* [PATCH 0/3] perf: arm-spe: Fix addresses of synthesized Arm SPE events
@ 2022-04-21 16:52 Timothy Hayes
2022-04-21 16:52 ` [PATCH 1/3] perf: arm-spe: Fix addresses of synthesized " Timothy Hayes
` (2 more replies)
0 siblings, 3 replies; 11+ messages in thread
From: Timothy Hayes @ 2022-04-21 16:52 UTC (permalink / raw)
To: linux-kernel, linux-perf-users, acme
Cc: John Garry, Will Deacon, Mathieu Poirier, Leo Yan, Mark Rutland,
Alexander Shishkin, Jiri Olsa, Namhyung Kim, Martin KaFai Lau,
Song Liu, Yonghong Song, John Fastabend, KP Singh,
linux-arm-kernel, netdev, bpf
This patch set fixes problems related to address in synthesized events from SPE.
Committer testing:
perf record --no-bpf-event -e arm_spe_0/pa_enable=1/ -- sleep 1
perf inject -i perf.data -o perf-inj.data --itrace --strip
perf script -i perf-inj.data --fields hw:+addr,+phys_addr
Before:
sleep 49337 [004] 20133.731889: 1 l1d-access: ffffcbebcde1d5b8 [unknown] ([unknown])
sleep 49337 [004] 20133.731889: 1 tlb-access: ffffcbebcde1d5b8 [unknown] ([unknown])
sleep 49337 [004] 20133.731889: 1 memory: ffffcbebcde1d5b8 [unknown] ([unknown])
After:
sleep 49337 [004] 20133.731889: 1 l1d-access: ffff800034123970 ffffcbebcde1d5b8 [unknown] ([unknown]) 153d19970
sleep 49337 [004] 20133.731889: 1 tlb-access: ffff800034123970 ffffcbebcde1d5b8 [unknown] ([unknown]) 153d19970
sleep 49337 [004] 20133.731889: 1 memory: ffff800034123970 ffffcbebcde1d5b8 [unknown] ([unknown]) 153d19970
tools/perf/arch/arm64/util/arm-spe.c | 10 ++++++++++
tools/perf/tests/attr/README | 1 +
tools/perf/tests/attr/test-record-spe-physical-address | 12 ++++++++++++
tools/perf/util/arm-spe.c | 5 +++--
4 files changed, 26 insertions(+), 2 deletions(-)
^ permalink raw reply [flat|nested] 11+ messages in thread
* [PATCH 1/3] perf: arm-spe: Fix addresses of synthesized SPE events
2022-04-21 16:52 [PATCH 0/3] perf: arm-spe: Fix addresses of synthesized Arm SPE events Timothy Hayes
@ 2022-04-21 16:52 ` Timothy Hayes
2022-04-24 12:28 ` Leo Yan
2022-04-21 16:52 ` [PATCH 2/3] perf: arm-spe: Fix SPE events with phys addresses Timothy Hayes
2022-04-21 16:52 ` [PATCH 3/3] perf test: Add perf_event_attr test for Arm SPE Timothy Hayes
2 siblings, 1 reply; 11+ messages in thread
From: Timothy Hayes @ 2022-04-21 16:52 UTC (permalink / raw)
To: linux-kernel, linux-perf-users, acme
Cc: Timothy Hayes, John Garry, Will Deacon, Mathieu Poirier, Leo Yan,
Mark Rutland, Alexander Shishkin, Jiri Olsa, Namhyung Kim,
Martin KaFai Lau, Song Liu, Yonghong Song, John Fastabend,
KP Singh, linux-arm-kernel, netdev, bpf
This patch corrects a bug whereby synthesized events from SPE
samples are missing virtual addresses.
Signed-off-by: Timothy Hayes <timothy.hayes@arm.com>
---
tools/perf/util/arm-spe.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/tools/perf/util/arm-spe.c b/tools/perf/util/arm-spe.c
index d2b64e3f588b..151cc38a171c 100644
--- a/tools/perf/util/arm-spe.c
+++ b/tools/perf/util/arm-spe.c
@@ -1036,7 +1036,7 @@ arm_spe_synth_events(struct arm_spe *spe, struct perf_session *session)
attr.sample_type = evsel->core.attr.sample_type & PERF_SAMPLE_MASK;
attr.sample_type |= PERF_SAMPLE_IP | PERF_SAMPLE_TID |
PERF_SAMPLE_PERIOD | PERF_SAMPLE_DATA_SRC |
- PERF_SAMPLE_WEIGHT;
+ PERF_SAMPLE_WEIGHT | PERF_SAMPLE_ADDR;
if (spe->timeless_decoding)
attr.sample_type &= ~(u64)PERF_SAMPLE_TIME;
else
--
2.25.1
^ permalink raw reply related [flat|nested] 11+ messages in thread
* [PATCH 2/3] perf: arm-spe: Fix SPE events with phys addresses
2022-04-21 16:52 [PATCH 0/3] perf: arm-spe: Fix addresses of synthesized Arm SPE events Timothy Hayes
2022-04-21 16:52 ` [PATCH 1/3] perf: arm-spe: Fix addresses of synthesized " Timothy Hayes
@ 2022-04-21 16:52 ` Timothy Hayes
2022-04-24 12:59 ` Leo Yan
2022-04-21 16:52 ` [PATCH 3/3] perf test: Add perf_event_attr test for Arm SPE Timothy Hayes
2 siblings, 1 reply; 11+ messages in thread
From: Timothy Hayes @ 2022-04-21 16:52 UTC (permalink / raw)
To: linux-kernel, linux-perf-users, acme
Cc: Timothy Hayes, John Garry, Will Deacon, Mathieu Poirier, Leo Yan,
Mark Rutland, Alexander Shishkin, Jiri Olsa, Namhyung Kim,
Martin KaFai Lau, Song Liu, Yonghong Song, John Fastabend,
KP Singh, linux-arm-kernel, netdev, bpf
This patch corrects a bug whereby SPE collection is invoked with
pa_enable=1 but synthesized events fail to show physical addresses.
