perf test: Add perf_event_attr tests for the arm_spe event
Subsystems (4)
PATCH 1/1 perf test: Add perf_event_attr tests for the arm_spe event
Adds a couple of perf_event_attr tests for the fix introduced in [1]. The tests check that the correct sample_period value is set in the struct perf_event_attr of the arm_spe events.
[1]: https://lore.kernel.org/all/[email protected]/
--- tools/perf/tests/attr/README | 2 + tools/perf/tests/attr/base-record-spe | 40 +++++++++++++++++++ tools/perf/tests/attr/test-record-spe-period | 12 ++++++ .../tests/attr/test-record-spe-period-term | 12 ++++++ 4 files changed, 66 insertions(+) create mode 100644 tools/perf/tests/attr/base-record-spe create mode 100644 tools/perf/tests/attr/test-record-spe-period create mode 100644 tools/perf/tests/attr/test-record-spe-period-term
tools/perf/tests/attr/README +2 -0
@@ -58,6 +58,8 @@ Following tests are defined (with perf commands): perf record -c 100 -P kill (test-record-period) perf record -c 1 --pfm-events=cycles:period=2 (test-record-pfm-period) perf record -R kill (test-record-raw) + perf record -c 2 -e arm_spe_0// -- kill (test-record-spe-period) + perf record -e arm_spe_0/period=3/ -- kill (test-record-spe-period-term) perf stat -e cycles kill (test-stat-basic) perf stat kill (test-stat-default) perf stat -d kill (test-stat-detailed-1)
tools/perf/tests/attr/base-record-spe new file +40 -0
@@ -0,0 +1,40 @@ +[event] +fd=* +group_fd=-1 +flags=* +cpu=* +type=* +size=* +config=* +sample_period=* +sample_type=* +read_format=* +disabled=* +inherit=* +pinned=* +exclusive=* +exclude_user=* +exclude_kernel=* +exclude_hv=* +exclude_idle=* +mmap=* +comm=* +freq=* +inherit_stat=* +enable_on_exec=* +task=* +watermark=* +precise_ip=* +mmap_data=* +sample_id_all=* +exclude_host=* +exclude_guest=* +exclude_callchain_kernel=* +exclude_callchain_user=* +wakeup_events=* +bp_type=* +config1=* +config2=* +branch_sample_type=* +sample_regs_user=* +sample_stack_user=*
tools/perf/tests/attr/test-record-spe-period new file +12 -0
@@ -0,0 +1,12 @@ +[config] +command = record +args = --no-bpf-event -c 2 -e arm_spe_0// -- kill >/dev/null 2>&1 +ret = 1 +arch = aarch64 + +[event-10:base-record-spe] +sample_period=2 +freq=0 + +# dummy event +[event-1:base-record-spe]
tools/perf/tests/attr/test-record-spe-period-term new file +12 -0
@@ -0,0 +1,12 @@ +[config] +command = record +args = --no-bpf-event -e arm_spe_0/period=3/ -- kill >/dev/null 2>&1 +ret = 1 +arch = aarch64 + +[event-10:base-record-spe] +sample_period=3 +freq=0 + +# dummy event +[event-1:base-record-spe]
Signature · 1 line
2.25.1
Hi German,
Quoted 7 lines · German Gomez
Adds a couple of
perf_event_attrtests for the fix introduced in [1]. The tests check that the correctsample_periodvalue is set in the structperf_event_attrof thearm_speevents.[1]: https://lore.kernel.org/all/[email protected]/
Signed-off-by German Gomez <[email protected]>
I tested this patch with two commands:
# PERF_TEST_ATTR=/tmp /usr/bin/python2 ./tests/attr.py -d ./tests/attr/ \
-p ./perf -vvvvv -t test-record-spe-period
# PERF_TEST_ATTR=/tmp /usr/bin/python2 ./tests/attr.py -d ./tests/attr/ \
-p ./perf -vvvvv -t test-record-spe-period-term
Both testing can pass on Hisilicon D06 board.
