Skip to contentSkip to conversation

perf test: Improve perf record tests (v2)

PATCH Applied Namhyung Kim 4 participants 32 messages Oct 20, 2022 – Oct 26, 2022 perf events

lore
Share & tools
lore

Namhyung Kim

author maintainer
#0 lore raw

cover 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
author maintainer
#1 lore raw

PATCH 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
index 301f95427159..747c33a1ec45 100755
@@ -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
reviewer Acked-by on v1 1/8
#2 lore raw

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.

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
index 301f95427159..747c33a1ec45 100755
@@ -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

author maintainer
#3 lore raw

PATCH 2/8 perf test: Fix shellcheck issues in the record test

In reply to Namhyung Kim

Basically 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
index 747c33a1ec45..464071462809 100755
@@ -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
reviewer Acked-by on v1 2/8
#4 lore raw

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 names

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
index 747c33a1ec45..464071462809 100755
@@ -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

reviewer Reviewed-by on v1 2/8
#5 lore raw

Re: [PATCH 2/8] perf test: Fix shellcheck issues in the record test

In reply to Namhyung Kim
Quoted 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 names

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
index 747c33a1ec45..464071462809 100755
@@ -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
author maintainer
#6 lore raw

PATCH 3/8 perf test: Use a test program in perf record tests

In reply to 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'.

---
 tools/perf/tests/shell/record.sh | 64 ++++++++++++++++++++++++++++++--
 1 file changed, 60 insertions(+), 4 deletions(-)
tools/perf/tests/shell/record.sh +60 -4
index 464071462809..952981481239 100755
@@ -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
reviewer Acked-by on v1 3/8
#7 lore raw

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'.

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
index 464071462809..952981481239 100755
@@ -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

#8 lore raw

Re: [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

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,
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
index 464071462809..952981481239 100755
@@ -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
reviewer
#9 lore raw

Re: [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'.

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.

#10 lore raw

Re: [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

author maintainer
#11 lore raw

Re: [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 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

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

#12 lore raw

Re: [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 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

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

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

reviewer
#13 lore raw

Re: [PATCH 3/8] perf test: Use a test program in perf record tests

In reply to Arnaldo Carvalho de Melo
Quoted 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.

author maintainer
#14 lore raw

Re: [PATCH 3/8] perf test: Use a test program in perf record tests

In reply to Arnaldo Carvalho de Melo
Quoted 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'.

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

reviewer Reviewed-by on v1 3/8
#15 lore raw

Re: [PATCH 3/8] perf test: Use a test program in perf record tests

In reply to Namhyung Kim
Quoted 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
---
 tools/perf/tests/shell/record.sh | 64 ++++++++++++++++++++++++++++++--
 1 file changed, 60 insertions(+), 4 deletions(-)
tools/perf/tests/shell/record.sh +60 -4
index 464071462809..952981481239 100755
@@ -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
 
author maintainer
#16 lore raw

PATCH 4/8 perf test: Wait for a new thread when testing --per-thread record

In reply to 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_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
index 952981481239..d1640d1daf2e 100755
@@ -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
reviewer Acked-by on v1 4/8
#17 lore raw

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_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.

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
index 952981481239..d1640d1daf2e 100755
@@ -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

reviewer Reviewed-by on v1 4/8
#18 lore raw

Re: [PATCH 4/8] perf test: Wait for a new thread when testing --per-thread record

In reply to Namhyung Kim
Quoted 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_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.

Quoted 49 lines
---
 tools/perf/tests/shell/record.sh | 27 +++++++++++++++++++++++++++
 1 file changed, 27 insertions(+)
tools/perf/tests/shell/record.sh +27 -0
index 952981481239..d1640d1daf2e 100755
@@ -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]"
 }
 
author maintainer
#19 lore raw

PATCH 5/8 perf test: Add system-wide mode in perf record tests

In reply to 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 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
index d1640d1daf2e..345764afb745 100755
@@ -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
reviewer Acked-by on v1 5/8
#20 lore raw

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.

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
index d1640d1daf2e..345764afb745 100755
@@ -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

#21 lore raw

Re: [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.

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
index d1640d1daf2e..345764afb745 100755
@@ -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
author maintainer
#22 lore raw

PATCH 6/8 perf test: Add target workload test in perf record tests

In reply to 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.

---
 tools/perf/tests/shell/record.sh | 18 ++++++++++++++++++
 1 file changed, 18 insertions(+)
tools/perf/tests/shell/record.sh +18 -0
index 345764afb745..c59d1459c960 100755
@@ -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
reviewer Acked-by on v1 6/8
#23 lore raw

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.

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
index 345764afb745..c59d1459c960 100755
@@ -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

reviewer Reviewed-by on v1 6/8
#24 lore raw

Re: [PATCH 6/8] perf test: Add target workload test in perf record tests

In reply to Namhyung Kim
Quoted 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.

Quoted 38 lines
---
 tools/perf/tests/shell/record.sh | 18 ++++++++++++++++++
 1 file changed, 18 insertions(+)
tools/perf/tests/shell/record.sh +18 -0
index 345764afb745..c59d1459c960 100755
@@ -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
author maintainer
#25 lore raw

PATCH 7/8 perf test: Test record with --threads option

In reply to 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.

---
 tools/perf/tests/shell/record.sh | 30 ++++++++++++++++++++++++++++--
 1 file changed, 28 insertions(+), 2 deletions(-)
tools/perf/tests/shell/record.sh +28 -2
index c59d1459c960..01aa9531b369 100755
@@ -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
reviewer Acked-by on v1 7/8
#26 lore raw

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.

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
index c59d1459c960..01aa9531b369 100755
@@ -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

reviewer Reviewed-by on v1 7/8
#27 lore raw

Re: [PATCH 7/8] perf test: Test record with --threads option

In reply to Namhyung Kim
Quoted 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.

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
index c59d1459c960..01aa9531b369 100755
@@ -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]"
 }
 
author maintainer
#28 lore raw

PATCH 8/8 perf test: Do not set TEST_SKIP for record subtests

In reply to 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.

---
 tools/perf/tests/shell/record.sh | 12 ------------
 1 file changed, 12 deletions(-)
tools/perf/tests/shell/record.sh +0 -12
index 01aa9531b369..e93b3a8871fe 100755
@@ -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
reviewer Acked-by on v1 8/8
#29 lore raw

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.

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
index 01aa9531b369..e93b3a8871fe 100755
@@ -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

#30 lore raw

Re: [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

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,
Ian

Quoted 44 lines
---
 tools/perf/tests/shell/record.sh | 12 ------------
 1 file changed, 12 deletions(-)
tools/perf/tests/shell/record.sh +0 -13
index 01aa9531b369..e93b3a8871fe 100755
@@ -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
maintainer Applied
#31 lore raw
In reply to Namhyung Kim

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 recent waiting.sh library
* 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: 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
Signature · 1 line
- Arnaldo