Skip to contentSkip to conversation

Test/uniquification related fixes

PATCH Applied Ian Rogers 2 participants 12 messages Aug 21, 2025 – Sep 18, 2025 perf events

v1 of 2 Applied Latest is v2 →

lore
Share & tools
lore

Ian Rogers

author reviewer
#0 lore raw

cover Test/uniquification related fixes

Failures discovered running as non-root and with v6.17-rc1 changes.

Ian Rogers (4):
perf test shell lbr: Avoid failures with perf event paranoia perf test: Don't leak workload gopipe in PERF_RECORD_* perf evsel: Fix uniquification when PMU given without suffix perf test: Avoid uncore_imc/clockticks in uniquification test

 tools/perf/tests/perf-record.c                |   4 +
 tools/perf/tests/shell/record_lbr.sh          |  26 ++++-
 .../tests/shell/stat+event_uniquifying.sh     | 105 ++++++++----------
 tools/perf/util/evsel.c                       |  28 +++--
 4 files changed, 86 insertions(+), 77 deletions(-)
Signature · 1 line
2.51.0.rc2.233.g662b1ed5c5-goog
author reviewer
#1 lore raw

PATCH 1/4 perf test shell lbr: Avoid failures with perf event paranoia

When not running as root and with higher perf event paranoia values the perf record LBR tests could fail rather than skipping the problematic tests. Add the sensitivity to the test and confirm it passes with paranoia values from -1 to 2.

Fixes 32559b99e0f5 ("perf test: Add set of perf record LBR tests")
---
 tools/perf/tests/shell/record_lbr.sh | 26 ++++++++++++++++++++------
 1 file changed, 20 insertions(+), 6 deletions(-)
tools/perf/tests/shell/record_lbr.sh +20 -6
index 6fcb5e52b9b4..78a02e90ece1 100755
@@ -4,6 +4,10 @@
 
 set -e
 
+ParanoidAndNotRoot() {
+  [ "$(id -u)" != 0 ] && [ "$(cat /proc/sys/kernel/perf_event_paranoid)" -gt $1 ]
+}
+
 if [ ! -f /sys/bus/event_source/devices/cpu/caps/branches ] &&
    [ ! -f /sys/bus/event_source/devices/cpu_core/caps/branches ]
 then
@@ -23,6 +27,7 @@ cleanup() {
 }
 
 trap_cleanup() {
+  echo "Unexpected signal in ${FUNCNAME[1]}"
   cleanup
   exit 1
 }
@@ -123,8 +128,11 @@ lbr_test "-j ind_call" "any indirect call" 2
 lbr_test "-j ind_jmp" "any indirect jump" 100
 lbr_test "-j call" "direct calls" 2
 lbr_test "-j ind_call,u" "any indirect user call" 100
-lbr_test "-a -b" "system wide any branch" 2
-lbr_test "-a -j any_call" "system wide any call" 2
+if ! ParanoidAndNotRoot 1
+then
+  lbr_test "-a -b" "system wide any branch" 2
+  lbr_test "-a -j any_call" "system wide any call" 2
+fi
 
 # Parallel
 parallel_lbr_test "-b" "parallel any branch" 100 &
@@ -141,10 +149,16 @@ parallel_lbr_test "-j call" "parallel direct calls" 100 &
 pid6=$!
 parallel_lbr_test "-j ind_call,u" "parallel any indirect user call" 100 &
 pid7=$!
-parallel_lbr_test "-a -b" "parallel system wide any branch" 100 &
-pid8=$!
-parallel_lbr_test "-a -j any_call" "parallel system wide any call" 100 &
-pid9=$!
+if ParanoidAndNotRoot 1
+then
+  pid8=
+  pid9=
+else
+  parallel_lbr_test "-a -b" "parallel system wide any branch" 100 &
+  pid8=$!
+  parallel_lbr_test "-a -j any_call" "parallel system wide any call" 100 &
+  pid9=$!
+fi
 
 for pid in $pid1 $pid2 $pid3 $pid4 $pid5 $pid6 $pid7 $pid8 $pid9
 do
Signature · 1 line
2.51.0.rc2.233.g662b1ed5c5-goog
maintainer Applied
#2 lore raw

Re: [PATCH v1 1/4] perf test shell lbr: Avoid failures with perf event paranoia

Quoted 4 lines · Ian Rogers

When not running as root and with higher perf event paranoia values the perf record LBR tests could fail rather than skipping the problematic tests. Add the sensitivity to the test and confirm it passes with paranoia values from -1 to 2.

Thanks, applied to perf-tools-next,

Committer testing:

Testing with '$ perf test -vv lbr', i.e. as non root, and then comparing the output shows the mentioned errors before this patch:

  acme@x1:~$ grep -m1 "model name" /proc/cpuinfo
  model name    : 13th Gen Intel(R) Core(TM) i7-1365U
  acme@x1:~$ 

Before:

132: perf record LBR tests : Skip

After:

132: perf record LBR tests : Ok

- Arnaldo

Quoted 66 lines
Fixes 32559b99e0f5 ("perf test: Add set of perf record LBR tests")
---
 tools/perf/tests/shell/record_lbr.sh | 26 ++++++++++++++++++++------
 1 file changed, 20 insertions(+), 6 deletions(-)
tools/perf/tests/shell/record_lbr.sh +20 -6
index 6fcb5e52b9b4..78a02e90ece1 100755
@@ -4,6 +4,10 @@
 
 set -e
 
+ParanoidAndNotRoot() {
+  [ "$(id -u)" != 0 ] && [ "$(cat /proc/sys/kernel/perf_event_paranoid)" -gt $1 ]
+}
+
 if [ ! -f /sys/bus/event_source/devices/cpu/caps/branches ] &&
    [ ! -f /sys/bus/event_source/devices/cpu_core/caps/branches ]
 then
@@ -23,6 +27,7 @@ cleanup() {
 }
 
 trap_cleanup() {
+  echo "Unexpected signal in ${FUNCNAME[1]}"
   cleanup
   exit 1
 }
@@ -123,8 +128,11 @@ lbr_test "-j ind_call" "any indirect call" 2
 lbr_test "-j ind_jmp" "any indirect jump" 100
 lbr_test "-j call" "direct calls" 2
 lbr_test "-j ind_call,u" "any indirect user call" 100
-lbr_test "-a -b" "system wide any branch" 2
-lbr_test "-a -j any_call" "system wide any call" 2
+if ! ParanoidAndNotRoot 1
+then
+  lbr_test "-a -b" "system wide any branch" 2
+  lbr_test "-a -j any_call" "system wide any call" 2
+fi
 
 # Parallel
 parallel_lbr_test "-b" "parallel any branch" 100 &
@@ -141,10 +149,16 @@ parallel_lbr_test "-j call" "parallel direct calls" 100 &
 pid6=$!
 parallel_lbr_test "-j ind_call,u" "parallel any indirect user call" 100 &
 pid7=$!
-parallel_lbr_test "-a -b" "parallel system wide any branch" 100 &
-pid8=$!
-parallel_lbr_test "-a -j any_call" "parallel system wide any call" 100 &
-pid9=$!
+if ParanoidAndNotRoot 1
+then
+  pid8=
+  pid9=
+else
+  parallel_lbr_test "-a -b" "parallel system wide any branch" 100 &
+  pid8=$!
+  parallel_lbr_test "-a -j any_call" "parallel system wide any call" 100 &
+  pid9=$!
+fi
 
 for pid in $pid1 $pid2 $pid3 $pid4 $pid5 $pid6 $pid7 $pid8 $pid9
 do
Signature · 1 line
2.51.0.rc2.233.g662b1ed5c5-goog
author reviewer
#3 lore raw

PATCH 2/4 perf test: Don't leak workload gopipe in PERF_RECORD_*

In reply to Ian Rogers

The test starts a workload and then opens events. If the events fail to open, for example because of perf_event_paranoid, the gopipe of the workload is leaked and the file descriptor leak check fails when the test exits. To avoid this cancel the workload when opening the events fails.