Signed-off-by: Timothy Hayes <timothy.hayes@arm.com>
---
tools/perf/arch/arm64/util/arm-spe.c | 10 ++++++++++
tools/perf/util/arm-spe.c | 3 ++-
2 files changed, 12 insertions(+), 1 deletion(-)
diff --git a/tools/perf/arch/arm64/util/arm-spe.c b/tools/perf/arch/arm64/util/arm-spe.c
index af4d63af8072..e8b577d33e53 100644
--- a/tools/perf/arch/arm64/util/arm-spe.c
+++ b/tools/perf/arch/arm64/util/arm-spe.c
@@ -148,6 +148,7 @@ static int arm_spe_recording_options(struct auxtrace_record *itr,
bool privileged = perf_event_paranoid_check(-1);
struct evsel *tracking_evsel;
int err;
+ u64 bit;
sper->evlist = evlist;
@@ -245,6 +246,15 @@ static int arm_spe_recording_options(struct auxtrace_record *itr,
*/
evsel__set_sample_bit(arm_spe_evsel, DATA_SRC);
+ /*
+ * The PHYS_ADDR flag does not affect the driver behaviour, it is used to
+ * inform that the resulting output's SPE samples contain physical addresses
+ * where applicable.
+ */
+ bit = perf_pmu__format_bits(&arm_spe_pmu->format, "pa_enable");
+ if (arm_spe_evsel->core.attr.config & bit)
+ evsel__set_sample_bit(arm_spe_evsel, PHYS_ADDR);
+
/* Add dummy event to keep tracking */
err = parse_events(evlist, "dummy:u", NULL);
if (err)
diff --git a/tools/perf/util/arm-spe.c b/tools/perf/util/arm-spe.c
index 151cc38a171c..1a80151baed9 100644
--- a/tools/perf/util/arm-spe.c
+++ b/tools/perf/util/arm-spe.c
@@ -1033,7 +1033,8 @@ arm_spe_synth_events(struct arm_spe *spe, struct perf_session *session)
memset(&attr, 0, sizeof(struct perf_event_attr));
attr.size = sizeof(struct perf_event_attr);
attr.type = PERF_TYPE_HARDWARE;
- attr.sample_type = evsel->core.attr.sample_type & PERF_SAMPLE_MASK;
+ attr.sample_type = evsel->core.attr.sample_type &
+ (PERF_SAMPLE_MASK | PERF_SAMPLE_PHYS_ADDR);
attr.sample_type |= PERF_SAMPLE_IP | PERF_SAMPLE_TID |
PERF_SAMPLE_PERIOD | PERF_SAMPLE_DATA_SRC |
PERF_SAMPLE_WEIGHT | PERF_SAMPLE_ADDR;
--
2.25.1
^ permalink raw reply related [flat|nested] 11+ messages in thread
* [PATCH 3/3] perf test: Add perf_event_attr test for Arm SPE
2022-04-21 16:52 [PATCH 0/3] perf: arm-spe: Fix addresses of synthesized Arm SPE events Timothy Hayes
2022-04-21 16:52 ` [PATCH 1/3] perf: arm-spe: Fix addresses of synthesized " Timothy Hayes
2022-04-21 16:52 ` [PATCH 2/3] perf: arm-spe: Fix SPE events with phys addresses Timothy Hayes
@ 2022-04-21 16:52 ` Timothy Hayes
2022-04-24 14:53 ` Leo Yan
2 siblings, 1 reply; 11+ messages in thread
From: Timothy Hayes @ 2022-04-21 16:52 UTC (permalink / raw)
To: linux-kernel, linux-perf-users, acme
Cc: Timothy Hayes, John Garry, Will Deacon, Mathieu Poirier, Leo Yan,
Mark Rutland, Alexander Shishkin, Jiri Olsa, Namhyung Kim,
Martin KaFai Lau, Song Liu, Yonghong Song, John Fastabend,
KP Singh, linux-arm-kernel, netdev, bpf
Adds a perf_event_attr test for Arm SPE in which the presence of
physical addresses are checked when SPE unit is run with pa_enable=1.