One question: I'm a bit concern this case will fail on some Arm64 platforms which doesn't contain Arm SPE modules. E.g. below commands will always fail on Arm64 platforms if SPE module is absent. So I am wandering if we can add extra checking ARM SPE event is existed or not?
# ./perf test list 17: Setup struct perf_event_attr # ./perf test 17
Thanks,
Leo
Quoted 108 lines
--- tools/perf/tests/attr/README | 2 + tools/perf/tests/attr/base-record-spe | 40 +++++++++++++++++++ tools/perf/tests/attr/test-record-spe-period | 12 ++++++ .../tests/attr/test-record-spe-period-term | 12 ++++++ 4 files changed, 66 insertions(+) create mode 100644 tools/perf/tests/attr/base-record-spe create mode 100644 tools/perf/tests/attr/test-record-spe-period create mode 100644 tools/perf/tests/attr/test-record-spe-period-termtools/perf/tests/attr/README +2
-0@@ -58,6 +58,8 @@ Following tests are defined (with perf commands): perf record -c 100 -P kill (test-record-period) perf record -c 1 --pfm-events=cycles:period=2 (test-record-pfm-period) perf record -R kill (test-record-raw) + perf record -c 2 -e arm_spe_0// -- kill (test-record-spe-period) + perf record -e arm_spe_0/period=3/ -- kill (test-record-spe-period-term) perf stat -e cycles kill (test-stat-basic) perf stat kill (test-stat-default) perf stat -d kill (test-stat-detailed-1)tools/perf/tests/attr/base-record-spe new file +40
-0@@ -0,0 +1,40 @@ +[event] +fd=* +group_fd=-1 +flags=* +cpu=* +type=* +size=* +config=* +sample_period=* +sample_type=* +read_format=* +disabled=* +inherit=* +pinned=* +exclusive=* +exclude_user=* +exclude_kernel=* +exclude_hv=* +exclude_idle=* +mmap=* +comm=* +freq=* +inherit_stat=* +enable_on_exec=* +task=* +watermark=* +precise_ip=* +mmap_data=* +sample_id_all=* +exclude_host=* +exclude_guest=* +exclude_callchain_kernel=* +exclude_callchain_user=* +wakeup_events=* +bp_type=* +config1=* +config2=* +branch_sample_type=* +sample_regs_user=* +sample_stack_user=*tools/perf/tests/attr/test-record-spe-period new file +12
-0@@ -0,0 +1,12 @@ +[config] +command = record +args = --no-bpf-event -c 2 -e arm_spe_0// -- kill >/dev/null 2>&1 +ret = 1 +arch = aarch64 + +[event-10:base-record-spe] +sample_period=2 +freq=0 + +# dummy event +[event-1:base-record-spe]tools/perf/tests/attr/test-record-spe-period-term new file +12
-0@@ -0,0 +1,12 @@ +[config] +command = record +args = --no-bpf-event -e arm_spe_0/period=3/ -- kill >/dev/null 2>&1 +ret = 1 +arch = aarch64 + +[event-10:base-record-spe] +sample_period=3 +freq=0 + +# dummy event +[event-1:base-record-spe]Signature · 1 line
2.25.1
Hi Leo, thanks for checking this
Quoted 17 lines · Leo Yan
Hi German,
Quoted 7 lines · German Gomez
Adds a couple of
perf_event_attrtests for the fix introduced in [1]. The tests check that the correctsample_periodvalue is set in the structperf_event_attrof thearm_speevents.[1]: https://lore.kernel.org/all/[email protected]/
Signed-off-by German Gomez <[email protected]>I tested this patch with two commands:
# PERF_TEST_ATTR=/tmp
/usr/bin/python2./tests/attr.py-d ./tests/attr/ \
-p ./perf -vvvvv -t test-record-spe-period
# PERF_TEST_ATTR=/tmp/usr/bin/python2./tests/attr.py-d ./tests/attr/ \
-p ./perf -vvvvv -t test-record-spe-period-term
Quoted 6 lines
Both testing can pass on Hisilicon D06 board.