Before:

```
$ perf test -vv 7
  7: PERF_RECORD_* events & perf_sample fields:
--- start ---
test child forked, pid 1189568
Using CPUID GenuineIntel-6-B7-1
------------------------------------------------------------
perf_event_attr:
  type                             0 (PERF_TYPE_HARDWARE)
  config                           0xa00000000 (cpu_atom/PERF_COUNT_HW_CPU_CYCLES/)
  disabled                         1
------------------------------------------------------------
sys_perf_event_open: pid 0  cpu -1  group_fd -1  flags 0x8
sys_perf_event_open failed, error -13
------------------------------------------------------------
perf_event_attr:
  type                             0 (PERF_TYPE_HARDWARE)
  config                           0xa00000000 (cpu_atom/PERF_COUNT_HW_CPU_CYCLES/)
  disabled                         1
  exclude_kernel                   1
------------------------------------------------------------
sys_perf_event_open: pid 0  cpu -1  group_fd -1  flags 0x8 = 3
------------------------------------------------------------
perf_event_attr:
  type                             0 (PERF_TYPE_HARDWARE)
  config                           0x400000000 (cpu_core/PERF_COUNT_HW_CPU_CYCLES/)
  disabled                         1
------------------------------------------------------------
sys_perf_event_open: pid 0  cpu -1  group_fd -1  flags 0x8
sys_perf_event_open failed, error -13
------------------------------------------------------------
perf_event_attr:
  type                             0 (PERF_TYPE_HARDWARE)
  config                           0x400000000 (cpu_core/PERF_COUNT_HW_CPU_CYCLES/)
  disabled                         1
  exclude_kernel                   1
------------------------------------------------------------
sys_perf_event_open: pid 0  cpu -1  group_fd -1  flags 0x8 = 3
Attempt to add: software/cpu-clock/
..after resolving event: software/config=0/
cpu-clock -> software/cpu-clock/
------------------------------------------------------------
perf_event_attr:
  type                             1 (PERF_TYPE_SOFTWARE)
  size                             136
  config                           0x9 (PERF_COUNT_SW_DUMMY)
  sample_type                      IP|TID|TIME|CPU
  read_format                      ID|LOST
  disabled                         1
  inherit                          1
  mmap                             1
  comm                             1
  enable_on_exec                   1
  task                             1
  sample_id_all                    1
  mmap2                            1
  comm_exec                        1
  ksymbol                          1
  bpf_event                        1
  { wakeup_events, wakeup_watermark } 1
------------------------------------------------------------
sys_perf_event_open: pid 1189569  cpu 0  group_fd -1  flags 0x8
sys_perf_event_open failed, error -13
perf_evlist__open: Permission denied
---- end(-2) ----
Leak of file descriptor 6 that opened: 'pipe:[14200347]'

---- unexpected signal (6) ----
Failed to read build ID for //anon
Failed to read build ID for //anon
Failed to read build ID for //anon
Failed to read build ID for //anon
Failed to read build ID for //anon
Failed to read build ID for //anon
Failed to read build ID for //anon
Failed to read build ID for //anon
Failed to read build ID for //anon
Failed to read build ID for //anon
Failed to read build ID for //anon
Failed to read build ID for //anon
Failed to read build ID for //anon
Failed to read build ID for //anon
Failed to read build ID for //anon
Failed to read build ID for //anon
Failed to read build ID for //anon
Failed to read build ID for //anon
Failed to read build ID for //anon
Failed to read build ID for //anon
Failed to read build ID for //anon
Failed to read build ID for //anon
Failed to read build ID for //anon
Failed to read build ID for //anon
Failed to read build ID for //anon
Failed to read build ID for //anon
Failed to read build ID for //anon
Failed to read build ID for //anon
Failed to read build ID for //anon
Failed to read build ID for //anon
Failed to read build ID for //anon
Failed to read build ID for //anon
    #0 0x565358f6666e in child_test_sig_handler builtin-test.c:311
    #1 0x7f29ce849df0 in __restore_rt libc_sigaction.c:0
    #2 0x7f29ce89e95c in __pthread_kill_implementation pthread_kill.c:44
    #3 0x7f29ce849cc2 in raise raise.c:27
    #4 0x7f29ce8324ac in abort abort.c:81
    #5 0x565358f662d4 in check_leaks builtin-test.c:226
    #6 0x565358f6682e in run_test_child builtin-test.c:344
    #7 0x565358ef7121 in start_command run-command.c:128
    #8 0x565358f67273 in start_test builtin-test.c:545
    #9 0x565358f6771d in __cmd_test builtin-test.c:647
    #10 0x565358f682bd in cmd_test builtin-test.c:849
    #11 0x565358ee5ded in run_builtin perf.c:349
    #12 0x565358ee6085 in handle_internal_command perf.c:401
    #13 0x565358ee61de in run_argv perf.c:448
    #14 0x565358ee6527 in main perf.c:555
    #15 0x7f29ce833ca8 in __libc_start_call_main libc_start_call_main.h:74
    #16 0x7f29ce833d65 in __libc_start_main@@GLIBC_2.34 libc-start.c:128
    #17 0x565358e391c1 in _start perf[851c1]
  7: PERF_RECORD_* events & perf_sample fields                       : FAILED!
```

After:

```
$ perf test 7
  7: PERF_RECORD_* events & perf_sample fields                       : Skip (permissions)
```
Fixes 16d00fee7038 ("perf tests: Move test__PERF_RECORD into separate object")
---
 tools/perf/tests/perf-record.c | 4 ++++
 1 file changed, 4 insertions(+)
tools/perf/tests/perf-record.c +4 -0
index 0b3c37e66871..8c79b5166a05 100644
@@ -115,6 +115,7 @@ static int test__PERF_RECORD(struct test_suite *test __maybe_unused, int subtest
 	if (err < 0) {
 		pr_debug("sched__get_first_possible_cpu: %s\n",
 			 str_error_r(errno, sbuf, sizeof(sbuf)));
+		evlist__cancel_workload(evlist);
 		goto out_delete_evlist;
 	}
 
@@ -126,6 +127,7 @@ static int test__PERF_RECORD(struct test_suite *test __maybe_unused, int subtest
 	if (sched_setaffinity(evlist->workload.pid, cpu_mask_size, &cpu_mask) < 0) {
 		pr_debug("sched_setaffinity: %s\n",
 			 str_error_r(errno, sbuf, sizeof(sbuf)));
+		evlist__cancel_workload(evlist);
 		goto out_delete_evlist;
 	}
 
@@ -137,6 +139,7 @@ static int test__PERF_RECORD(struct test_suite *test __maybe_unused, int subtest
 	if (err < 0) {
 		pr_debug("perf_evlist__open: %s\n",
 			 str_error_r(errno, sbuf, sizeof(sbuf)));
+		evlist__cancel_workload(evlist);
 		goto out_delete_evlist;
 	}
 
@@ -149,6 +152,7 @@ static int test__PERF_RECORD(struct test_suite *test __maybe_unused, int subtest
 	if (err < 0) {
 		pr_debug("evlist__mmap: %s\n",
 			 str_error_r(errno, sbuf, sizeof(sbuf)));
+		evlist__cancel_workload(evlist);
 		goto out_delete_evlist;
 	}
 
Signature · 1 line
2.51.0.rc2.233.g662b1ed5c5-goog
#4 lore raw

Re: [PATCH v1 2/4] perf test: Don't leak workload gopipe in PERF_RECORD_*

Quoted 11 lines · Ian Rogers

The test starts a workload and then opens events. If the events fail to open, for example because of perf_event_paranoid, the gopipe of the workload is leaked and the file descriptor leak check fails when the test exits. To avoid this cancel the workload when opening the events fails.