Signed-off-by: Timothy Hayes <timothy.hayes@arm.com>
---
tools/perf/tests/attr/README | 1 +
.../perf/tests/attr/test-record-spe-physical-address | 12 ++++++++++++
2 files changed, 13 insertions(+)
create mode 100644 tools/perf/tests/attr/test-record-spe-physical-address
diff --git a/tools/perf/tests/attr/README b/tools/perf/tests/attr/README
index 454505d343fa..eb3f7d4bb324 100644
--- a/tools/perf/tests/attr/README
+++ b/tools/perf/tests/attr/README
@@ -60,6 +60,7 @@ Following tests are defined (with perf commands):
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 record -e arm_spe_0/pa_enable=1/ -- kill (test-record-spe-physical-address)
perf stat -e cycles kill (test-stat-basic)
perf stat kill (test-stat-default)
perf stat -d kill (test-stat-detailed-1)
diff --git a/tools/perf/tests/attr/test-record-spe-physical-address b/tools/perf/tests/attr/test-record-spe-physical-address
new file mode 100644
index 000000000000..7ebcf5012ce3
--- /dev/null
+++ b/tools/perf/tests/attr/test-record-spe-physical-address
@@ -0,0 +1,12 @@
+[config]
+command = record
+args = --no-bpf-event -e arm_spe_0/pa_enable=1/ -- kill >/dev/null 2>&1
+ret = 1
+arch = aarch64
+
+[event-10:base-record-spe]
+# 622727 is the decimal of IP|TID|TIME|CPU|IDENTIFIER|DATA_SRC|PHYS_ADDR
+sample_type=622727
+
+# dummy event
+[event-1:base-record-spe]
\ No newline at end of file
--
2.25.1
^ permalink raw reply related [flat|nested] 11+ messages in thread
* Re: [PATCH 1/3] perf: arm-spe: Fix addresses of synthesized SPE events
2022-04-21 16:52 ` [PATCH 1/3] perf: arm-spe: Fix addresses of synthesized " Timothy Hayes
@ 2022-04-24 12:28 ` Leo Yan
2022-04-24 15:22 ` Leo Yan
0 siblings, 1 reply; 11+ messages in thread
From: Leo Yan @ 2022-04-24 12:28 UTC (permalink / raw)
To: Timothy Hayes
Cc: linux-kernel, linux-perf-users, acme, John Garry, Will Deacon,
Mathieu Poirier, Mark Rutland, Alexander Shishkin, Jiri Olsa,
Namhyung Kim, Martin KaFai Lau, Song Liu, Yonghong Song,
John Fastabend, KP Singh, linux-arm-kernel, netdev, bpf
On Thu, Apr 21, 2022 at 05:52:03PM +0100, Timothy Hayes wrote:
> This patch corrects a bug whereby synthesized events from SPE
> samples are missing virtual addresses.
>
> Signed-off-by: Timothy Hayes <timothy.hayes@arm.com>
Reviewed-by: Leo Yan <leo.yan@linaro.org>
> ---
> tools/perf/util/arm-spe.c | 2 +-
> 1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/tools/perf/util/arm-spe.c b/tools/perf/util/arm-spe.c
> index d2b64e3f588b..151cc38a171c 100644
> --- a/tools/perf/util/arm-spe.c
> +++ b/tools/perf/util/arm-spe.c
> @@ -1036,7 +1036,7 @@ arm_spe_synth_events(struct arm_spe *spe, struct perf_session *session)
> attr.sample_type = evsel->core.attr.sample_type & PERF_SAMPLE_MASK;
> attr.sample_type |= PERF_SAMPLE_IP | PERF_SAMPLE_TID |
> PERF_SAMPLE_PERIOD | PERF_SAMPLE_DATA_SRC |
> - PERF_SAMPLE_WEIGHT;
> + PERF_SAMPLE_WEIGHT | PERF_SAMPLE_ADDR;
> if (spe->timeless_decoding)
> attr.sample_type &= ~(u64)PERF_SAMPLE_TIME;
> else
> --
> 2.25.1
>
^ permalink raw reply [flat|nested] 11+ messages in thread
* Re: [PATCH 2/3] perf: arm-spe: Fix SPE events with phys addresses
2022-04-21 16:52 ` [PATCH 2/3] perf: arm-spe: Fix SPE events with phys addresses Timothy Hayes
@ 2022-04-24 12:59 ` Leo Yan
2022-04-25 9:12 ` James Clark
0 siblings, 1 reply; 11+ messages in thread
From: Leo Yan @ 2022-04-24 12:59 UTC (permalink / raw)
To: Timothy Hayes
Cc: linux-kernel, linux-perf-users, acme, John Garry, Will Deacon,
Mathieu Poirier, Mark Rutland, Alexander Shishkin, Jiri Olsa,
Namhyung Kim, Martin KaFai Lau, Song Liu, Yonghong Song,
John Fastabend, KP Singh, linux-arm-kernel, netdev, bpf
Hi Timothy,
On Thu, Apr 21, 2022 at 05:52:04PM +0100, Timothy Hayes wrote:
> This patch corrects a bug whereby SPE collection is invoked with
> pa_enable=1 but synthesized events fail to show physical addresses.
