr/lib/.build-id/c0/8b255f02ce8f3e90df2b6156d2d38b1221995d/usr/lib/.build-id/d5/e964c3413f52492e6355852d25b2ab23b03e38/usr/lib/.build-id/e8/usr/lib/.build-id/e8/2de614bf77f501a4511cb70151b98392669c77
Interesting, it contains some build-ids for system libraries?
Thanks,
Namhyung
Arnaldo Carvalho de Melo
maintainerRe: [PATCH 3/8] perf test: Use a test program in perf record tests
Em Mon, Oct 24, 2022 at 09:12:21PM -0700, Namhyung Kim escreveu:
Quoted 83 lines · Arnaldo Carvalho de Melo
Quoted 73 lines
Em Mon, Oct 24, 2022 at 05:00:14PM +0300, Adrian Hunter escreveu:
Quoted 6 lines · Arnaldo Carvalho de Melo
Quoted 4 lines
Em Thu, Oct 20, 2022 at 04:52:14PM -0700, Ian Rogers escreveu:
Quoted 3 lines
I wonder if these utilities should just be built into perf to avoid the cc dependency. Perhaps we can have a hidden option built into perf test.
Quoted 7 lines
Quoted 3 lines
Agreed, not depending on a compiler makes 'perf test' usable in more systems, particularly production ones where we may want to check if perf is passing all tests applicable to that system.
I haven't seen anyone package anything except the perf executable, so I presume the only people running these tests install the source, and so need a compiler anyway.
Humm?
[root@quaco ~]# head -3
/etc/os-release
NAME="Fedora Linux"
VERSION="36 (Workstation Edition)"
ID=fedora
[root@quaco ~]#G[root@quaco ~]# rpm -qi perf
Name : perf
Version : 5.19.4
Release : 200.fc36
Architecture:x86_64
Install Date: Mon 24 Oct 2022 12:57:34 PM -03
Group : Unspecified
Size : 12663136
License : GPLv2
Signature : RSA/SHA256, Thu 25 Aug 2022 07:16:04 PM -03, Key ID999f7cbf38ab71f4
Source RPM : kernel-tools-5.19.4-200.fc36.src.rpm
Build Date : Thu 25 Aug 2022 06:30:42 PM -03
Build Host : bkernel02.iad2.fedoraproject.org
Packager : Fedora Project
Vendor : Fedora Project
URL : http://www.kernel.org/
Bug URL : https://bugz.fedoraproject.org/kernel-tools
Summary : Performance monitoring for the Linux kernel
Description :
This package contains the perf tool, which enables performance monitoring of the Linux kernel.
[root@quaco ~]#[root@quaco ~]# rpm -ql perf
/etc/bash_completion.d/perf/usr/bin/perf/usr/lib/.build-id/usr/lib/.build-id/0c/usr/lib/.build-id/0c/54d587cab1b533b9ab777717ddd256ff84f241/usr/lib/.build-id/15/usr/lib/.build-id/15/6552556be0d4ca8cffb8e002d8a67add55aed5/usr/lib/.build-id/15/c4dfaa3861f530ef60def5506385d4b0b8924f/usr/lib/.build-id/1f/usr/lib/.build-id/1f/799c084e326516a24a647cfd5f484cd054e0a2/usr/lib/.build-id/23/usr/lib/.build-id/23/311fe148dd3f3d487ca5ace54ff2faa72ea8da/usr/lib/.build-id/37/usr/lib/.build-id/37/c9ee70f321fae7fc1fc00fca91514c63f9f052/usr/lib/.build-id/54/usr/lib/.build-id/54/43b6bc332ad188b907fdde69ac02c9a69d158f/usr/lib/.build-id/7e/usr/lib/.build-id/7e/55e7a8e4df03137ec23b4135e205bc8eb77e05/usr/lib/.build-id/84/usr/lib/.build-id/84/3c3863a123d129c98fa7cb99a58d620c9e5edc/usr/lib/.build-id/86/usr/lib/.build-id/86/d063d2f9833a447d96504e5f9e472048d12c49/usr/lib/.build-id/98/usr/lib/.build-id/98/a28589fde6ce95d9f3d305f7c9853333a9415b/usr/lib/.build-id/9eQuoted 8 lines
/usr/lib/.build-id/9e/567f55ca9c2b4867eb95d4182da07d4843976d/usr/lib/.build-id/b6/usr/lib/.build-id/b6/f8c9a6d6e1990f8577a9fbac875f812a316a20/usr/lib/.build-id/c0/usr/lib/.build-id/c0/8b255f02ce8f3e90df2b6156d2d38b1221995d/usr/lib/.build-id/d5/e964c3413f52492e6355852d25b2ab23b03e38/usr/lib/.build-id/e8/usr/lib/.build-id/e8/2de614bf77f501a4511cb70151b98392669c77
Quoted 1 line
Interesting, it contains some build-ids for system libraries?