Before:
```

$ perf test -vv 7
  7: PERF_RECORD_* events & perf_sample fields:
--- start ---

⬢ [acme@toolbx perf-tools-next]$ patch -p1 < b
patching file tools/perf/tests/perf-record.c
Hunk #1 succeeded at 130 (offset 15 lines).
Hunk #2 succeeded at 142 with fuzz 1 (offset 15 lines).
Hunk #3 succeeded at 154 (offset 15 lines).
Hunk #4 succeeded at 167 (offset 15 lines).
⬢ [acme@toolbx perf-tools-next]$
⬢ [acme@toolbx perf-tools-next]$ git log --oneline -5 tools/perf/tests/perf-record.c
576bd7a8c90c48e9 (x1/perf-tools-next, x1/HEAD, five/perf-tools-next, five/HEAD) perf tests record: Update testcase to fix usage of affinity for machines with #CPUs > 1K b4c658d4d63d6149 perf target: Remove uid from target
dc6d2bc2d893a878 perf sample: Make user_regs and intr_regs optional
fd8d5a3b076c033f perf tests: Add missing event.h include
9823147da6c893d9 perf tools: Move 'struct perf_sample' to a separate header file to disentangle headers
⬢ [acme@toolbx perf-tools-next]$

Can you please check that it is still ok?

I processed the first in the series and now I'm going thru the other two.

- Arnaldo

Quoted 167 lines

test child forked, pid 1189568
Using CPUID GenuineIntel-6-B7-1
------------------------------------------------------------
perf_event_attr:

  type                             0 (PERF_TYPE_HARDWARE)
  config                           0xa00000000 (cpu_atom/PERF_COUNT_HW_CPU_CYCLES/)
  disabled                         1

------------------------------------------------------------
sys_perf_event_open: pid 0 cpu -1 group_fd -1 flags 0x8 sys_perf_event_open failed, error -13
------------------------------------------------------------
perf_event_attr:

  type                             0 (PERF_TYPE_HARDWARE)
  config                           0xa00000000 (cpu_atom/PERF_COUNT_HW_CPU_CYCLES/)
  disabled                         1
  exclude_kernel                   1

------------------------------------------------------------
sys_perf_event_open: pid 0 cpu -1 group_fd -1 flags 0x8 = 3
------------------------------------------------------------
perf_event_attr:

  type                             0 (PERF_TYPE_HARDWARE)
  config                           0x400000000 (cpu_core/PERF_COUNT_HW_CPU_CYCLES/)
  disabled                         1

------------------------------------------------------------
sys_perf_event_open: pid 0 cpu -1 group_fd -1 flags 0x8 sys_perf_event_open failed, error -13
------------------------------------------------------------
perf_event_attr:

  type                             0 (PERF_TYPE_HARDWARE)
  config                           0x400000000 (cpu_core/PERF_COUNT_HW_CPU_CYCLES/)
  disabled                         1
  exclude_kernel                   1

------------------------------------------------------------
sys_perf_event_open: pid 0 cpu -1 group_fd -1 flags 0x8 = 3 Attempt to add: software/cpu-clock/
..after resolving event: software/config=0/ cpu-clock -> software/cpu-clock/
------------------------------------------------------------
perf_event_attr:

  type                             1 (PERF_TYPE_SOFTWARE)
  size                             136
  config                           0x9 (PERF_COUNT_SW_DUMMY)
  sample_type                      IP|TID|TIME|CPU
  read_format                      ID|LOST
  disabled                         1
  inherit                          1
  mmap                             1
  comm                             1
  enable_on_exec                   1
  task                             1
  sample_id_all                    1
  mmap2                            1
  comm_exec                        1
  ksymbol                          1
  bpf_event                        1

{ wakeup_events, wakeup_watermark } 1
------------------------------------------------------------
sys_perf_event_open: pid 1189569 cpu 0 group_fd -1 flags 0x8 sys_perf_event_open failed, error -13
perf_evlist__open: Permission denied
---- end(-2) ----
Leak of file descriptor 6 that opened: 'pipe:[14200347]'

---- unexpected signal (6) ----
Failed to read build ID for //anon
Failed to read build ID for //anon
Failed to read build ID for //anon
Failed to read build ID for //anon
Failed to read build ID for //anon
Failed to read build ID for //anon
Failed to read build ID for //anon
Failed to read build ID for //anon
Failed to read build ID for //anon
Failed to read build ID for //anon
Failed to read build ID for //anon
Failed to read build ID for //anon
Failed to read build ID for //anon
Failed to read build ID for //anon
Failed to read build ID for //anon
Failed to read build ID for //anon
Failed to read build ID for //anon
Failed to read build ID for //anon
Failed to read build ID for //anon
Failed to read build ID for //anon
Failed to read build ID for //anon
Failed to read build ID for //anon
Failed to read build ID for //anon
Failed to read build ID for //anon
Failed to read build ID for //anon
Failed to read build ID for //anon
Failed to read build ID for //anon
Failed to read build ID for //anon
Failed to read build ID for //anon
Failed to read build ID for //anon
Failed to read build ID for //anon
Failed to read build ID for //anon

    #0 0x565358f6666e in child_test_sig_handler builtin-test.c:311
    #1 0x7f29ce849df0 in __restore_rt libc_sigaction.c:0
    #2 0x7f29ce89e95c in __pthread_kill_implementation pthread_kill.c:44
    #3 0x7f29ce849cc2 in raise raise.c:27
    #4 0x7f29ce8324ac in abort abort.c:81
    #5 0x565358f662d4 in check_leaks builtin-test.c:226
    #6 0x565358f6682e in run_test_child builtin-test.c:344
    #7 0x565358ef7121 in start_command run-command.c:128
    #8 0x565358f67273 in start_test builtin-test.c:545
    #9 0x565358f6771d in __cmd_test builtin-test.c:647
    #10 0x565358f682bd in cmd_test builtin-test.c:849
    #11 0x565358ee5ded in run_builtin perf.c:349
    #12 0x565358ee6085 in handle_internal_command perf.c:401
    #13 0x565358ee61de in run_argv perf.c:448
    #14 0x565358ee6527 in main perf.c:555
    #15 0x7f29ce833ca8 in __libc_start_call_main libc_start_call_main.h:74
    #16 0x7f29ce833d65 in __libc_start_main@@GLIBC_2.34 libc-start.c:128
    #17 0x565358e391c1 in _start perf[851c1]
  7: PERF_RECORD_* events & perf_sample fields                       : FAILED!
```

After:
```
$ perf test 7
  7: PERF_RECORD_* events & perf_sample fields                       : Skip (permissions)
```
Fixes 16d00fee7038 ("perf tests: Move test__PERF_RECORD into separate object")
---
 tools/perf/tests/perf-record.c | 4 ++++
 1 file changed, 4 insertions(+)
tools/perf/tests/perf-record.c +4 -0
index 0b3c37e66871..8c79b5166a05 100644
@@ -115,6 +115,7 @@ static int test__PERF_RECORD(struct test_suite *test __maybe_unused, int subtest
 	if (err < 0) {
 		pr_debug("sched__get_first_possible_cpu: %s\n",
 			 str_error_r(errno, sbuf, sizeof(sbuf)));
+		evlist__cancel_workload(evlist);
 		goto out_delete_evlist;
 	}
 
@@ -126,6 +127,7 @@ static int test__PERF_RECORD(struct test_suite *test __maybe_unused, int subtest
 	if (sched_setaffinity(evlist->workload.pid, cpu_mask_size, &cpu_mask) < 0) {
 		pr_debug("sched_setaffinity: %s\n",
 			 str_error_r(errno, sbuf, sizeof(sbuf)));
+		evlist__cancel_workload(evlist);
 		goto out_delete_evlist;
 	}
 
@@ -137,6 +139,7 @@ static int test__PERF_RECORD(struct test_suite *test __maybe_unused, int subtest
 	if (err < 0) {
 		pr_debug("perf_evlist__open: %s\n",
 			 str_error_r(errno, sbuf, sizeof(sbuf)));
+		evlist__cancel_workload(evlist);
 		goto out_delete_evlist;
 	}
 
@@ -149,6 +152,7 @@ static int test__PERF_RECORD(struct test_suite *test __maybe_unused, int subtest
 	if (err < 0) {
 		pr_debug("evlist__mmap: %s\n",
 			 str_error_r(errno, sbuf, sizeof(sbuf)));
+		evlist__cancel_workload(evlist);
 		goto out_delete_evlist;
 	}
 
Signature · 1 line
2.51.0.rc2.233.g662b1ed5c5-goog
author reviewer
#5 lore raw

Re: [PATCH v1 2/4] perf test: Don't leak workload gopipe in PERF_RECORD_*

Quoted 29 lines · Arnaldo Carvalho de Melo
Quoted 11 lines · Ian Rogers

The test starts a workload and then opens events. If the events fail to open, for example because of perf_event_paranoid, the gopipe of the workload is leaked and the file descriptor leak check fails when the test exits. To avoid this cancel the workload when opening the events fails.