One question: I'm a bit concern this case will fail on some Arm64 platforms which doesn't contain Arm SPE modules. E.g. below commands will always fail on Arm64 platforms if SPE module is absent. So I am wandering if we can add extra checking ARM SPE event is existed or not?
The test reports "unsupported" if the return code and the 'ret' field don't match.
When I unload the SPE module:
running './tests/attr//test-record-spe-period-term' test limitation 'aarch64'
unsupp './tests/attr//test-record-spe-period-term'
Quoted 116 lines
# ./perf test list 17: Setup struct perf_event_attr # ./perf test 17Thanks,
LeoQuoted 108 lines
--- tools/perf/tests/attr/README | 2 + tools/perf/tests/attr/base-record-spe | 40 +++++++++++++++++++ tools/perf/tests/attr/test-record-spe-period | 12 ++++++ .../tests/attr/test-record-spe-period-term | 12 ++++++ 4 files changed, 66 insertions(+) create mode 100644 tools/perf/tests/attr/base-record-spe create mode 100644 tools/perf/tests/attr/test-record-spe-period create mode 100644 tools/perf/tests/attr/test-record-spe-period-termtools/perf/tests/attr/README +2
-0@@ -58,6 +58,8 @@ Following tests are defined (with perf commands): perf record -c 100 -P kill (test-record-period) perf record -c 1 --pfm-events=cycles:period=2 (test-record-pfm-period) perf record -R kill (test-record-raw) + perf record -c 2 -e arm_spe_0// -- kill (test-record-spe-period) + perf record -e arm_spe_0/period=3/ -- kill (test-record-spe-period-term) perf stat -e cycles kill (test-stat-basic) perf stat kill (test-stat-default) perf stat -d kill (test-stat-detailed-1)tools/perf/tests/attr/base-record-spe new file +40
-0@@ -0,0 +1,40 @@ +[event] +fd=* +group_fd=-1 +flags=* +cpu=* +type=* +size=* +config=* +sample_period=* +sample_type=* +read_format=* +disabled=* +inherit=* +pinned=* +exclusive=* +exclude_user=* +exclude_kernel=* +exclude_hv=* +exclude_idle=* +mmap=* +comm=* +freq=* +inherit_stat=* +enable_on_exec=* +task=* +watermark=* +precise_ip=* +mmap_data=* +sample_id_all=* +exclude_host=* +exclude_guest=* +exclude_callchain_kernel=* +exclude_callchain_user=* +wakeup_events=* +bp_type=* +config1=* +config2=* +branch_sample_type=* +sample_regs_user=* +sample_stack_user=*tools/perf/tests/attr/test-record-spe-period new file +12
-0@@ -0,0 +1,12 @@ +[config] +command = record +args = --no-bpf-event -c 2 -e arm_spe_0// -- kill >/dev/null 2>&1 +ret = 1 +arch = aarch64 + +[event-10:base-record-spe] +sample_period=2 +freq=0 + +# dummy event +[event-1:base-record-spe]tools/perf/tests/attr/test-record-spe-period-term new file +12
-0@@ -0,0 +1,12 @@ +[config] +command = record +args = --no-bpf-event -e arm_spe_0/period=3/ -- kill >/dev/null 2>&1 +ret = 1 +arch = aarch64 + +[event-10:base-record-spe] +sample_period=3 +freq=0 + +# dummy event +[event-1:base-record-spe]Signature · 1 line
2.25.1
Leo Yan
Tested-byQuoted 26 lines · German Gomez
Hi Leo, thanks for checking this
Quoted 17 lines · Leo Yan
Hi German,
Quoted 7 lines · German Gomez
Adds a couple of
perf_event_attrtests for the fix introduced in [1]. The tests check that the correctsample_periodvalue is set in the structperf_event_attrof thearm_speevents.[1]: https://lore.kernel.org/all/[email protected]/
Signed-off-by German Gomez <[email protected]>I tested this patch with two commands:
# PERF_TEST_ATTR=/tmp
/usr/bin/python2./tests/attr.py-d ./tests/attr/ \
-p ./perf -vvvvv -t test-record-spe-period
# PERF_TEST_ATTR=/tmp/usr/bin/python2./tests/attr.py-d ./tests/attr/ \
-p ./perf -vvvvv -t test-record-spe-period-termQuoted 6 lines
Both testing can pass on Hisilicon D06 board.