[acme@quaco ~]$ ls -la /usr/lib/.build-id/e8/2de614bf77f501a4511cb70151b98392669c77
lrwxrwxrwx. 1 root root 67 Aug 25 18:33 /usr/lib/.build-id/e8/2de614bf77f501a4511cb70151b98392669c77 -> ../../../../usr/libexec/perf-core/dlfilters/dlfilter-test-api-v0.so [acme@quaco ~]$
Yeah, strange:
[acme@quaco ~]$ rpm -ql perf | grep build-id/../ /usr/lib/.build-id/0c/54d587cab1b533b9ab777717ddd256ff84f241 /usr/lib/.build-id/15/6552556be0d4ca8cffb8e002d8a67add55aed5 /usr/lib/.build-id/15/c4dfaa3861f530ef60def5506385d4b0b8924f /usr/lib/.build-id/1f/799c084e326516a24a647cfd5f484cd054e0a2 /usr/lib/.build-id/23/311fe148dd3f3d487ca5ace54ff2faa72ea8da /usr/lib/.build-id/37/c9ee70f321fae7fc1fc00fca91514c63f9f052 /usr/lib/.build-id/54/43b6bc332ad188b907fdde69ac02c9a69d158f /usr/lib/.build-id/7e/55e7a8e4df03137ec23b4135e205bc8eb77e05 /usr/lib/.build-id/84/3c3863a123d129c98fa7cb99a58d620c9e5edc /usr/lib/.build-id/86/d063d2f9833a447d96504e5f9e472048d12c49 /usr/lib/.build-id/98/a28589fde6ce95d9f3d305f7c9853333a9415b /usr/lib/.build-id/9e/567f55ca9c2b4867eb95d4182da07d4843976d /usr/lib/.build-id/b6/f8c9a6d6e1990f8577a9fbac875f812a316a20 /usr/lib/.build-id/c0/8b255f02ce8f3e90df2b6156d2d38b1221995d /usr/lib/.build-id/d5/e964c3413f52492e6355852d25b2ab23b03e38 /usr/lib/.build-id/e8/2de614bf77f501a4511cb70151b98392669c77 [acme@quaco ~]$
[acme@quaco ~]$ rpm -ql perf | grep build-id/../ | xargs realpath /usr/lib64/traceevent/plugins/plugin_sched_switch.so /usr/lib64/traceevent/plugins/plugin_kmem.so /usr/lib64/traceevent/plugins/plugin_tlb.so /usr/lib64/traceevent/plugins/plugin_jbd2.so /usr/lib64/traceevent/plugins/plugin_hrtimer.so /usr/lib64/traceevent/plugins/plugin_mac80211.so /usr/lib64/libperf-jvmti.so /usr/lib64/traceevent/plugins/plugin_cfg80211.so /usr/lib64/traceevent/plugins/plugin_xen.so /usr/lib64/traceevent/plugins/plugin_kvm.so /usr/libexec/perf-core/dlfilters/dlfilter-show-cycles.so /usr/lib64/traceevent/plugins/plugin_scsi.so /usr/lib64/traceevent/plugins/plugin_futex.so/usr/bin/perf/usr/lib64/traceevent/plugins/plugin_function.so /usr/libexec/perf-core/dlfilters/dlfilter-test-api-v0.so
[acme@quaco ~]$ rpm -ql perf | grep build-id/../ | xargs realpath | xargs rpm -qf libtraceevent-1.5.3-2.fc36.x86_64
libtraceevent-1.5.3-2.fc36.x86_64
libtraceevent-1.5.3-2.fc36.x86_64
libtraceevent-1.5.3-2.fc36.x86_64
libtraceevent-1.5.3-2.fc36.x86_64
libtraceevent-1.5.3-2.fc36.x86_64
perf-5.19.4-200.fc36.x86_64
libtraceevent-1.5.3-2.fc36.x86_64
libtraceevent-1.5.3-2.fc36.x86_64
libtraceevent-1.5.3-2.fc36.x86_64
perf-5.19.4-200.fc36.x86_64
libtraceevent-1.5.3-2.fc36.x86_64
libtraceevent-1.5.3-2.fc36.x86_64
perf-5.19.4-200.fc36.x86_64
libtraceevent-1.5.3-2.fc36.x86_64
perf-5.19.4-200.fc36.x86_64
[acme@quaco ~]$
Humm, but then the build-ids are different...
[acme@quaco ~]$ rpm -ql libtraceevent | grep build-id/../ /usr/lib/.build-id/00/59bd37dc1c345eb2f98cf2bc1163bc1fd4b63e /usr/lib/.build-id/00/b22aa8cc2b81e4cb383d8b4dc14cbeb958fda6 /usr/lib/.build-id/11/6a8d17c89795d2269d3d65906ea7c7c5bcb856 /usr/lib/.build-id/24/8bfb4534954eb71ab3e5f7887507cb8981f921 /usr/lib/.build-id/57/0768e8ea5155cf2759c4f5dca3736a73c3dfad /usr/lib/.build-id/61/2744cd9cb2d1b0c569b0c86578d11be7227f27 /usr/lib/.build-id/9d/630834441f4e382ddcecef810b100737bb56e4 /usr/lib/.build-id/b0/57189b558351ace427ae0a650c248deca4badd /usr/lib/.build-id/b1/a0003b83e307cf61d7c8a2751e5c78e6ccfc87 /usr/lib/.build-id/b4/b8a30ebf86cb849bcb0cabc4d452c1a44487cc /usr/lib/.build-id/ce/5e0647f756ec2c30b2e6df9a5a98e552871cb6 /usr/lib/.build-id/e0/a6a6207f56674ffacd7b063fd2acd4cec3f661 /usr/lib/.build-id/f4/86fb5b3072945405d0590045715d18adc7aa18
[acme@quaco ~]$
[acme@quaco ~]$ rpm -ql libtraceevent | grep build-id/../ | xargs realpath /usr/lib64/traceevent/plugins/plugin_function.so /usr/lib64/traceevent/plugins/plugin_scsi.so /usr/lib64/traceevent/plugins/plugin_futex.so /usr/lib64/traceevent/plugins/plugin_cfg80211.so /usr/lib64/traceevent/plugins/plugin_hrtimer.so /usr/lib64/traceevent/plugins/plugin_mac80211.so /usr/lib64/traceevent/plugins/plugin_sched_switch.so /usr/lib64/traceevent/plugins/plugin_jbd2.so /usr/lib64/traceevent/plugins/plugin_tlb.so /usr/lib64/traceevent/plugins/plugin_xen.so /usr/lib64/traceevent/plugins/plugin_kvm.so /usr/lib64/traceevent/plugins/plugin_kmem.so /usr/lib64/libtraceevent.so.1.5.3
[acme@quaco ~]$ rpm -ql libtraceevent | grep build-id/../ | xargs realpath | xargs rpm -qf libtraceevent-1.5.3-2.fc36.x86_64
libtraceevent-1.5.3-2.fc36.x86_64
libtraceevent-1.5.3-2.fc36.x86_64
libtraceevent-1.5.3-2.fc36.x86_64
libtraceevent-1.5.3-2.fc36.x86_64