Before:
```

$ perf test -vv 7
  7: PERF_RECORD_* events & perf_sample fields:
--- start ---

⬢ [acme@toolbx perf-tools-next]$ patch -p1 < b
patching file tools/perf/tests/perf-record.c
Hunk #1 succeeded at 130 (offset 15 lines).
Hunk #2 succeeded at 142 with fuzz 1 (offset 15 lines).
Hunk #3 succeeded at 154 (offset 15 lines).
Hunk #4 succeeded at 167 (offset 15 lines).
⬢ [acme@toolbx perf-tools-next]$
⬢ [acme@toolbx perf-tools-next]$ git log --oneline -5 tools/perf/tests/perf-record.c
576bd7a8c90c48e9 (x1/perf-tools-next, x1/HEAD, five/perf-tools-next, five/HEAD) perf tests record: Update testcase to fix usage of affinity for machines with #CPUs > 1K b4c658d4d63d6149 perf target: Remove uid from target
dc6d2bc2d893a878 perf sample: Make user_regs and intr_regs optional
fd8d5a3b076c033f perf tests: Add missing event.h include
9823147da6c893d9 perf tools: Move 'struct perf_sample' to a separate header file to disentangle headers
⬢ [acme@toolbx perf-tools-next]$

Quoted 4 lines

Can you please check that it is still ok?

I processed the first in the series and now I'm going thru the other two.

Thanks Arnaldo! I'm not seeing the patch on:
https://web.git.kernel.org/pub/scm/linux/kernel/git/perf/perf-tools-next.git/log/?h=tmp.perf-tools-next
I'm happy to check.

Ian

Quoted 169 lines

- Arnaldo

Quoted 167 lines

test child forked, pid 1189568
Using CPUID GenuineIntel-6-B7-1
------------------------------------------------------------
perf_event_attr:

  type                             0 (PERF_TYPE_HARDWARE)
  config                           0xa00000000 (cpu_atom/PERF_COUNT_HW_CPU_CYCLES/)
  disabled                         1

------------------------------------------------------------
sys_perf_event_open: pid 0 cpu -1 group_fd -1 flags 0x8 sys_perf_event_open failed, error -13
------------------------------------------------------------
perf_event_attr:

  type                             0 (PERF_TYPE_HARDWARE)
  config                           0xa00000000 (cpu_atom/PERF_COUNT_HW_CPU_CYCLES/)
  disabled                         1
  exclude_kernel                   1

------------------------------------------------------------
sys_perf_event_open: pid 0 cpu -1 group_fd -1 flags 0x8 = 3
------------------------------------------------------------
perf_event_attr:

  type                             0 (PERF_TYPE_HARDWARE)
  config                           0x400000000 (cpu_core/PERF_COUNT_HW_CPU_CYCLES/)
  disabled                         1

------------------------------------------------------------
sys_perf_event_open: pid 0 cpu -1 group_fd -1 flags 0x8 sys_perf_event_open failed, error -13
------------------------------------------------------------
perf_event_attr:

  type                             0 (PERF_TYPE_HARDWARE)
  config                           0x400000000 (cpu_core/PERF_COUNT_HW_CPU_CYCLES/)
  disabled                         1
  exclude_kernel                   1

------------------------------------------------------------
sys_perf_event_open: pid 0 cpu -1 group_fd -1 flags 0x8 = 3 Attempt to add: software/cpu-clock/
..after resolving event: software/config=0/ cpu-clock -> software/cpu-clock/
------------------------------------------------------------
perf_event_attr:

  type                             1 (PERF_TYPE_SOFTWARE)
  size                             136
  config                           0x9 (PERF_COUNT_SW_DUMMY)
  sample_type                      IP|TID|TIME|CPU
  read_format                      ID|LOST
  disabled                         1
  inherit                          1
  mmap                             1
  comm                             1
  enable_on_exec                   1
  task                             1
  sample_id_all                    1
  mmap2                            1
  comm_exec                        1
  ksymbol                          1
  bpf_event                        1

{ wakeup_events, wakeup_watermark } 1
------------------------------------------------------------
sys_perf_event_open: pid 1189569 cpu 0 group_fd -1 flags 0x8 sys_perf_event_open failed, error -13
perf_evlist__open: Permission denied
---- end(-2) ----
Leak of file descriptor 6 that opened: 'pipe:[14200347]'

---- unexpected signal (6) ----
Failed to read build ID for //anon
Failed to read build ID for //anon
Failed to read build ID for //anon
Failed to read build ID for //anon
Failed to read build ID for //anon
Failed to read build ID for //anon
Failed to read build ID for //anon
Failed to read build ID for //anon
Failed to read build ID for //anon
Failed to read build ID for //anon
Failed to read build ID for //anon
Failed to read build ID for //anon
Failed to read build ID for //anon
Failed to read build ID for //anon
Failed to read build ID for //anon
Failed to read build ID for //anon
Failed to read build ID for //anon
Failed to read build ID for //anon
Failed to read build ID for //anon
Failed to read build ID for //anon
Failed to read build ID for //anon
Failed to read build ID for //anon
Failed to read build ID for //anon
Failed to read build ID for //anon
Failed to read build ID for //anon
Failed to read build ID for //anon
Failed to read build ID for //anon
Failed to read build ID for //anon
Failed to read build ID for //anon
Failed to read build ID for //anon
Failed to read build ID for //anon
Failed to read build ID for //anon

    #0 0x565358f6666e in child_test_sig_handler builtin-test.c:311
    #1 0x7f29ce849df0 in __restore_rt libc_sigaction.c:0
    #2 0x7f29ce89e95c in __pthread_kill_implementation pthread_kill.c:44
    #3 0x7f29ce849cc2 in raise raise.c:27
    #4 0x7f29ce8324ac in abort abort.c:81
    #5 0x565358f662d4 in check_leaks builtin-test.c:226
    #6 0x565358f6682e in run_test_child builtin-test.c:344
    #7 0x565358ef7121 in start_command run-command.c:128
    #8 0x565358f67273 in start_test builtin-test.c:545
    #9 0x565358f6771d in __cmd_test builtin-test.c:647
    #10 0x565358f682bd in cmd_test builtin-test.c:849
    #11 0x565358ee5ded in run_builtin perf.c:349
    #12 0x565358ee6085 in handle_internal_command perf.c:401
    #13 0x565358ee61de in run_argv perf.c:448
    #14 0x565358ee6527 in main perf.c:555
    #15 0x7f29ce833ca8 in __libc_start_call_main libc_start_call_main.h:74
    #16 0x7f29ce833d65 in __libc_start_main@@GLIBC_2.34 libc-start.c:128
    #17 0x565358e391c1 in _start perf[851c1]
  7: PERF_RECORD_* events & perf_sample fields                       : FAILED!
```

After:
```
$ perf test 7
  7: PERF_RECORD_* events & perf_sample fields                       : Skip (permissions)
```
Fixes 16d00fee7038 ("perf tests: Move test__PERF_RECORD into separate object")
---
 tools/perf/tests/perf-record.c | 4 ++++
 1 file changed, 4 insertions(+)
tools/perf/tests/perf-record.c +4 -1
index 0b3c37e66871..8c79b5166a05 100644
@@ -115,6 +115,7 @@ static int test__PERF_RECORD(struct test_suite *test __maybe_unused, int subtest
      if (err < 0) {
              pr_debug("sched__get_first_possible_cpu: %s\n",
                       str_error_r(errno, sbuf, sizeof(sbuf)));
+             evlist__cancel_workload(evlist);
              goto out_delete_evlist;
      }
 