One question: I'm a bit concern this case will fail on some Arm64 platforms which doesn't contain Arm SPE modules. E.g. below commands will always fail on Arm64 platforms if SPE module is absent. So I am wandering if we can add extra checking ARM SPE event is existed or not?
Quoted 8 lines
The test reports "unsupported" if the return code and the 'ret' field don't match.When I unload the SPE module:
running './tests/attr//test-record-spe-period-term' test limitation 'aarch64'
unsupp './tests/attr//test-record-spe-period-term'
Thanks for confirmation, German.
You could add my testing tag for this patch:
Quoted 6 lines · Leo Yan
[...]
Thanks for confirmation, German.You could add my testing tag for this patch:
Tested-by Leo Yan <[email protected]>
Thanks Leo
Quoted 7 lines · German Gomez
Adds a couple of
perf_event_attrtests for the fix introduced in [1]. The tests check that the correctsample_periodvalue is set in the structperf_event_attrof thearm_speevents.[1]: https://lore.kernel.org/all/[email protected]/
Signed-off-by German Gomez <[email protected]>
Quoted 105 lines
--- tools/perf/tests/attr/README | 2 + tools/perf/tests/attr/base-record-spe | 40 +++++++++++++++++++ tools/perf/tests/attr/test-record-spe-period | 12 ++++++ .../tests/attr/test-record-spe-period-term | 12 ++++++ 4 files changed, 66 insertions(+) create mode 100644 tools/perf/tests/attr/base-record-spe create mode 100644 tools/perf/tests/attr/test-record-spe-period create mode 100644 tools/perf/tests/attr/test-record-spe-period-termtools/perf/tests/attr/README +2
-0@@ -58,6 +58,8 @@ Following tests are defined (with perf commands): perf record -c 100 -P kill (test-record-period) perf record -c 1 --pfm-events=cycles:period=2 (test-record-pfm-period) perf record -R kill (test-record-raw) + perf record -c 2 -e arm_spe_0// -- kill (test-record-spe-period) + perf record -e arm_spe_0/period=3/ -- kill (test-record-spe-period-term) perf stat -e cycles kill (test-stat-basic) perf stat kill (test-stat-default) perf stat -d kill (test-stat-detailed-1)tools/perf/tests/attr/base-record-spe new file +40
-0@@ -0,0 +1,40 @@ +[event] +fd=* +group_fd=-1 +flags=* +cpu=* +type=* +size=* +config=* +sample_period=* +sample_type=* +read_format=* +disabled=* +inherit=* +pinned=* +exclusive=* +exclude_user=* +exclude_kernel=* +exclude_hv=* +exclude_idle=* +mmap=* +comm=* +freq=* +inherit_stat=* +enable_on_exec=* +task=* +watermark=* +precise_ip=* +mmap_data=* +sample_id_all=* +exclude_host=* +exclude_guest=* +exclude_callchain_kernel=* +exclude_callchain_user=* +wakeup_events=* +bp_type=* +config1=* +config2=* +branch_sample_type=* +sample_regs_user=* +sample_stack_user=*tools/perf/tests/attr/test-record-spe-period new file +12
-0@@ -0,0 +1,12 @@ +[config] +command = record +args = --no-bpf-event -c 2 -e arm_spe_0// -- kill >/dev/null 2>&1 +ret = 1 +arch = aarch64 + +[event-10:base-record-spe] +sample_period=2 +freq=0 + +# dummy event +[event-1:base-record-spe]tools/perf/tests/attr/test-record-spe-period-term new file +12
-0@@ -0,0 +1,12 @@ +[config] +command = record +args = --no-bpf-event -e arm_spe_0/period=3/ -- kill >/dev/null 2>&1 +ret = 1 +arch = aarch64 + +[event-10:base-record-spe] +sample_period=3 +freq=0 + +# dummy event +[event-1:base-record-spe]
Em Mon, Feb 28, 2022 at 03:05:21PM +0000, James Clark escreveu:
Quoted 7 lines · German Gomez
Quoted 4 lines
Adds a couple of
perf_event_attrtests for the fix introduced in [1]. The tests check that the correctsample_periodvalue is set in the structperf_event_attrof thearm_speevents.