libtraceevent-1.5.3-2.fc36.x86_64
libtraceevent-1.5.3-2.fc36.x86_64
libtraceevent-1.5.3-2.fc36.x86_64
libtraceevent-1.5.3-2.fc36.x86_64
libtraceevent-1.5.3-2.fc36.x86_64
libtraceevent-1.5.3-2.fc36.x86_64
libtraceevent-1.5.3-2.fc36.x86_64
libtraceevent-1.5.3-2.fc36.x86_64
[acme@quaco ~]$
I.e. libtraceevent has the plugins and the build-id symlinks, perf has just the symlinks:
[acme@quaco ~]$ rpm -ql libtraceevent | grep plugin_.*so /usr/lib64/traceevent/plugins/plugin_cfg80211.so /usr/lib64/traceevent/plugins/plugin_function.so /usr/lib64/traceevent/plugins/plugin_futex.so /usr/lib64/traceevent/plugins/plugin_hrtimer.so /usr/lib64/traceevent/plugins/plugin_jbd2.so /usr/lib64/traceevent/plugins/plugin_kmem.so /usr/lib64/traceevent/plugins/plugin_kvm.so /usr/lib64/traceevent/plugins/plugin_mac80211.so /usr/lib64/traceevent/plugins/plugin_sched_switch.so /usr/lib64/traceevent/plugins/plugin_scsi.so /usr/lib64/traceevent/plugins/plugin_tlb.so /usr/lib64/traceevent/plugins/plugin_xen.so [acme@quaco ~]$ rpm -ql perf | grep plugin_.*so [acme@quaco ~]$
And perf is being built with the libtraceevent that is in fedora:
[acme@quaco ~]$ ldd which perf | grep traceevent
libtraceevent.so.1 => /lib64/libtraceevent.so.1 (0x00007f43496dc000)
[acme@quaco ~]$ rpm -qf /lib64/libtraceevent.so.1
libtraceevent-1.5.3-2.fc36.x86_64
[acme@quaco ~]$ rpm -qf /usr/bin/perf
perf-5.19.4-200.fc36.x86_64
[acme@quaco ~]$
[acme@quaco ~]$ file /usr/bin/perf/usr/bin/perf: ELF 64-bit LSB pie executable, x86-64, version 1 (SYSV), dynamically linked, interpreter /lib64/ld-linux-x86-64.so.2, BuildID[sha1]=c08b255f02ce8f3e90df2b6156d2d38b1221995d, for GNU/Linux 3.2.0, stripped [acme@quaco ~]$
So it seems its just a matter of removing those /usr/lib/.build-id/ symlinks to the libtraceevent plugins from the fedora perf rpm.
Justin?
- Arnaldo
Adrian Hunter
reviewerRe: [PATCH 3/8] perf test: Use a test program in perf record tests
In reply to Arnaldo Carvalho de MeloQuoted 15 lines · Arnaldo Carvalho de Melo
Em Mon, Oct 24, 2022 at 05:00:14PM +0300, Adrian Hunter escreveu:
Quoted 6 lines · Arnaldo Carvalho de Melo
Quoted 4 lines
Em Thu, Oct 20, 2022 at 04:52:14PM -0700, Ian Rogers escreveu:
Quoted 3 lines
I wonder if these utilities should just be built into perf to avoid the cc dependency. Perhaps we can have a hidden option built into perf test.
Quoted 7 lines
Quoted 3 lines
Agreed, not depending on a compiler makes 'perf test' usable in more systems, particularly production ones where we may want to check if perf is passing all tests applicable to that system.
I haven't seen anyone package anything except the perf executable, so I presume the only people running these tests install the source, and so need a compiler anyway.
Quoted 7 lines
Humm?
[root@quaco ~]# head -3
/etc/os-release
NAME="Fedora Linux"
VERSION="36 (Workstation Edition)"
ID=fedora
[root@quaco ~]#G
I guess I got confused - it seems it is only Ubuntu that does that.
Namhyung Kim
maintainerRe: [PATCH 3/8] perf test: Use a test program in perf record tests
In reply to Arnaldo Carvalho de MeloQuoted 14 lines · Arnaldo Carvalho de Melo
Em Thu, Oct 20, 2022 at 04:52:14PM -0700, Ian Rogers escreveu:
Quoted 12 lines · Namhyung Kim
Quoted 8 lines
If the system has cc it could build a test program with two threads and then use it for more detailed testing. Also it accepts an option to run a thread forever to ensure multi-thread runs.
If cc is not found, it falls back to use the default value 'true'.
Signed-off-by Namhyung Kim <[email protected]>Acked-by Ian Rogers <[email protected]>
Quoted 7 lines
Quoted 3 lines
I wonder if these utilities should just be built into perf to avoid the cc dependency. Perhaps we can have a hidden option built into perf test.
Agreed, not depending on a compiler makes 'perf test' usable in more systems, particularly production ones where we may want to check if perf is passing all tests applicable to that system.
Good idea.
I think we can add an option (like -w/--workload) to run the specified workload rather than executing the tests. Like below (assuming we have a workload called 'noploop'). Thoughts?
$ perf stat -- perf test -w noploop
Thanks,
Namhyung
Re: [PATCH 3/8] perf test: Use a test program in perf record tests
In reply to Namhyung KimQuoted 5 lines · Namhyung Kim
If the system has cc it could build a test program with two threads and then use it for more detailed testing. Also it accepts an option to run a thread forever to ensure multi-thread runs.
If cc is not found, it falls back to use the default value 'true'.