@@ -126,6 +127,7 @@ static int test__PERF_RECORD(struct test_suite *test __maybe_unused, int subtest
      if (sched_setaffinity(evlist->workload.pid, cpu_mask_size, &cpu_mask) < 0) {
              pr_debug("sched_setaffinity: %s\n",
                       str_error_r(errno, sbuf, sizeof(sbuf)));
+             evlist__cancel_workload(evlist);
              goto out_delete_evlist;
      }
 
@@ -137,6 +139,7 @@ static int test__PERF_RECORD(struct test_suite *test __maybe_unused, int subtest
      if (err < 0) {
              pr_debug("perf_evlist__open: %s\n",
                       str_error_r(errno, sbuf, sizeof(sbuf)));
+             evlist__cancel_workload(evlist);
              goto out_delete_evlist;
      }
 
@@ -149,6 +152,7 @@ static int test__PERF_RECORD(struct test_suite *test __maybe_unused, int subtest
      if (err < 0) {
              pr_debug("evlist__mmap: %s\n",
                       str_error_r(errno, sbuf, sizeof(sbuf)));
+             evlist__cancel_workload(evlist);
              goto out_delete_evlist;
      }
 
--

2.51.0.rc2.233.g662b1ed5c5-goog

#6 lore raw

Re: [PATCH v1 2/4] perf test: Don't leak workload gopipe in PERF_RECORD_*

Quoted 29 lines · Ian Rogers
Quoted 27 lines · Arnaldo Carvalho de Melo
Quoted 11 lines · Ian Rogers

The test starts a workload and then opens events. If the events fail to open, for example because of perf_event_paranoid, the gopipe of the workload is leaked and the file descriptor leak check fails when the test exits. To avoid this cancel the workload when opening the events fails.

Before:
```

$ perf test -vv 7
  7: PERF_RECORD_* events & perf_sample fields:
--- start ---

⬢ [acme@toolbx perf-tools-next]$ patch -p1 < b
patching file tools/perf/tests/perf-record.c
Hunk #1 succeeded at 130 (offset 15 lines).
Hunk #2 succeeded at 142 with fuzz 1 (offset 15 lines).
Hunk #3 succeeded at 154 (offset 15 lines).
Hunk #4 succeeded at 167 (offset 15 lines).
⬢ [acme@toolbx perf-tools-next]$
⬢ [acme@toolbx perf-tools-next]$ git log --oneline -5 tools/perf/tests/perf-record.c
576bd7a8c90c48e9 (x1/perf-tools-next, x1/HEAD, five/perf-tools-next, five/HEAD) perf tests record: Update testcase to fix usage of affinity for machines with #CPUs > 1K b4c658d4d63d6149 perf target: Remove uid from target
dc6d2bc2d893a878 perf sample: Make user_regs and intr_regs optional
fd8d5a3b076c033f perf tests: Add missing event.h include
9823147da6c893d9 perf tools: Move 'struct perf_sample' to a separate header file to disentangle headers
⬢ [acme@toolbx perf-tools-next]$

Quoted 7 lines
Quoted 4 lines

Can you please check that it is still ok?

I processed the first in the series and now I'm going thru the other two.

Thanks Arnaldo! I'm not seeing the patch on:
https://web.git.kernel.org/pub/scm/linux/kernel/git/perf/perf-tools-next.git/log/?h=tmp.perf-tools-next

Sorry, I thought I had pushed it earlier, should be there now.

Quoted 1 line

I'm happy to check.

Thanks!

- Arnaldo

author reviewer
#7 lore raw

Re: [PATCH v1 2/4] perf test: Don't leak workload gopipe in PERF_RECORD_*

Quoted 36 lines · Arnaldo Carvalho de Melo
Quoted 33 lines · Ian Rogers
Quoted 32 lines · Arnaldo Carvalho de Melo
Quoted 11 lines · Ian Rogers

The test starts a workload and then opens events. If the events fail to open, for example because of perf_event_paranoid, the gopipe of the workload is leaked and the file descriptor leak check fails when the test exits. To avoid this cancel the workload when opening the events fails.

Before:
```

$ perf test -vv 7
  7: PERF_RECORD_* events & perf_sample fields:
--- start ---

⬢ [acme@toolbx perf-tools-next]$ patch -p1 < b
patching file tools/perf/tests/perf-record.c
Hunk #1 succeeded at 130 (offset 15 lines).
Hunk #2 succeeded at 142 with fuzz 1 (offset 15 lines).
Hunk #3 succeeded at 154 (offset 15 lines).
Hunk #4 succeeded at 167 (offset 15 lines).
⬢ [acme@toolbx perf-tools-next]$
⬢ [acme@toolbx perf-tools-next]$ git log --oneline -5 tools/perf/tests/perf-record.c
576bd7a8c90c48e9 (x1/perf-tools-next, x1/HEAD, five/perf-tools-next, five/HEAD) perf tests record: Update testcase to fix usage of affinity for machines with #CPUs > 1K b4c658d4d63d6149 perf target: Remove uid from target
dc6d2bc2d893a878 perf sample: Make user_regs and intr_regs optional
fd8d5a3b076c033f perf tests: Add missing event.h include
9823147da6c893d9 perf tools: Move 'struct perf_sample' to a separate header file to disentangle headers
⬢ [acme@toolbx perf-tools-next]$

Can you please check that it is still ok?

I processed the first in the series and now I'm going thru the other two.

Quoted 8 lines
Quoted 2 lines

Thanks Arnaldo! I'm not seeing the patch on:
https://web.git.kernel.org/pub/scm/linux/kernel/git/perf/perf-tools-next.git/log/?h=tmp.perf-tools-next

Sorry, I thought I had pushed it earlier, should be there now.

Quoted 1 line

I'm happy to check.

Thanks!

Not seeing it, but I see the first patch. In any case I'll rebase and send as a v2.

Thanks,
Ian

Quoted 1 line

- Arnaldo

author reviewer
#8 lore raw

PATCH 3/4 perf evsel: Fix uniquification when PMU given without suffix

In reply to Ian Rogers

The PMU name is appearing twice in:

```
$ perf stat -e uncore_imc_free_running/data_total/ -A true

 Performance counter stats for 'system wide':

CPU0                 1.57 MiB  uncore_imc_free_running_0/uncore_imc_free_running,data_total/
CPU0                 1.58 MiB  uncore_imc_free_running_1/uncore_imc_free_running,data_total/
       0.000892376 seconds time elapsed
```

Use the pmu_name_len_no_suffix to avoid this problem.

Fixes 7d45f402d311 ("perf evlist: Make uniquifying counter names consistent")
---
 tools/perf/util/evsel.c | 28 ++++++++++++++++++----------
 1 file changed, 18 insertions(+), 10 deletions(-)
tools/perf/util/evsel.c +18 -10
index d264c143b592..9a6e2d556d35 100644
@@ -4048,9 +4048,9 @@ bool evsel__set_needs_uniquify(struct evsel *counter, const struct perf_stat_con
 
 void evsel__uniquify_counter(struct evsel *counter)
 {
-	const char *name, *pmu_name;
-	char *new_name, *config;
-	int ret;
+	const char *name, *pmu_name, *config;
+	char *new_name;
+	int len, ret;
 
 	/* No uniquification necessary. */
 	if (!counter->needs_uniquify)
@@ -4064,15 +4064,23 @@ void evsel__uniquify_counter(struct evsel *counter)
 	counter->uniquified_name = true;
 
 	name = evsel__name(counter);
+	config = strchr(name, '/');
 	pmu_name = counter->pmu->name;
-	/* Already prefixed by the PMU name. */
-	if (!strncmp(name, pmu_name, strlen(pmu_name)))
-		return;
 
-	config = strchr(name, '/');
-	if (config) {
-		int len = config - name;
+	/* Already prefixed by the PMU name? */
+	len = pmu_name_len_no_suffix(pmu_name);
+
+	if (!strncmp(name, pmu_name, len)) {
+		/*
+		 * If the PMU name is there, then there is no sense in not
+		 * having a slash. Do this for robustness.
+		 */
+		if (config == NULL)
+			config = name - 1;
 
+		ret = asprintf(&new_name, "%s/%s", pmu_name, config + 1);
+	} else if (config) {
+		len = config - name;
 		if (config[1] == '/') {
 			/* case: event// */
 			ret = asprintf(&new_name, "%s/%.*s/%s", pmu_name, len, name, config + 2);
@@ -4084,7 +4092,7 @@ void evsel__uniquify_counter(struct evsel *counter)
 		config = strchr(name, ':');
 		if (config) {
 			/* case: event:.. */
-			int len = config - name;
+			len = config - name;
 
 			ret = asprintf(&new_name, "%s/%.*s/%s", pmu_name, len, name, config + 1);
 		} else {
Signature · 1 line
2.51.0.rc2.233.g662b1ed5c5-goog
author reviewer
#9 lore raw

PATCH 4/4 perf test: Avoid uncore_imc/clockticks in uniquification test

In reply to Ian Rogers

The detection of uncore_imc may happen for free running PMUs and the clockticks event may be present on uncore_clock. Rewrite the test to detect duplicated/deduplicated events from perf list, not hardcoded to uncore_imc.