Quoted 5 lines
Quoted 3 lines
[1]: https://lore.kernel.org/all/[email protected]/
Signed-off-by German Gomez <[email protected]>Reviewed-by James Clark <[email protected]>
Thanks, applied.
- Arnaldo
Quoted 105 lines
Quoted 105 lines
--- tools/perf/tests/attr/README | 2 + tools/perf/tests/attr/base-record-spe | 40 +++++++++++++++++++ tools/perf/tests/attr/test-record-spe-period | 12 ++++++ .../tests/attr/test-record-spe-period-term | 12 ++++++ 4 files changed, 66 insertions(+) create mode 100644 tools/perf/tests/attr/base-record-spe create mode 100644 tools/perf/tests/attr/test-record-spe-period create mode 100644 tools/perf/tests/attr/test-record-spe-period-termtools/perf/tests/attr/README +2
-0@@ -58,6 +58,8 @@ Following tests are defined (with perf commands): perf record -c 100 -P kill (test-record-period) perf record -c 1 --pfm-events=cycles:period=2 (test-record-pfm-period) perf record -R kill (test-record-raw) + perf record -c 2 -e arm_spe_0// -- kill (test-record-spe-period) + perf record -e arm_spe_0/period=3/ -- kill (test-record-spe-period-term) perf stat -e cycles kill (test-stat-basic) perf stat kill (test-stat-default) perf stat -d kill (test-stat-detailed-1)tools/perf/tests/attr/base-record-spe new file +40
-0@@ -0,0 +1,40 @@ +[event] +fd=* +group_fd=-1 +flags=* +cpu=* +type=* +size=* +config=* +sample_period=* +sample_type=* +read_format=* +disabled=* +inherit=* +pinned=* +exclusive=* +exclude_user=* +exclude_kernel=* +exclude_hv=* +exclude_idle=* +mmap=* +comm=* +freq=* +inherit_stat=* +enable_on_exec=* +task=* +watermark=* +precise_ip=* +mmap_data=* +sample_id_all=* +exclude_host=* +exclude_guest=* +exclude_callchain_kernel=* +exclude_callchain_user=* +wakeup_events=* +bp_type=* +config1=* +config2=* +branch_sample_type=* +sample_regs_user=* +sample_stack_user=*tools/perf/tests/attr/test-record-spe-period new file +12
-0@@ -0,0 +1,12 @@ +[config] +command = record +args = --no-bpf-event -c 2 -e arm_spe_0// -- kill >/dev/null 2>&1 +ret = 1 +arch = aarch64 + +[event-10:base-record-spe] +sample_period=2 +freq=0 + +# dummy event +[event-1:base-record-spe]tools/perf/tests/attr/test-record-spe-period-term new file +12
-0@@ -0,0 +1,12 @@ +[config] +command = record +args = --no-bpf-event -e arm_spe_0/period=3/ -- kill >/dev/null 2>&1 +ret = 1 +arch = aarch64 + +[event-10:base-record-spe] +sample_period=3 +freq=0 + +# dummy event +[event-1:base-record-spe]
Signature · 1 line
- Arnaldo