Quoted 120 lines
Signed-off-by Namhyung Kim <[email protected]>--- tools/perf/tests/shell/record.sh | 64 ++++++++++++++++++++++++++++++-- 1 file changed, 60 insertions(+), 4 deletions(-)tools/perf/tests/shell/record.sh +60
-4@@ -6,10 +6,17 @@ set -e err=0 perfdata=$(mktemp /tmp/__perf_test.perf.data.XXXXX) +testprog=$(mktemp /tmp/__perf_test.prog.XXXXXX) +testsym="test_loop" cleanup() { rm -f "${perfdata}" rm -f "${perfdata}".old + + if [ "${testprog}" != "true" ]; then + rm -f "${testprog}" + fi + trap - EXIT TERM INT }@@ -19,9 +26,56 @@ trap_cleanup() { } trap trap_cleanup EXIT TERM INT +build_test_program() { + if ! [ -x "$(command -v cc)" ]; then + # No CC found. Fall back to 'true' + testprog=true + testsym=true + return + fi + + echo "Build a test program" + cat <<EOF | cc -o ${testprog} -xc - -pthread +#include <stdio.h> +#include <stdlib.h> +#include <pthread.h> + +void test_loop(void) { + volatile int count = 1000000; + + while (count--) + continue; +} + +void *thfunc(void *arg) { + int forever = *(int *)arg; + + do { + test_loop(); + } while (forever); + + return NULL; +} + +int main(int argc, char *argv[]) { + pthread_t th; + int forever = 0; + + if (argc > 1) + forever = atoi(argv[1]); + + pthread_create(&th, NULL, thfunc, &forever); + test_loop(); + pthread_join(th, NULL); + + return 0; +} +EOF +} + test_per_thread() { echo "Basic --per-thread mode test" - if ! perf record -o /dev/null --quiet true 2> /dev/null + if ! perf record -o /dev/null --quiet ${testprog} 2> /dev/null then echo "Per-thread record [Skipped event not supported]" if [ $err -ne 1 ]@@ -30,13 +84,13 @@ test_per_thread() { fi return fi - if ! perf record --per-thread -o "${perfdata}" true 2> /dev/null + if ! perf record --per-thread -o "${perfdata}" ${testprog} 2> /dev/null then echo "Per-thread record [Failed record]" err=1 return fi - if ! perf report -i "${perfdata}" -q | grep -q true + if ! perf report -i "${perfdata}" -q | grep -q "${testsym}" then echo "Per-thread record [Failed missing output]" err=1@@ -62,7 +116,7 @@ test_register_capture() { return fi if ! perf record -o - --intr-regs=di,r8,dx,cx -e br_inst_retired.near_call:p \ - -c 1000 --per-thread true 2> /dev/null \ + -c 1000 --per-thread ${testprog} 2> /dev/null \ | perf script -F ip,sym,iregs -i - 2> /dev/null \ | grep -q "DI:" then@@ -73,6 +127,8 @@ test_register_capture() { echo "Register capture test [Success]" } +build_test_program + test_per_thread test_register_capture
Namhyung Kim
maintainerPATCH 4/8 perf test: Wait for a new thread when testing --per-thread record
In reply to Namhyung KimJust running the target program is not enough to test multi-thread target because it'd be racy perf vs target startup. I used the initial delay but it cannot guarantee for perf to see the thread.
Instead, use wait_for_threads helper from shell/lib/waiting.sh to make sure it starts the sibling thread first. Then perf record can use -p option to profile the target process.
--- tools/perf/tests/shell/record.sh | 27 +++++++++++++++++++++++++++ 1 file changed, 27 insertions(+)
tools/perf/tests/shell/record.sh +27 -0
@@ -4,6 +4,9 @@ set -e +shelldir=$(dirname "$0") +. "${shelldir}"/lib/waiting.sh + err=0 perfdata=$(mktemp /tmp/__perf_test.perf.data.XXXXX) testprog=$(mktemp /tmp/__perf_test.prog.XXXXXX)
@@ -96,6 +99,30 @@ test_per_thread() { err=1 return fi + + # run the test program in background (forever) + ${testprog} 1 & + TESTPID=$! + + rm -f "${perfdata}" + + wait_for_threads ${TESTPID} 2 + perf record -p "${TESTPID}" --per-thread -o "${perfdata}" sleep 1 2> /dev/null + kill ${TESTPID} + + if [ ! -e "${perfdata}" ] + then + echo "Per-thread record [Failed record -p]" + err=1 + return + fi + if ! perf report -i "${perfdata}" -q | grep -q "${testsym}" + then + echo "Per-thread record [Failed -p missing output]" + err=1 + return + fi + echo "Basic --per-thread mode test [Success]" }
Signature · 1 line
2.38.0.135.g90850a2211-goog
Re: [PATCH 4/8] perf test: Wait for a new thread when testing --per-thread record
Quoted 10 lines · Namhyung Kim
Just running the target program is not enough to test multi-thread target because it'd be racy perf vs target startup. I used the initial delay but it cannot guarantee for perf to see the thread.
Instead, use
wait_for_threadshelper fromshell/lib/waiting.shto make sure it starts the sibling thread first. Then perf record can use -p option to profile the target process.Signed-off-by Namhyung Kim <[email protected]>
Thanks,
Ian
Quoted 52 lines
--- tools/perf/tests/shell/record.sh | 27 +++++++++++++++++++++++++++ 1 file changed, 27 insertions(+)tools/perf/tests/shell/record.sh +27
-1@@ -4,6 +4,9 @@ set -e +shelldir=$(dirname "$0") +. "${shelldir}"/lib/waiting.sh + err=0 perfdata=$(mktemp /tmp/__perf_test.perf.data.XXXXX) testprog=$(mktemp /tmp/__perf_test.prog.XXXXXX)@@ -96,6 +99,30 @@ test_per_thread() { err=1 return fi + + # run the test program in background (forever) + ${testprog} 1 & + TESTPID=$! + + rm -f "${perfdata}" + + wait_for_threads ${TESTPID} 2 + perf record -p "${TESTPID}" --per-thread -o "${perfdata}" sleep 1 2> /dev/null + kill ${TESTPID} + + if [ ! -e "${perfdata}" ] + then + echo "Per-thread record [Failed record -p]" + err=1 + return + fi + if ! perf report -i "${perfdata}" -q | grep -q "${testsym}" + then + echo "Per-thread record [Failed -p missing output]" + err=1 + return + fi + echo "Basic --per-thread mode test [Success]" } --2.38.0.135.g90850a2211-goog
Re: [PATCH 4/8] perf test: Wait for a new thread when testing --per-thread record
In reply to Namhyung KimQuoted 9 lines · Namhyung Kim
Just running the target program is not enough to test multi-thread target because it'd be racy perf vs target startup. I used the initial delay but it cannot guarantee for perf to see the thread.