Fixes 070b315333ee ("perf test: Restrict uniquifying test to machines with 'uncore_imc'")
---
 .../tests/shell/stat+event_uniquifying.sh     | 105 ++++++++----------
 1 file changed, 44 insertions(+), 61 deletions(-)
tools/perf/tests/shell/stat+event_uniquifying.sh +44 -61
index bf54bd6c3e2e..9bc7a1f520f9 100755
@@ -4,74 +4,57 @@
 
 set -e
 
-stat_output=$(mktemp /tmp/__perf_test.stat_output.XXXXX)
-perf_tool=perf
 err=0
+stat_output=$(mktemp /tmp/__perf_test.stat_output.XXXXX)
 
-test_event_uniquifying() {
-  # We use `clockticks` in `uncore_imc` to verify the uniquify behavior.
-  pmu="uncore_imc"
-  event="clockticks"
-
-  # If the `-A` option is added, the event should be uniquified.
-  #
-  # $perf list -v clockticks
-  #
-  # List of pre-defined events (to be used in -e or -M):
-  #
-  #   uncore_imc_0/clockticks/                           [Kernel PMU event]
-  #   uncore_imc_1/clockticks/                           [Kernel PMU event]
-  #   uncore_imc_2/clockticks/                           [Kernel PMU event]
-  #   uncore_imc_3/clockticks/                           [Kernel PMU event]
-  #   uncore_imc_4/clockticks/                           [Kernel PMU event]
-  #   uncore_imc_5/clockticks/                           [Kernel PMU event]
-  #
-  #   ...
-  #
-  # $perf stat -e clockticks -A -- true
-  #
-  #  Performance counter stats for 'system wide':
-  #
-  # CPU0            3,773,018      uncore_imc_0/clockticks/
-  # CPU0            3,609,025      uncore_imc_1/clockticks/
-  # CPU0                    0      uncore_imc_2/clockticks/
-  # CPU0            3,230,009      uncore_imc_3/clockticks/
-  # CPU0            3,049,897      uncore_imc_4/clockticks/
-  # CPU0                    0      uncore_imc_5/clockticks/
-  #
-  #        0.002029828 seconds time elapsed
-
-  echo "stat event uniquifying test"
-  uniquified_event_array=()
-
-  # Skip if the machine does not have `uncore_imc` device.
-  if ! ${perf_tool} list pmu | grep -q ${pmu}; then
-    echo "Target does not support PMU ${pmu} [Skipped]"
-    err=2
-    return
-  fi
+cleanup() {
+  rm -f "${stat_output}"
 
-  # Check how many uniquified events.
-  while IFS= read -r line; do
-    uniquified_event=$(echo "$line" | awk '{print $1}')
-    uniquified_event_array+=("${uniquified_event}")
-  done < <(${perf_tool} list -v ${event} | grep ${pmu})
+  trap - EXIT TERM INT
+}
 
-  perf_command="${perf_tool} stat -e $event -A -o ${stat_output} -- true"
-  $perf_command
+trap_cleanup() {
+  echo "Unexpected signal in ${FUNCNAME[1]}"
+  cleanup
+  exit 1
+}
+trap trap_cleanup EXIT TERM INT
 
-  # Check the output contains all uniquified events.
-  for uniquified_event in "${uniquified_event_array[@]}"; do
-    if ! cat "${stat_output}" | grep -q "${uniquified_event}"; then
-      echo "Event is not uniquified [Failed]"
-      echo "${perf_command}"
-      cat "${stat_output}"
-      err=1
-      break
-    fi
+test_event_uniquifying() {
+  echo "Uniquification of PMU sysfs events test"
+
+  # Read events from perf list with and without -v. With -v the duplicate PMUs
+  # aren't deduplicated. Note, json events are listed by perf list without a
+  # PMU.
+  read -ra pmu_events <<< "$(perf list --raw pmu)"
+  read -ra pmu_v_events <<< "$(perf list -v --raw pmu)"
+  # For all non-deduplicated events.
+  for pmu_v_event in "${pmu_v_events[@]}"; do
+    # If the event matches an event in the deduplicated events then it musn't
+    # be an event with duplicate PMUs, continue the outer loop.
+    for pmu_event in "${pmu_events[@]}"; do
+      if [[ "$pmu_v_event" == "$pmu_event" ]]; then
+        continue 2
+      fi
+    done
+    # Strip the suffix from the non-deduplicated event's PMU.
+    event=$(echo "$pmu_v_event" | sed -E 's/_[0-9]+//')
+    for pmu_event in "${pmu_events[@]}"; do
+      if [[ "$event" == "$pmu_event" ]]; then
+        echo "Testing event ${event} is uniquified to ${pmu_v_event}"
+        perf stat -e "$event" -A -o ${stat_output} -- true
+        # Ensure the non-deduplicated event appears in the output.
+        if ! grep -q "${pmu_v_event}" "${stat_output}"; then
+          echo "Uniquification of PMU sysfs events test [Failed]"
+          cat "${stat_output}"
+          err=1
+        fi
+        break
+      fi
+    done
   done
 }
 
 test_event_uniquifying
-rm -f "${stat_output}"
+cleanup
 exit $err
Signature · 1 line
2.51.0.rc2.233.g662b1ed5c5-goog
#10 lore raw

Re: [PATCH v1 4/4] perf test: Avoid uncore_imc/clockticks in uniquification test

Quoted 4 lines · Ian Rogers

The detection of uncore_imc may happen for free running PMUs and the clockticks event may be present on uncore_clock. Rewrite the test to detect duplicated/deduplicated events from perf list, not hardcoded to uncore_imc.

acme@x1:~/git/perf-tools-next$ perf test 96
 96: perf stat events uniquifying                                    : FAILED!
acme@x1:~/git/perf-tools-next$ perf test -vv 96
Couldn't bump rlimit(MEMLOCK), failures may take place when creating BPF maps, etc
 96: perf stat events uniquifying:
--- start ---
test child forked, pid 86847
Uniquification of PMU sysfs events test
Testing event uncore_imc_free_running/data_read/ is uniquified to uncore_imc_free_running_0/data_read/
Error:
The sys_perf_event_open() syscall returned with 22 (Invalid argument) for event (uncore_imc_free_running/data_read/u).
"dmesg | grep -i perf" may provide additional information.

Unexpected signal in test_event_uniquifying
---- end(-1) ----
96: perf stat events uniquifying : FAILED!