>
> Signed-off-by: Timothy Hayes <timothy.hayes@arm.com>
> ---
> tools/perf/arch/arm64/util/arm-spe.c | 10 ++++++++++
> tools/perf/util/arm-spe.c | 3 ++-
> 2 files changed, 12 insertions(+), 1 deletion(-)
>
> diff --git a/tools/perf/arch/arm64/util/arm-spe.c b/tools/perf/arch/arm64/util/arm-spe.c
> index af4d63af8072..e8b577d33e53 100644
> --- a/tools/perf/arch/arm64/util/arm-spe.c
> +++ b/tools/perf/arch/arm64/util/arm-spe.c
> @@ -148,6 +148,7 @@ static int arm_spe_recording_options(struct auxtrace_record *itr,
> bool privileged = perf_event_paranoid_check(-1);
> struct evsel *tracking_evsel;
> int err;
> + u64 bit;
>
> sper->evlist = evlist;
>
> @@ -245,6 +246,15 @@ static int arm_spe_recording_options(struct auxtrace_record *itr,
> */
> evsel__set_sample_bit(arm_spe_evsel, DATA_SRC);
>
> + /*
> + * The PHYS_ADDR flag does not affect the driver behaviour, it is used to
> + * inform that the resulting output's SPE samples contain physical addresses
> + * where applicable.
> + */
> + bit = perf_pmu__format_bits(&arm_spe_pmu->format, "pa_enable");
> + if (arm_spe_evsel->core.attr.config & bit)
> + evsel__set_sample_bit(arm_spe_evsel, PHYS_ADDR);
> +
> /* Add dummy event to keep tracking */
> err = parse_events(evlist, "dummy:u", NULL);
> if (err)
> diff --git a/tools/perf/util/arm-spe.c b/tools/perf/util/arm-spe.c
> index 151cc38a171c..1a80151baed9 100644
> --- a/tools/perf/util/arm-spe.c
> +++ b/tools/perf/util/arm-spe.c
> @@ -1033,7 +1033,8 @@ arm_spe_synth_events(struct arm_spe *spe, struct perf_session *session)
> memset(&attr, 0, sizeof(struct perf_event_attr));
> attr.size = sizeof(struct perf_event_attr);
> attr.type = PERF_TYPE_HARDWARE;
> - attr.sample_type = evsel->core.attr.sample_type & PERF_SAMPLE_MASK;
> + attr.sample_type = evsel->core.attr.sample_type &
> + (PERF_SAMPLE_MASK | PERF_SAMPLE_PHYS_ADDR);
I verified this patch and I can confirm the physical address can be
dumped successfully.
I have a more general question, seems to me, we need to change the
macro PERF_SAMPLE_MASK in the file util/event.h as below, so
here doesn't need to 'or' the flag PERF_SAMPLE_PHYS_ADDR anymore.
@Arnaldo, @Jiri, could you confirm if this is the right way to move
forward? I am not sure why PERF_SAMPLE_MASK doesn't contain the bit
PERF_SAMPLE_PHYS_ADDR in current code.
diff --git a/tools/perf/util/event.h b/tools/perf/util/event.h
index cdd72e05fd28..c905ac32ebad 100644
--- a/tools/perf/util/event.h
+++ b/tools/perf/util/event.h
@@ -39,7 +39,7 @@ struct perf_event_attr;
PERF_SAMPLE_TIME | PERF_SAMPLE_ADDR | \
PERF_SAMPLE_ID | PERF_SAMPLE_STREAM_ID | \
PERF_SAMPLE_CPU | PERF_SAMPLE_PERIOD | \
- PERF_SAMPLE_IDENTIFIER)
+ PERF_SAMPLE_IDENTIFIER | PERF_SAMPLE_PHYS_ADDR)
Thanks,
Leo
> attr.sample_type |= PERF_SAMPLE_IP | PERF_SAMPLE_TID |
> PERF_SAMPLE_PERIOD | PERF_SAMPLE_DATA_SRC |
> PERF_SAMPLE_WEIGHT | PERF_SAMPLE_ADDR;
> --
> 2.25.1
>
^ permalink raw reply related [flat|nested] 11+ messages in thread
* Re: [PATCH 3/3] perf test: Add perf_event_attr test for Arm SPE
2022-04-21 16:52 ` [PATCH 3/3] perf test: Add perf_event_attr test for Arm SPE Timothy Hayes
@ 2022-04-24 14:53 ` Leo Yan
2022-04-28 13:41 ` Arnaldo Carvalho de Melo
0 siblings, 1 reply; 11+ messages in thread
From: Leo Yan @ 2022-04-24 14:53 UTC (permalink / raw)
To: Timothy Hayes
Cc: linux-kernel, linux-perf-users, acme, John Garry, Will Deacon,
Mathieu Poirier, Mark Rutland, Alexander Shishkin, Jiri Olsa,
Namhyung Kim, Martin KaFai Lau, Song Liu, Yonghong Song,
John Fastabend, KP Singh, linux-arm-kernel, netdev, bpf
On Thu, Apr 21, 2022 at 05:52:05PM +0100, Timothy Hayes wrote:
> Adds a perf_event_attr test for Arm SPE in which the presence of
> physical addresses are checked when SPE unit is run with pa_enable=1.