Instead, use
wait_for_threadshelper fromshell/lib/waiting.shto make sure it starts the sibling thread first. Then perf record can use -p option to profile the target process.Signed-off-by Namhyung Kim <[email protected]>
Quoted 49 lines
--- tools/perf/tests/shell/record.sh | 27 +++++++++++++++++++++++++++ 1 file changed, 27 insertions(+)tools/perf/tests/shell/record.sh +27
-0@@ -4,6 +4,9 @@ set -e +shelldir=$(dirname "$0") +. "${shelldir}"/lib/waiting.sh + err=0 perfdata=$(mktemp /tmp/__perf_test.perf.data.XXXXX) testprog=$(mktemp /tmp/__perf_test.prog.XXXXXX)@@ -96,6 +99,30 @@ test_per_thread() { err=1 return fi + + # run the test program in background (forever) + ${testprog} 1 & + TESTPID=$! + + rm -f "${perfdata}" + + wait_for_threads ${TESTPID} 2 + perf record -p "${TESTPID}" --per-thread -o "${perfdata}" sleep 1 2> /dev/null + kill ${TESTPID} + + if [ ! -e "${perfdata}" ] + then + echo "Per-thread record [Failed record -p]" + err=1 + return + fi + if ! perf report -i "${perfdata}" -q | grep -q "${testsym}" + then + echo "Per-thread record [Failed -p missing output]" + err=1 + return + fi + echo "Basic --per-thread mode test [Success]" }
Namhyung Kim
maintainerPATCH 5/8 perf test: Add system-wide mode in perf record tests
In reply to Namhyung KimAdd system wide recording test with the same pattern. It'd skip the test when it failes to run perf record. For system-wide mode, it needs to avoid build-id collection and synthesis because the test only cares about the test program and kernel would generates necessary events as the process starts.
--- tools/perf/tests/shell/record.sh | 21 +++++++++++++++++++++ 1 file changed, 21 insertions(+)
tools/perf/tests/shell/record.sh +21 -0
@@ -154,10 +154,31 @@ test_register_capture() { echo "Register capture test [Success]" } +test_system_wide() { + echo "Basic --system-wide mode test" + if ! perf record -aB --synth=no -o "${perfdata}" ${testprog} 2> /dev/null + then + echo "System-wide record [Skipped not supported]" + if [ $err -ne 1 ] + then + err=2 + fi + return + fi + if ! perf report -i "${perfdata}" -q | grep -q "${testsym}" + then + echo "System-wide record [Failed missing output]" + err=1 + return + fi + echo "Basic --system-wide mode test [Success]" +} + build_test_program test_per_thread test_register_capture +test_system_wide cleanup exit $err
Signature · 1 line
2.38.0.135.g90850a2211-goog
Re: [PATCH 5/8] perf test: Add system-wide mode in perf record tests
Quoted 3 lines · Namhyung Kim
Add system wide recording test with the same pattern. It'd skip the test when it failes to run perf record. For system-wide mode, it needs
nit: s/failes/fails/
Quoted 2 lines
to avoid build-id collection and synthesis because the test only cares about the test program and kernel would generates necessary events as
nit: s/generates/generate the/
Quoted 4 lines
the process starts.
Reviewed-by Adrian Hunter <[email protected]>Signed-off-by Namhyung Kim <[email protected]>
Thanks,
Ian
Quoted 43 lines
--- tools/perf/tests/shell/record.sh | 21 +++++++++++++++++++++ 1 file changed, 21 insertions(+)tools/perf/tests/shell/record.sh +21
-1@@ -154,10 +154,31 @@ test_register_capture() { echo "Register capture test [Success]" } +test_system_wide() { + echo "Basic --system-wide mode test" + if ! perf record -aB --synth=no -o "${perfdata}" ${testprog} 2> /dev/null + then + echo "System-wide record [Skipped not supported]" + if [ $err -ne 1 ] + then + err=2 + fi + return + fi + if ! perf report -i "${perfdata}" -q | grep -q "${testsym}" + then + echo "System-wide record [Failed missing output]" + err=1 + return + fi + echo "Basic --system-wide mode test [Success]" +} + build_test_program test_per_thread test_register_capture +test_system_wide cleanup exit $err --2.38.0.135.g90850a2211-goog
Arnaldo Carvalho de Melo
maintainerRe: [PATCH 5/8] perf test: Add system-wide mode in perf record tests
Em Thu, Oct 20, 2022 at 05:00:39PM -0700, Ian Rogers escreveu:
Quoted 11 lines · Namhyung Kim
Quoted 3 lines
Add system wide recording test with the same pattern. It'd skip the test when it failes to run perf record. For system-wide mode, it needs
nit: s/failes/fails/
Quoted 2 lines
to avoid build-id collection and synthesis because the test only cares about the test program and kernel would generates necessary events as
nit: s/generates/generate the/
fixed
Quoted 53 lines
Quoted 4 lines
the process starts.
Reviewed-by Adrian Hunter <[email protected]>Signed-off-by Namhyung Kim <[email protected]>Acked-by Ian Rogers <[email protected]>Thanks,
IanQuoted 43 lines
--- tools/perf/tests/shell/record.sh | 21 +++++++++++++++++++++ 1 file changed, 21 insertions(+)tools/perf/tests/shell/record.sh +21
-1@@ -154,10 +154,31 @@ test_register_capture() { echo "Register capture test [Success]" } +test_system_wide() { + echo "Basic --system-wide mode test" + if ! perf record -aB --synth=no -o "${perfdata}" ${testprog} 2> /dev/null + then + echo "System-wide record [Skipped not supported]" + if [ $err -ne 1 ] + then + err=2 + fi + return + fi + if ! perf report -i "${perfdata}" -q | grep -q "${testsym}" + then + echo "System-wide record [Failed missing output]" + err=1 + return + fi + echo "Basic --system-wide mode test [Success]" +} + build_test_program test_per_thread test_register_capture +test_system_wide cleanup exit $err --2.38.0.135.g90850a2211-goog
Signature · 1 line
- Arnaldo
Namhyung Kim
maintainerPATCH 6/8 perf test: Add target workload test in perf record tests
In reply to Namhyung KimAdd a subtest which profiles the given workload on the command line. As it's a minimal requirement, test should run ok so it doesn't skip the test even if it failed to run the perf record.