Works for root:

acme@x1:~/git/perf-tools-next$ sudo su -
[sudo] password for acme: 
root@x1:~# perf test 96
 96: perf stat events uniquifying                                    : Ok
root@x1:~#
acme@x1:~/git/perf-tools-next$ ls -la /sys/bus/event_source/devices/uncore_imc_free_running_*/events/
/sys/bus/event_source/devices/uncore_imc_free_running_0/events/:
total 0
drwxr-xr-x. 2 root root    0 Sep 16 16:09 .
drwxr-xr-x. 5 root root    0 Sep 16 16:09 ..
-r--r--r--. 1 root root 4096 Sep 16 16:09 data_read
-r--r--r--. 1 root root 4096 Sep 16 16:09 data_read.scale
-r--r--r--. 1 root root 4096 Sep 16 16:09 data_read.unit
-r--r--r--. 1 root root 4096 Sep 16 16:09 data_total
-r--r--r--. 1 root root 4096 Sep 16 16:09 data_total.scale
-r--r--r--. 1 root root 4096 Sep 16 16:09 data_total.unit
-r--r--r--. 1 root root 4096 Sep 16 16:09 data_write
-r--r--r--. 1 root root 4096 Sep 16 16:09 data_write.scale
-r--r--r--. 1 root root 4096 Sep 16 16:09 data_write.unit

/sys/bus/event_source/devices/uncore_imc_free_running_1/events/:
total 0
drwxr-xr-x. 2 root root 0 Sep 16 16:09 .
drwxr-xr-x. 5 root root 0 Sep 16 16:09 ..
-r--r--r--. 1 root root 4096 Sep 16 16:09 data_read -r--r--r--. 1 root root 4096 Sep 16 16:09 data_read.scale -r--r--r--. 1 root root 4096 Sep 16 16:09 data_read.unit -r--r--r--. 1 root root 4096 Sep 16 16:09 data_total -r--r--r--. 1 root root 4096 Sep 16 16:09 data_total.scale -r--r--r--. 1 root root 4096 Sep 16 16:09 data_total.unit -r--r--r--. 1 root root 4096 Sep 16 16:09 data_write -r--r--r--. 1 root root 4096 Sep 16 16:09 data_write.scale -r--r--r--. 1 root root 4096 Sep 16 16:09 data_write.unit acme@x1:~/git/perf-tools-next$

Quoted 131 lines
Fixes 070b315333ee ("perf test: Restrict uniquifying test to machines with 'uncore_imc'")
---
 .../tests/shell/stat+event_uniquifying.sh     | 105 ++++++++----------
 1 file changed, 44 insertions(+), 61 deletions(-)
tools/perf/tests/shell/stat+event_uniquifying.sh +44 -61
index bf54bd6c3e2e..9bc7a1f520f9 100755
@@ -4,74 +4,57 @@
 
 set -e
 
-stat_output=$(mktemp /tmp/__perf_test.stat_output.XXXXX)
-perf_tool=perf
 err=0
+stat_output=$(mktemp /tmp/__perf_test.stat_output.XXXXX)
 
-test_event_uniquifying() {
-  # We use `clockticks` in `uncore_imc` to verify the uniquify behavior.
-  pmu="uncore_imc"
-  event="clockticks"
-
-  # If the `-A` option is added, the event should be uniquified.
-  #
-  # $perf list -v clockticks
-  #
-  # List of pre-defined events (to be used in -e or -M):
-  #
-  #   uncore_imc_0/clockticks/                           [Kernel PMU event]
-  #   uncore_imc_1/clockticks/                           [Kernel PMU event]
-  #   uncore_imc_2/clockticks/                           [Kernel PMU event]
-  #   uncore_imc_3/clockticks/                           [Kernel PMU event]
-  #   uncore_imc_4/clockticks/                           [Kernel PMU event]
-  #   uncore_imc_5/clockticks/                           [Kernel PMU event]
-  #
-  #   ...
-  #
-  # $perf stat -e clockticks -A -- true
-  #
-  #  Performance counter stats for 'system wide':
-  #
-  # CPU0            3,773,018      uncore_imc_0/clockticks/
-  # CPU0            3,609,025      uncore_imc_1/clockticks/
-  # CPU0                    0      uncore_imc_2/clockticks/
-  # CPU0            3,230,009      uncore_imc_3/clockticks/
-  # CPU0            3,049,897      uncore_imc_4/clockticks/
-  # CPU0                    0      uncore_imc_5/clockticks/
-  #
-  #        0.002029828 seconds time elapsed
-
-  echo "stat event uniquifying test"
-  uniquified_event_array=()
-
-  # Skip if the machine does not have `uncore_imc` device.
-  if ! ${perf_tool} list pmu | grep -q ${pmu}; then
-    echo "Target does not support PMU ${pmu} [Skipped]"
-    err=2
-    return
-  fi
+cleanup() {
+  rm -f "${stat_output}"
 
-  # Check how many uniquified events.
-  while IFS= read -r line; do
-    uniquified_event=$(echo "$line" | awk '{print $1}')
-    uniquified_event_array+=("${uniquified_event}")
-  done < <(${perf_tool} list -v ${event} | grep ${pmu})
+  trap - EXIT TERM INT
+}
 
-  perf_command="${perf_tool} stat -e $event -A -o ${stat_output} -- true"
-  $perf_command
+trap_cleanup() {
+  echo "Unexpected signal in ${FUNCNAME[1]}"
+  cleanup
+  exit 1
+}
+trap trap_cleanup EXIT TERM INT
 
-  # Check the output contains all uniquified events.
-  for uniquified_event in "${uniquified_event_array[@]}"; do
-    if ! cat "${stat_output}" | grep -q "${uniquified_event}"; then
-      echo "Event is not uniquified [Failed]"
-      echo "${perf_command}"
-      cat "${stat_output}"
-      err=1
-      break
-    fi
+test_event_uniquifying() {
+  echo "Uniquification of PMU sysfs events test"
+
+  # Read events from perf list with and without -v. With -v the duplicate PMUs
+  # aren't deduplicated. Note, json events are listed by perf list without a
+  # PMU.
+  read -ra pmu_events <<< "$(perf list --raw pmu)"
+  read -ra pmu_v_events <<< "$(perf list -v --raw pmu)"
+  # For all non-deduplicated events.
+  for pmu_v_event in "${pmu_v_events[@]}"; do
+    # If the event matches an event in the deduplicated events then it musn't
+    # be an event with duplicate PMUs, continue the outer loop.
+    for pmu_event in "${pmu_events[@]}"; do
+      if [[ "$pmu_v_event" == "$pmu_event" ]]; then
+        continue 2
+      fi
+    done
+    # Strip the suffix from the non-deduplicated event's PMU.
+    event=$(echo "$pmu_v_event" | sed -E 's/_[0-9]+//')
+    for pmu_event in "${pmu_events[@]}"; do
+      if [[ "$event" == "$pmu_event" ]]; then
+        echo "Testing event ${event} is uniquified to ${pmu_v_event}"
+        perf stat -e "$event" -A -o ${stat_output} -- true
+        # Ensure the non-deduplicated event appears in the output.
+        if ! grep -q "${pmu_v_event}" "${stat_output}"; then
+          echo "Uniquification of PMU sysfs events test [Failed]"
+          cat "${stat_output}"
+          err=1
+        fi
+        break
+      fi
+    done
   done
 }
 
 test_event_uniquifying
-rm -f "${stat_output}"
+cleanup
 exit $err
Signature · 1 line
2.51.0.rc2.233.g662b1ed5c5-goog
author reviewer
#11 lore raw

Re: [PATCH v1 4/4] perf test: Avoid uncore_imc/clockticks in uniquification test

Quoted 61 lines · Arnaldo Carvalho de Melo
Quoted 4 lines · Ian Rogers

The detection of uncore_imc may happen for free running PMUs and the clockticks event may be present on uncore_clock. Rewrite the test to detect duplicated/deduplicated events from perf list, not hardcoded to uncore_imc.

acme@x1:~/git/perf-tools-next$ perf test 96
 96: perf stat events uniquifying                                    : FAILED!
acme@x1:~/git/perf-tools-next$ perf test -vv 96
Couldn't bump rlimit(MEMLOCK), failures may take place when creating BPF maps, etc
 96: perf stat events uniquifying:
--- start ---
test child forked, pid 86847
Uniquification of PMU sysfs events test
Testing event uncore_imc_free_running/data_read/ is uniquified to uncore_imc_free_running_0/data_read/
Error:
The sys_perf_event_open() syscall returned with 22 (Invalid argument) for event (uncore_imc_free_running/data_read/u).
"dmesg | grep -i perf" may provide additional information.