>
> Signed-off-by: Timothy Hayes <timothy.hayes@arm.com>
Reviewed-by: Leo Yan <leo.yan@linaro.org>
Tested-by: Leo Yan <leo.yan@linaro.org>
> ---
> tools/perf/tests/attr/README | 1 +
> .../perf/tests/attr/test-record-spe-physical-address | 12 ++++++++++++
> 2 files changed, 13 insertions(+)
> create mode 100644 tools/perf/tests/attr/test-record-spe-physical-address
>
> diff --git a/tools/perf/tests/attr/README b/tools/perf/tests/attr/README
> index 454505d343fa..eb3f7d4bb324 100644
> --- a/tools/perf/tests/attr/README
> +++ b/tools/perf/tests/attr/README
> @@ -60,6 +60,7 @@ Following tests are defined (with perf commands):
> 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 record -e arm_spe_0/pa_enable=1/ -- kill (test-record-spe-physical-address)
> perf stat -e cycles kill (test-stat-basic)
> perf stat kill (test-stat-default)
> perf stat -d kill (test-stat-detailed-1)
> diff --git a/tools/perf/tests/attr/test-record-spe-physical-address b/tools/perf/tests/attr/test-record-spe-physical-address
> new file mode 100644
> index 000000000000..7ebcf5012ce3
> --- /dev/null
> +++ b/tools/perf/tests/attr/test-record-spe-physical-address
> @@ -0,0 +1,12 @@
> +[config]
> +command = record
> +args = --no-bpf-event -e arm_spe_0/pa_enable=1/ -- kill >/dev/null 2>&1
> +ret = 1
> +arch = aarch64
> +
> +[event-10:base-record-spe]
> +# 622727 is the decimal of IP|TID|TIME|CPU|IDENTIFIER|DATA_SRC|PHYS_ADDR
> +sample_type=622727
> +
> +# dummy event
> +[event-1:base-record-spe]
> \ No newline at end of file
> --
> 2.25.1
>
^ permalink raw reply [flat|nested] 11+ messages in thread
* Re: [PATCH 1/3] perf: arm-spe: Fix addresses of synthesized SPE events
2022-04-24 12:28 ` Leo Yan
@ 2022-04-24 15:22 ` Leo Yan
0 siblings, 0 replies; 11+ messages in thread
From: Leo Yan @ 2022-04-24 15:22 UTC (permalink / raw)
To: Timothy Hayes
Cc: linux-kernel, linux-perf-users, acme, John Garry, Will Deacon,
Mathieu Poirier, Mark Rutland, Alexander Shishkin, Jiri Olsa,
Namhyung Kim, Martin KaFai Lau, Song Liu, Yonghong Song,
John Fastabend, KP Singh, linux-arm-kernel, netdev, bpf
On Sun, Apr 24, 2022 at 08:28:31PM +0800, Leo Yan wrote:
> On Thu, Apr 21, 2022 at 05:52:03PM +0100, Timothy Hayes wrote:
> > This patch corrects a bug whereby synthesized events from SPE
> > samples are missing virtual addresses.
> >
> > Signed-off-by: Timothy Hayes <timothy.hayes@arm.com>
>
> Reviewed-by: Leo Yan <leo.yan@linaro.org>
Supplement for fixed tag:
Since patches 01, 02 are fixing bugs, please add fixed tag [1]:
Fixes: 54f7815efef7 ("perf arm-spe: Fill address info for samples")
Thanks,
Leo
[1] https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/Documentation/process/submitting-patches.rst#n138
^ permalink raw reply [flat|nested] 11+ messages in thread
* Re: [PATCH 2/3] perf: arm-spe: Fix SPE events with phys addresses
2022-04-24 12:59 ` Leo Yan
@ 2022-04-25 9:12 ` James Clark
2022-04-26 13:19 ` Leo Yan
0 siblings, 1 reply; 11+ messages in thread
From: James Clark @ 2022-04-25 9:12 UTC (permalink / raw)
To: Leo Yan, Timothy Hayes
Cc: linux-kernel, linux-perf-users, acme, John Garry, Will Deacon,
Mathieu Poirier, Mark Rutland, Alexander Shishkin, Jiri Olsa,
Namhyung Kim, Martin KaFai Lau, Song Liu, Yonghong Song,
John Fastabend, KP Singh, linux-arm-kernel, netdev, bpf
On 24/04/2022 13:59, Leo Yan wrote:
> Hi Timothy,
>
> On Thu, Apr 21, 2022 at 05:52:04PM +0100, Timothy Hayes wrote:
>> This patch corrects a bug whereby SPE collection is invoked with
>> pa_enable=1 but synthesized events fail to show physical addresses.