--- tools/perf/tests/shell/record.sh | 18 ++++++++++++++++++ 1 file changed, 18 insertions(+)
tools/perf/tests/shell/record.sh +18 -0
@@ -174,11 +174,29 @@ test_system_wide() { echo "Basic --system-wide mode test [Success]" } +test_workload() { + echo "Basic target workload test" + if ! perf record -o "${perfdata}" ${testprog} 2> /dev/null + then + echo "Workload record [Failed record]" + err=1 + return + fi + if ! perf report -i "${perfdata}" -q | grep -q "${testsym}" + then + echo "Workload record [Failed missing output]" + err=1 + return + fi + echo "Basic target workload test [Success]" +} + build_test_program test_per_thread test_register_capture test_system_wide +test_workload cleanup exit $err
Signature · 1 line
2.38.0.135.g90850a2211-goog
Re: [PATCH 6/8] perf test: Add target workload test in perf record tests
Quoted 6 lines · Namhyung Kim
Add a subtest which profiles the given workload on the command line. As it's a minimal requirement, test should run ok so it doesn't skip the test even if it failed to run the perf record.
Signed-off-by Namhyung Kim <[email protected]>
Thanks,
Ian
Quoted 41 lines
--- tools/perf/tests/shell/record.sh | 18 ++++++++++++++++++ 1 file changed, 18 insertions(+)tools/perf/tests/shell/record.sh +18
-1@@ -174,11 +174,29 @@ test_system_wide() { echo "Basic --system-wide mode test [Success]" } +test_workload() { + echo "Basic target workload test" + if ! perf record -o "${perfdata}" ${testprog} 2> /dev/null + then + echo "Workload record [Failed record]" + err=1 + return + fi + if ! perf report -i "${perfdata}" -q | grep -q "${testsym}" + then + echo "Workload record [Failed missing output]" + err=1 + return + fi + echo "Basic target workload test [Success]" +} + build_test_program test_per_thread test_register_capture test_system_wide +test_workload cleanup exit $err --2.38.0.135.g90850a2211-goog
Re: [PATCH 6/8] perf test: Add target workload test in perf record tests
In reply to Namhyung KimQuoted 5 lines · Namhyung Kim
Add a subtest which profiles the given workload on the command line. As it's a minimal requirement, test should run ok so it doesn't skip the test even if it failed to run the perf record.
Signed-off-by Namhyung Kim <[email protected]>
Quoted 38 lines
--- tools/perf/tests/shell/record.sh | 18 ++++++++++++++++++ 1 file changed, 18 insertions(+)tools/perf/tests/shell/record.sh +18
-0@@ -174,11 +174,29 @@ test_system_wide() { echo "Basic --system-wide mode test [Success]" } +test_workload() { + echo "Basic target workload test" + if ! perf record -o "${perfdata}" ${testprog} 2> /dev/null + then + echo "Workload record [Failed record]" + err=1 + return + fi + if ! perf report -i "${perfdata}" -q | grep -q "${testsym}" + then + echo "Workload record [Failed missing output]" + err=1 + return + fi + echo "Basic target workload test [Success]" +} + build_test_program test_per_thread test_register_capture test_system_wide +test_workload cleanup exit $err
Namhyung Kim
maintainerPATCH 7/8 perf test: Test record with --threads option
In reply to Namhyung KimThe --threads option changed the perf record behavior significantly, so it'd be nice if we test it separately. Add --threads options with different argument in each test supported and check the result.
Also update the cleanup routine because threads recording produces data in a directory.
--- tools/perf/tests/shell/record.sh | 30 ++++++++++++++++++++++++++++-- 1 file changed, 28 insertions(+), 2 deletions(-)
tools/perf/tests/shell/record.sh +28 -2
@@ -13,8 +13,8 @@ testprog=$(mktemp /tmp/__perf_test.prog.XXXXXX) testsym="test_loop" cleanup() { - rm -f "${perfdata}" - rm -f "${perfdata}".old + rm -rf "${perfdata}" + rm -rf "${perfdata}".old if [ "${testprog}" != "true" ]; then rm -f "${testprog}"
@@ -171,6 +171,19 @@ test_system_wide() { err=1 return fi + if ! perf record -aB --synth=no -e cpu-clock,cs --threads=cpu \ + -o "${perfdata}" ${testprog} 2> /dev/null + then + echo "System-wide record [Failed record --threads option]" + err=1 + return + fi + if ! perf report -i "${perfdata}" -q | grep -q "${testsym}" + then + echo "System-wide record [Failed --threads missing output]" + err=1 + return + fi echo "Basic --system-wide mode test [Success]" }
@@ -188,6 +201,19 @@ test_workload() { err=1 return fi + if ! perf record -e cpu-clock,cs --threads=package \ + -o "${perfdata}" ${testprog} 2> /dev/null + then + echo "Workload record [Failed record --threads option]" + err=1 + return + fi + if ! perf report -i "${perfdata}" -q | grep -q "${testsym}" + then + echo "Workload record [Failed --threads missing output]" + err=1 + return + fi echo "Basic target workload test [Success]" }
Signature · 1 line
2.38.0.135.g90850a2211-goog
Re: [PATCH 7/8] perf test: Test record with --threads option
Quoted 9 lines · Namhyung Kim
The --threads option changed the perf record behavior significantly, so it'd be nice if we test it separately. Add --threads options with different argument in each test supported and check the result.