Unexpected signal in test_event_uniquifying
---- end(-1) ----
96: perf stat events uniquifying : FAILED!

Works for root:

acme@x1:~/git/perf-tools-next$ sudo su -
[sudo] password for acme:
root@x1:~# perf test 96
 96: perf stat events uniquifying                                    : Ok
root@x1:~#
acme@x1:~/git/perf-tools-next$ ls -la /sys/bus/event_source/devices/uncore_imc_free_running_*/events/
/sys/bus/event_source/devices/uncore_imc_free_running_0/events/:
total 0
drwxr-xr-x. 2 root root    0 Sep 16 16:09 .
drwxr-xr-x. 5 root root    0 Sep 16 16:09 ..
(quoted diff) +0 -9
-r--r--r--. 1 root root 4096 Sep 16 16:09 data_read
-r--r--r--. 1 root root 4096 Sep 16 16:09 data_read.scale
-r--r--r--. 1 root root 4096 Sep 16 16:09 data_read.unit
-r--r--r--. 1 root root 4096 Sep 16 16:09 data_total
-r--r--r--. 1 root root 4096 Sep 16 16:09 data_total.scale
-r--r--r--. 1 root root 4096 Sep 16 16:09 data_total.unit
-r--r--r--. 1 root root 4096 Sep 16 16:09 data_write
-r--r--r--. 1 root root 4096 Sep 16 16:09 data_write.scale
-r--r--r--. 1 root root 4096 Sep 16 16:09 data_write.unit

/sys/bus/event_source/devices/uncore_imc_free_running_1/events/:
total 0
drwxr-xr-x. 2 root root 0 Sep 16 16:09 .
drwxr-xr-x. 5 root root 0 Sep 16 16:09 ..

(quoted diff) +0 -9
-r--r--r--. 1 root root 4096 Sep 16 16:09 data_read
-r--r--r--. 1 root root 4096 Sep 16 16:09 data_read.scale
-r--r--r--. 1 root root 4096 Sep 16 16:09 data_read.unit
-r--r--r--. 1 root root 4096 Sep 16 16:09 data_total
-r--r--r--. 1 root root 4096 Sep 16 16:09 data_total.scale
-r--r--r--. 1 root root 4096 Sep 16 16:09 data_total.unit
-r--r--r--. 1 root root 4096 Sep 16 16:09 data_write
-r--r--r--. 1 root root 4096 Sep 16 16:09 data_write.scale
-r--r--r--. 1 root root 4096 Sep 16 16:09 data_write.unit
Quoted 1 line

acme@x1:~/git/perf-tools-next$

Looks like a permissions issue (opening an uncore event with user filter is a bad idea, but is tried when open fails and we do fallbacks). I'll do a v2 with a permission check fix and resolving the earlier merge conflict.

Thanks,
Ian

Quoted 131 lines
Quoted 131 lines
Fixes 070b315333ee ("perf test: Restrict uniquifying test to machines with 'uncore_imc'")
---
 .../tests/shell/stat+event_uniquifying.sh     | 105 ++++++++----------
 1 file changed, 44 insertions(+), 61 deletions(-)
tools/perf/tests/shell/stat+event_uniquifying.sh +44 -62
index bf54bd6c3e2e..9bc7a1f520f9 100755
@@ -4,74 +4,57 @@
 
 set -e
 
-stat_output=$(mktemp /tmp/__perf_test.stat_output.XXXXX)
-perf_tool=perf
 err=0
+stat_output=$(mktemp /tmp/__perf_test.stat_output.XXXXX)
 
-test_event_uniquifying() {
-  # We use `clockticks` in `uncore_imc` to verify the uniquify behavior.
-  pmu="uncore_imc"
-  event="clockticks"
-
-  # If the `-A` option is added, the event should be uniquified.
-  #
-  # $perf list -v clockticks
-  #
-  # List of pre-defined events (to be used in -e or -M):
-  #
-  #   uncore_imc_0/clockticks/                           [Kernel PMU event]
-  #   uncore_imc_1/clockticks/                           [Kernel PMU event]
-  #   uncore_imc_2/clockticks/                           [Kernel PMU event]
-  #   uncore_imc_3/clockticks/                           [Kernel PMU event]
-  #   uncore_imc_4/clockticks/                           [Kernel PMU event]
-  #   uncore_imc_5/clockticks/                           [Kernel PMU event]
-  #
-  #   ...
-  #
-  # $perf stat -e clockticks -A -- true
-  #
-  #  Performance counter stats for 'system wide':
-  #
-  # CPU0            3,773,018      uncore_imc_0/clockticks/
-  # CPU0            3,609,025      uncore_imc_1/clockticks/
-  # CPU0                    0      uncore_imc_2/clockticks/
-  # CPU0            3,230,009      uncore_imc_3/clockticks/
-  # CPU0            3,049,897      uncore_imc_4/clockticks/
-  # CPU0                    0      uncore_imc_5/clockticks/
-  #
-  #        0.002029828 seconds time elapsed
-
-  echo "stat event uniquifying test"
-  uniquified_event_array=()
-
-  # Skip if the machine does not have `uncore_imc` device.
-  if ! ${perf_tool} list pmu | grep -q ${pmu}; then
-    echo "Target does not support PMU ${pmu} [Skipped]"
-    err=2
-    return
-  fi
+cleanup() {
+  rm -f "${stat_output}"
 
-  # Check how many uniquified events.
-  while IFS= read -r line; do
-    uniquified_event=$(echo "$line" | awk '{print $1}')
-    uniquified_event_array+=("${uniquified_event}")
-  done < <(${perf_tool} list -v ${event} | grep ${pmu})
+  trap - EXIT TERM INT
+}
 
-  perf_command="${perf_tool} stat -e $event -A -o ${stat_output} -- true"
-  $perf_command
+trap_cleanup() {
+  echo "Unexpected signal in ${FUNCNAME[1]}"
+  cleanup
+  exit 1
+}
+trap trap_cleanup EXIT TERM INT
 
-  # Check the output contains all uniquified events.
-  for uniquified_event in "${uniquified_event_array[@]}"; do
-    if ! cat "${stat_output}" | grep -q "${uniquified_event}"; then
-      echo "Event is not uniquified [Failed]"
-      echo "${perf_command}"
-      cat "${stat_output}"
-      err=1
-      break
-    fi
+test_event_uniquifying() {
+  echo "Uniquification of PMU sysfs events test"
+
+  # Read events from perf list with and without -v. With -v the duplicate PMUs
+  # aren't deduplicated. Note, json events are listed by perf list without a
+  # PMU.
+  read -ra pmu_events <<< "$(perf list --raw pmu)"
+  read -ra pmu_v_events <<< "$(perf list -v --raw pmu)"
+  # For all non-deduplicated events.
+  for pmu_v_event in "${pmu_v_events[@]}"; do
+    # If the event matches an event in the deduplicated events then it musn't
+    # be an event with duplicate PMUs, continue the outer loop.
+    for pmu_event in "${pmu_events[@]}"; do
+      if [[ "$pmu_v_event" == "$pmu_event" ]]; then
+        continue 2
+      fi
+    done
+    # Strip the suffix from the non-deduplicated event's PMU.
+    event=$(echo "$pmu_v_event" | sed -E 's/_[0-9]+//')
+    for pmu_event in "${pmu_events[@]}"; do
+      if [[ "$event" == "$pmu_event" ]]; then
+        echo "Testing event ${event} is uniquified to ${pmu_v_event}"
+        perf stat -e "$event" -A -o ${stat_output} -- true
+        # Ensure the non-deduplicated event appears in the output.
+        if ! grep -q "${pmu_v_event}" "${stat_output}"; then
+          echo "Uniquification of PMU sysfs events test [Failed]"
+          cat "${stat_output}"
+          err=1
+        fi
+        break
+      fi
+    done
   done
 }
 
 test_event_uniquifying
-rm -f "${stat_output}"
+cleanup
 exit $err
--

2.51.0.rc2.233.g662b1ed5c5-goog