>>
>> Signed-off-by: Timothy Hayes <timothy.hayes@arm.com>
>> ---
>> tools/perf/arch/arm64/util/arm-spe.c | 10 ++++++++++
>> tools/perf/util/arm-spe.c | 3 ++-
>> 2 files changed, 12 insertions(+), 1 deletion(-)
>>
>> diff --git a/tools/perf/arch/arm64/util/arm-spe.c b/tools/perf/arch/arm64/util/arm-spe.c
>> index af4d63af8072..e8b577d33e53 100644
>> --- a/tools/perf/arch/arm64/util/arm-spe.c
>> +++ b/tools/perf/arch/arm64/util/arm-spe.c
>> @@ -148,6 +148,7 @@ static int arm_spe_recording_options(struct auxtrace_record *itr,
>> bool privileged = perf_event_paranoid_check(-1);
>> struct evsel *tracking_evsel;
>> int err;
>> + u64 bit;
>>
>> sper->evlist = evlist;
>>
>> @@ -245,6 +246,15 @@ static int arm_spe_recording_options(struct auxtrace_record *itr,
>> */
>> evsel__set_sample_bit(arm_spe_evsel, DATA_SRC);
>>
>> + /*
>> + * The PHYS_ADDR flag does not affect the driver behaviour, it is used to
>> + * inform that the resulting output's SPE samples contain physical addresses
>> + * where applicable.
>> + */
>> + bit = perf_pmu__format_bits(&arm_spe_pmu->format, "pa_enable");
>> + if (arm_spe_evsel->core.attr.config & bit)
>> + evsel__set_sample_bit(arm_spe_evsel, PHYS_ADDR);
>> +
>> /* Add dummy event to keep tracking */
>> err = parse_events(evlist, "dummy:u", NULL);
>> if (err)
>> diff --git a/tools/perf/util/arm-spe.c b/tools/perf/util/arm-spe.c
>> index 151cc38a171c..1a80151baed9 100644
>> --- a/tools/perf/util/arm-spe.c
>> +++ b/tools/perf/util/arm-spe.c
>> @@ -1033,7 +1033,8 @@ arm_spe_synth_events(struct arm_spe *spe, struct perf_session *session)
>> memset(&attr, 0, sizeof(struct perf_event_attr));
>> attr.size = sizeof(struct perf_event_attr);
>> attr.type = PERF_TYPE_HARDWARE;
>> - attr.sample_type = evsel->core.attr.sample_type & PERF_SAMPLE_MASK;
>> + attr.sample_type = evsel->core.attr.sample_type &
>> + (PERF_SAMPLE_MASK | PERF_SAMPLE_PHYS_ADDR);
>
> I verified this patch and I can confirm the physical address can be
> dumped successfully.
>
> I have a more general question, seems to me, we need to change the
> macro PERF_SAMPLE_MASK in the file util/event.h as below, so
> here doesn't need to 'or' the flag PERF_SAMPLE_PHYS_ADDR anymore.
>
> @Arnaldo, @Jiri, could you confirm if this is the right way to move
> forward? I am not sure why PERF_SAMPLE_MASK doesn't contain the bit
> PERF_SAMPLE_PHYS_ADDR in current code.
I think there is a reason that PERF_SAMPLE_MASK is a subset of all the
bits. This comment below suggests it. Is it so the mask only includes fields
that are 64bits? That makes the __evsel__sample_size() function a simple
multiplication of a count of all the fields that are 64bits.
static int
perf_event__check_size(union perf_event *event, unsigned int sample_size)
{
/*
* The evsel's sample_size is based on PERF_SAMPLE_MASK which includes
* up to PERF_SAMPLE_PERIOD. After that overflow() must be used to
* check the format does not go past the end of the event.
*/
if (sample_size + sizeof(event->header) > event->header.size)
return -EFAULT;
return 0;
}
Having said that, the mask was updated once to add PERF_SAMPLE_IDENTIFIER to
it, so that comment is slightly out of date now.