Also update the cleanup routine because threads recording produces data in a directory.
Signed-off-by Namhyung Kim <[email protected]>
Thanks,
Ian
Quoted 62 lines
--- tools/perf/tests/shell/record.sh | 30 ++++++++++++++++++++++++++++-- 1 file changed, 28 insertions(+), 2 deletions(-)tools/perf/tests/shell/record.sh +28
-3@@ -13,8 +13,8 @@ testprog=$(mktemp /tmp/__perf_test.prog.XXXXXX) testsym="test_loop" cleanup() { - rm -f "${perfdata}" - rm -f "${perfdata}".old + rm -rf "${perfdata}" + rm -rf "${perfdata}".old if [ "${testprog}" != "true" ]; then rm -f "${testprog}"@@ -171,6 +171,19 @@ test_system_wide() { err=1 return fi + if ! perf record -aB --synth=no -e cpu-clock,cs --threads=cpu \ + -o "${perfdata}" ${testprog} 2> /dev/null + then + echo "System-wide record [Failed record --threads option]" + err=1 + return + fi + if ! perf report -i "${perfdata}" -q | grep -q "${testsym}" + then + echo "System-wide record [Failed --threads missing output]" + err=1 + return + fi echo "Basic --system-wide mode test [Success]" }@@ -188,6 +201,19 @@ test_workload() { err=1 return fi + if ! perf record -e cpu-clock,cs --threads=package \ + -o "${perfdata}" ${testprog} 2> /dev/null + then + echo "Workload record [Failed record --threads option]" + err=1 + return + fi + if ! perf report -i "${perfdata}" -q | grep -q "${testsym}" + then + echo "Workload record [Failed --threads missing output]" + err=1 + return + fi echo "Basic target workload test [Success]" } --2.38.0.135.g90850a2211-goog
Re: [PATCH 7/8] perf test: Test record with --threads option
In reply to Namhyung KimQuoted 8 lines · Namhyung Kim
The --threads option changed the perf record behavior significantly, so it'd be nice if we test it separately. Add --threads options with different argument in each test supported and check the result.
Also update the cleanup routine because threads recording produces data in a directory.
Signed-off-by Namhyung Kim <[email protected]>
Quoted 59 lines
--- tools/perf/tests/shell/record.sh | 30 ++++++++++++++++++++++++++++-- 1 file changed, 28 insertions(+), 2 deletions(-)tools/perf/tests/shell/record.sh +28
-2@@ -13,8 +13,8 @@ testprog=$(mktemp /tmp/__perf_test.prog.XXXXXX) testsym="test_loop" cleanup() { - rm -f "${perfdata}" - rm -f "${perfdata}".old + rm -rf "${perfdata}" + rm -rf "${perfdata}".old if [ "${testprog}" != "true" ]; then rm -f "${testprog}"@@ -171,6 +171,19 @@ test_system_wide() { err=1 return fi + if ! perf record -aB --synth=no -e cpu-clock,cs --threads=cpu \ + -o "${perfdata}" ${testprog} 2> /dev/null + then + echo "System-wide record [Failed record --threads option]" + err=1 + return + fi + if ! perf report -i "${perfdata}" -q | grep -q "${testsym}" + then + echo "System-wide record [Failed --threads missing output]" + err=1 + return + fi echo "Basic --system-wide mode test [Success]" }@@ -188,6 +201,19 @@ test_workload() { err=1 return fi + if ! perf record -e cpu-clock,cs --threads=package \ + -o "${perfdata}" ${testprog} 2> /dev/null + then + echo "Workload record [Failed record --threads option]" + err=1 + return + fi + if ! perf report -i "${perfdata}" -q | grep -q "${testsym}" + then + echo "Workload record [Failed --threads missing output]" + err=1 + return + fi echo "Basic target workload test [Success]" }
Namhyung Kim
maintainerPATCH 8/8 perf test: Do not set TEST_SKIP for record subtests
In reply to Namhyung KimIt now has 4 sub tests and one of them should run at least. But once TEST_SKIP (= 2) return value is set, it won't be overwritten unless there's a failure. I think we should return success when one or more tested are skipped but the remaining subtests are passed.
So update the test code not to set the err variable when it skips the test.
--- tools/perf/tests/shell/record.sh | 12 ------------ 1 file changed, 12 deletions(-)
tools/perf/tests/shell/record.sh +0 -12
@@ -81,10 +81,6 @@ test_per_thread() { if ! perf record -o /dev/null --quiet ${testprog} 2> /dev/null then echo "Per-thread record [Skipped event not supported]" - if [ $err -ne 1 ] - then - err=2 - fi return fi if ! perf record --per-thread -o "${perfdata}" ${testprog} 2> /dev/null
@@ -131,10 +127,6 @@ test_register_capture() { if ! perf list | grep -q 'br_inst_retired.near_call' then echo "Register capture test [Skipped missing event]" - if [ $err -ne 1 ] - then - err=2 - fi return fi if ! perf record --intr-regs=\? 2>&1 | grep -q 'available registers: AX BX CX DX SI DI BP SP IP FLAGS CS SS R8 R9 R10 R11 R12 R13 R14 R15'
@@ -159,10 +151,6 @@ test_system_wide() { if ! perf record -aB --synth=no -o "${perfdata}" ${testprog} 2> /dev/null then echo "System-wide record [Skipped not supported]" - if [ $err -ne 1 ] - then - err=2 - fi return fi if ! perf report -i "${perfdata}" -q | grep -q "${testsym}"
Signature · 1 line
2.38.0.135.g90850a2211-goog
Re: [PATCH 8/8] perf test: Do not set TEST_SKIP for record subtests
Quoted 11 lines · Namhyung Kim
It now has 4 sub tests and one of them should run at least. But once
TEST_SKIP(= 2) return value is set, it won't be overwritten unless there's a failure. I think we should return success when one or more tested are skipped but the remaining subtests are passed.So update the test code not to set the err variable when it skips the test.