>
> diff --git a/tools/perf/util/event.h b/tools/perf/util/event.h
> index cdd72e05fd28..c905ac32ebad 100644
> --- a/tools/perf/util/event.h
> +++ b/tools/perf/util/event.h
> @@ -39,7 +39,7 @@ struct perf_event_attr;
> PERF_SAMPLE_TIME | PERF_SAMPLE_ADDR | \
> PERF_SAMPLE_ID | PERF_SAMPLE_STREAM_ID | \
> PERF_SAMPLE_CPU | PERF_SAMPLE_PERIOD | \
> - PERF_SAMPLE_IDENTIFIER)
> + PERF_SAMPLE_IDENTIFIER | PERF_SAMPLE_PHYS_ADDR)
>
> Thanks,
> Leo
>
>> attr.sample_type |= PERF_SAMPLE_IP | PERF_SAMPLE_TID |
>> PERF_SAMPLE_PERIOD | PERF_SAMPLE_DATA_SRC |
>> PERF_SAMPLE_WEIGHT | PERF_SAMPLE_ADDR;
>> --
>> 2.25.1
>>
^ permalink raw reply [flat|nested] 11+ messages in thread
* Re: [PATCH 2/3] perf: arm-spe: Fix SPE events with phys addresses
2022-04-25 9:12 ` James Clark
@ 2022-04-26 13:19 ` Leo Yan
0 siblings, 0 replies; 11+ messages in thread
From: Leo Yan @ 2022-04-26 13:19 UTC (permalink / raw)
To: James Clark
Cc: Timothy Hayes, linux-kernel, linux-perf-users, acme, John Garry,
Will Deacon, Mathieu Poirier, Mark Rutland, Alexander Shishkin,
Jiri Olsa, Namhyung Kim, Martin KaFai Lau, Song Liu,
Yonghong Song, John Fastabend, KP Singh, linux-arm-kernel,
netdev, bpf
On Mon, Apr 25, 2022 at 10:12:36AM +0100, James Clark wrote:
[...]
> >> diff --git a/tools/perf/util/arm-spe.c b/tools/perf/util/arm-spe.c
> >> index 151cc38a171c..1a80151baed9 100644
> >> --- a/tools/perf/util/arm-spe.c
> >> +++ b/tools/perf/util/arm-spe.c
> >> @@ -1033,7 +1033,8 @@ arm_spe_synth_events(struct arm_spe *spe, struct perf_session *session)
> >> memset(&attr, 0, sizeof(struct perf_event_attr));
> >> attr.size = sizeof(struct perf_event_attr);
> >> attr.type = PERF_TYPE_HARDWARE;
> >> - attr.sample_type = evsel->core.attr.sample_type & PERF_SAMPLE_MASK;
> >> + attr.sample_type = evsel->core.attr.sample_type &
> >> + (PERF_SAMPLE_MASK | PERF_SAMPLE_PHYS_ADDR);
> >
> > I verified this patch and I can confirm the physical address can be
> > dumped successfully.
> >
> > I have a more general question, seems to me, we need to change the
> > macro PERF_SAMPLE_MASK in the file util/event.h as below, so
> > here doesn't need to 'or' the flag PERF_SAMPLE_PHYS_ADDR anymore.
> >
> > @Arnaldo, @Jiri, could you confirm if this is the right way to move
> > forward? I am not sure why PERF_SAMPLE_MASK doesn't contain the bit
> > PERF_SAMPLE_PHYS_ADDR in current code.
>
> I think there is a reason that PERF_SAMPLE_MASK is a subset of all the
> bits. This comment below suggests it. Is it so the mask only includes fields
> that are 64bits? That makes the __evsel__sample_size() function a simple
> multiplication of a count of all the fields that are 64bits.
After reading code, seems the conclusion "a count of all the fields
that are 64bits" is not valid. PERF_SAMPLE_MASK contains bits
PERF_SAMPLE_IP and PERF_SAMPLE_TID, the corresponding fields 'pid'
and 'tid' in the structure perf_sample are u32 type.
> static int
> perf_event__check_size(union perf_event *event, unsigned int sample_size)
> {
> /*
> * The evsel's sample_size is based on PERF_SAMPLE_MASK which includes
> * up to PERF_SAMPLE_PERIOD. After that overflow() must be used to
> * check the format does not go past the end of the event.
> */
> if (sample_size + sizeof(event->header) > event->header.size)
> return -EFAULT;
Okay, thanks for sharing the info, it does show that it's deliberately to
not contain all fields in PERF_SAMPLE_MASK. If so, this patch is fine
for me:
Reviewed-by: Leo Yan <leo.yan@linaro.org>
> return 0;
> }
>
> Having said that, the mask was updated once to add PERF_SAMPLE_IDENTIFIER to
> it, so that comment is slightly out of date now.