Reviewed-by Adrian Hunter <[email protected]>Signed-off-by Namhyung Kim <[email protected]>
Perhaps we can have proper sub-tests in shell tests like we do for the C ones.
Thanks,
Ian
Quoted 44 lines
--- tools/perf/tests/shell/record.sh | 12 ------------ 1 file changed, 12 deletions(-)tools/perf/tests/shell/record.sh +0
-13@@ -81,10 +81,6 @@ test_per_thread() { if ! perf record -o /dev/null --quiet ${testprog} 2> /dev/null then echo "Per-thread record [Skipped event not supported]" - if [ $err -ne 1 ] - then - err=2 - fi return fi if ! perf record --per-thread -o "${perfdata}" ${testprog} 2> /dev/null@@ -131,10 +127,6 @@ test_register_capture() { if ! perf list | grep -q 'br_inst_retired.near_call' then echo "Register capture test [Skipped missing event]" - if [ $err -ne 1 ] - then - err=2 - fi return fi if ! perf record --intr-regs=\? 2>&1 | grep -q 'available registers: AX BX CX DX SI DI BP SP IP FLAGS CS SS R8 R9 R10 R11 R12 R13 R14 R15'@@ -159,10 +151,6 @@ test_system_wide() { if ! perf record -aB --synth=no -o "${perfdata}" ${testprog} 2> /dev/null then echo "System-wide record [Skipped not supported]" - if [ $err -ne 1 ] - then - err=2 - fi return fi if ! perf report -i "${perfdata}" -q | grep -q "${testsym}" --2.38.0.135.g90850a2211-goog
Arnaldo Carvalho de Melo
maintainerRe: [PATCH 8/8] perf test: Do not set TEST_SKIP for record subtests
Em Thu, Oct 20, 2022 at 05:05:49PM -0700, Ian Rogers escreveu:
Quoted 10 lines · Namhyung Kim
Quoted 9 lines
It now has 4 sub tests and one of them should run at least. But once
TEST_SKIP(= 2) return value is set, it won't be overwritten unless there's a failure. I think we should return success when one or more tested are skipped but the remaining subtests are passed.So update the test code not to set the err variable when it skips the test.
Quoted 6 lines
Quoted 2 lines
Reviewed-by Adrian Hunter <[email protected]>Signed-off-by Namhyung Kim <[email protected]>Acked-by Ian Rogers <[email protected]>Perhaps we can have proper sub-tests in shell tests like we do for the C ones.
Yeah, giving info about the various subtests helps in giving progress information and overal confidence that more stuff is being tested.
- Arnaldo
Quoted 47 lines
Thanks,
IanQuoted 44 lines
--- tools/perf/tests/shell/record.sh | 12 ------------ 1 file changed, 12 deletions(-)tools/perf/tests/shell/record.sh +0
-13@@ -81,10 +81,6 @@ test_per_thread() { if ! perf record -o /dev/null --quiet ${testprog} 2> /dev/null then echo "Per-thread record [Skipped event not supported]" - if [ $err -ne 1 ] - then - err=2 - fi return fi if ! perf record --per-thread -o "${perfdata}" ${testprog} 2> /dev/null@@ -131,10 +127,6 @@ test_register_capture() { if ! perf list | grep -q 'br_inst_retired.near_call' then echo "Register capture test [Skipped missing event]" - if [ $err -ne 1 ] - then - err=2 - fi return fi if ! perf record --intr-regs=\? 2>&1 | grep -q 'available registers: AX BX CX DX SI DI BP SP IP FLAGS CS SS R8 R9 R10 R11 R12 R13 R14 R15'@@ -159,10 +151,6 @@ test_system_wide() { if ! perf record -aB --synth=no -o "${perfdata}" ${testprog} 2> /dev/null then echo "System-wide record [Skipped not supported]" - if [ $err -ne 1 ] - then - err=2 - fi return fi if ! perf report -i "${perfdata}" -q | grep -q "${testsym}" --2.38.0.135.g90850a2211-goog
Signature · 1 line
- Arnaldo
Em Thu, Oct 20, 2022 at 10:26:35AM -0700, Namhyung Kim escreveu:
Quoted 13 lines
Hello,
This patchset improves the perf record tests to check more cases so that it can find problems early. The motivation was a failure per-thread mmap with multi-threaded targets which Adrian is working on the fix.
Changes in v2)
* fix shellcheck issues
* drop unsupported --per-thread and --threads combination
* do not use initial delay (-D option); instead it runs the target and wait for it separately using the recentwaiting.shlibrary
* add Adrian's Reviewed-by tags
Quoted 3 lines
I added a custom test program and more combinations like system-wide and command line workload (in per-process mode) testing with multi-threaded recording mode.
Thanks, applied to perf/core.
- Arnaldo
Quoted 44 lines
Currently it succeeds every tests when running as root!
$ sudo ./perf test -v 91 91: perf record tests : --- start --- test child forked, pid 108975 Build a test program Basic --per-thread mode test Basic --per-thread mode test [Success] Register capture test Register capture test [Success] Basic --system-wide mode test Basic --system-wide mode test [Success] Basic target workload test Basic target workload test [Success] test child finished with 0 ---- end ---- perf record tests: OkYou can find it in 'perf/record-test-v2' branch in
git://git.kernel.org/pub/scm/linux/kernel/git/namhyung/linux-perf.gitThanks,
NamhyungNamhyung Kim (8):
perf test: Do not use instructions:u explicitly
perf test: Fix shellcheck issues in the record test perf test: Use a test program in perf record tests perf test: Wait for a new thread when testing --per-thread record perf test: Add system-wide mode in perf record tests perf test: Add target workload test in perf record tests perf test: Test record with --threads option
perf test: Do not setTEST_SKIPfor record subteststools/perf/tests/shell/record.sh | 180 +++++++++++++++++++++++++++---- 1 file changed, 158 insertions(+), 22 deletions(-)base-commit:
a3a365655a28f12f07eddf4f3fd596987b175e1dSignature · 1 line
2.38.0.135.g90850a2211-goog
Signature · 1 line
- Arnaldo