>
>
> >
> > diff --git a/tools/perf/util/event.h b/tools/perf/util/event.h
> > index cdd72e05fd28..c905ac32ebad 100644
> > --- a/tools/perf/util/event.h
> > +++ b/tools/perf/util/event.h
> > @@ -39,7 +39,7 @@ struct perf_event_attr;
> > PERF_SAMPLE_TIME | PERF_SAMPLE_ADDR | \
> > PERF_SAMPLE_ID | PERF_SAMPLE_STREAM_ID | \
> > PERF_SAMPLE_CPU | PERF_SAMPLE_PERIOD | \
> > - PERF_SAMPLE_IDENTIFIER)
> > + PERF_SAMPLE_IDENTIFIER | PERF_SAMPLE_PHYS_ADDR)
> >
> > Thanks,
> > Leo
> >
> >> attr.sample_type |= PERF_SAMPLE_IP | PERF_SAMPLE_TID |
> >> PERF_SAMPLE_PERIOD | PERF_SAMPLE_DATA_SRC |
> >> PERF_SAMPLE_WEIGHT | PERF_SAMPLE_ADDR;
> >> --
> >> 2.25.1
> >>
^ permalink raw reply [flat|nested] 11+ messages in thread
* Re: [PATCH 3/3] perf test: Add perf_event_attr test for Arm SPE
2022-04-24 14:53 ` Leo Yan
@ 2022-04-28 13:41 ` Arnaldo Carvalho de Melo
0 siblings, 0 replies; 11+ messages in thread
From: Arnaldo Carvalho de Melo @ 2022-04-28 13:41 UTC (permalink / raw)
To: Leo Yan
Cc: Timothy Hayes, linux-kernel, linux-perf-users, John Garry,
Will Deacon, Mathieu Poirier, Mark Rutland, Alexander Shishkin,
Jiri Olsa, Namhyung Kim, Martin KaFai Lau, Song Liu,
Yonghong Song, John Fastabend, KP Singh, linux-arm-kernel,
netdev, bpf
Em Sun, Apr 24, 2022 at 10:53:07PM +0800, Leo Yan escreveu:
> On Thu, Apr 21, 2022 at 05:52:05PM +0100, Timothy Hayes wrote:
> > Adds a perf_event_attr test for Arm SPE in which the presence of
> > physical addresses are checked when SPE unit is run with pa_enable=1.
> >
> > Signed-off-by: Timothy Hayes <timothy.hayes@arm.com>
>
> Reviewed-by: Leo Yan <leo.yan@linaro.org>
> Tested-by: Leo Yan <leo.yan@linaro.org>
Thanks, applied the set to perf/urgent.
- Arnaldo
> > ---
> > tools/perf/tests/attr/README | 1 +
> > .../perf/tests/attr/test-record-spe-physical-address | 12 ++++++++++++
> > 2 files changed, 13 insertions(+)
> > create mode 100644 tools/perf/tests/attr/test-record-spe-physical-address
> >
> > diff --git a/tools/perf/tests/attr/README b/tools/perf/tests/attr/README
> > index 454505d343fa..eb3f7d4bb324 100644
> > --- a/tools/perf/tests/attr/README
> > +++ b/tools/perf/tests/attr/README
> > @@ -60,6 +60,7 @@ Following tests are defined (with perf commands):
> > 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 record -e arm_spe_0/pa_enable=1/ -- kill (test-record-spe-physical-address)
> > perf stat -e cycles kill (test-stat-basic)
> > perf stat kill (test-stat-default)
> > perf stat -d kill (test-stat-detailed-1)
> > diff --git a/tools/perf/tests/attr/test-record-spe-physical-address b/tools/perf/tests/attr/test-record-spe-physical-address
> > new file mode 100644
> > index 000000000000..7ebcf5012ce3
> > --- /dev/null
> > +++ b/tools/perf/tests/attr/test-record-spe-physical-address
> > @@ -0,0 +1,12 @@
> > +[config]
> > +command = record
> > +args = --no-bpf-event -e arm_spe_0/pa_enable=1/ -- kill >/dev/null 2>&1
> > +ret = 1
> > +arch = aarch64
> > +
> > +[event-10:base-record-spe]
> > +# 622727 is the decimal of IP|TID|TIME|CPU|IDENTIFIER|DATA_SRC|PHYS_ADDR
> > +sample_type=622727
> > +
> > +# dummy event
> > +[event-1:base-record-spe]
> > \ No newline at end of file
> > --
> > 2.25.1
> >
--
- Arnaldo
^ permalink raw reply [flat|nested] 11+ messages in thread
end of thread, other threads:[~2022-04-28 13:41 UTC | newest]
Thread overview: 11+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2022-04-21 16:52 [PATCH 0/3] perf: arm-spe: Fix addresses of synthesized Arm SPE events Timothy Hayes
2022-04-21 16:52 ` [PATCH 1/3] perf: arm-spe: Fix addresses of synthesized " Timothy Hayes
2022-04-24 12:28 ` Leo Yan
2022-04-24 15:22 ` Leo Yan
2022-04-21 16:52 ` [PATCH 2/3] perf: arm-spe: Fix SPE events with phys addresses Timothy Hayes
2022-04-24 12:59 ` Leo Yan
2022-04-25 9:12 ` James Clark
2022-04-26 13:19 ` Leo Yan
2022-04-21 16:52 ` [PATCH 3/3] perf test: Add perf_event_attr test for Arm SPE Timothy Hayes
2022-04-24 14:53 ` Leo Yan
2022-04-28 13:41 ` Arnaldo Carvalho de Melo
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).