Skip to contentSkip to conversation

perf evsel: Don't set sample_regs_intr/sample_regs_user for dummy event

PATCH In review Jin Yao 4 participants 8 messages Jul 20, 2020 – Aug 4, 2020 perf events

lore
Share & tools
lore

Jin Yao

author
#0 lore raw

PATCH v2 1/1 perf evsel: Don't set sample_regs_intr/sample_regs_user for dummy event

Since commit 0a892c1c9472 ("perf record: Add dummy event during system wide synthesis"), a dummy event is added to capture mmaps.

But if we run perf-record as,

 # perf record -e cycles:p -IXMM0 -a -- sleep 1
 Error:
 dummy:HG: PMU Hardware doesn't support sampling/overflow-interrupts. Try 'perf stat'

The issue is, if we enable the extended regs (-IXMM0), but the pmu->capabilities is not set with PERF_PMU_CAP_EXTENDED_REGS, the kernel will return -EOPNOTSUPP error.

See following code:

/* in kernel/events/core.c */
static int perf_try_init_event(struct pmu *pmu, struct perf_event *event)

{
        ....
        if (!(pmu->capabilities & PERF_PMU_CAP_EXTENDED_REGS) &&
            has_extended_regs(event))
                ret = -EOPNOTSUPP;
        ....
}

For software dummy event, the PMU should not be set with PERF_PMU_CAP_EXTENDED_REGS. But unfortunately now, the dummy event has possibility to be set with PERF_REG_EXTENDED_MASK bit.

In evsel__config, /* tools/perf/util/evsel.c */

if (opts->sample_intr_regs) {
        attr->sample_regs_intr = opts->sample_intr_regs;
}

If we use -IXMM0, the attr>sample_regs_intr will be set with PERF_REG_EXTENDED_MASK bit.

It doesn't make sense to set attr->sample_regs_intr for a software dummy event.

This patch adds dummy event checking before setting attr->sample_regs_intr and attr->sample_regs_user.

After:

  # ./perf record -e cycles:p -IXMM0 -a -- sleep 1
  [ perf record: Woken up 1 times to write data ]
  [ perf record: Captured and wrote 0.413 MB perf.data (45 samples) ]

v2:
---
Rebase to perf/core

Fixes 0a892c1c9472 ("perf record: Add dummy event during system wide synthesis")
---
 tools/perf/util/evsel.c | 6 ++++--
 1 file changed, 4 insertions(+), 2 deletions(-)
tools/perf/util/evsel.c +4 -2
index 9aa51a65593d..11794d3b7879 100644
@@ -1014,12 +1014,14 @@ void evsel__config(struct evsel *evsel, struct record_opts *opts,
 	if (callchain && callchain->enabled && !evsel->no_aux_samples)
 		evsel__config_callchain(evsel, opts, callchain);
 
-	if (opts->sample_intr_regs && !evsel->no_aux_samples) {
+	if (opts->sample_intr_regs && !evsel->no_aux_samples &&
+	    !evsel__is_dummy_event(evsel)) {
 		attr->sample_regs_intr = opts->sample_intr_regs;
 		evsel__set_sample_bit(evsel, REGS_INTR);
 	}
 
-	if (opts->sample_user_regs && !evsel->no_aux_samples) {
+	if (opts->sample_user_regs && !evsel->no_aux_samples &&
+	    !evsel__is_dummy_event(evsel)) {
 		attr->sample_regs_user |= opts->sample_user_regs;
 		evsel__set_sample_bit(evsel, REGS_USER);
 	}
Signature · 1 line
2.17.1
#1 lore raw
Quoted 65 lines · Jin Yao

Since commit 0a892c1c9472 ("perf record: Add dummy event during system wide synthesis"), a dummy event is added to capture mmaps.

But if we run perf-record as,

 # perf record -e cycles:p -IXMM0 -a -- sleep 1
 Error:
 dummy:HG: PMU Hardware doesn't support sampling/overflow-interrupts. Try 'perf stat'

The issue is, if we enable the extended regs (-IXMM0), but the pmu->capabilities is not set with PERF_PMU_CAP_EXTENDED_REGS, the kernel will return -EOPNOTSUPP error.

See following code:

/* in kernel/events/core.c */
static int perf_try_init_event(struct pmu *pmu, struct perf_event *event)

{
        ....
        if (!(pmu->capabilities & PERF_PMU_CAP_EXTENDED_REGS) &&
            has_extended_regs(event))
                ret = -EOPNOTSUPP;
        ....
}

For software dummy event, the PMU should not be set with PERF_PMU_CAP_EXTENDED_REGS. But unfortunately now, the dummy event has possibility to be set with PERF_REG_EXTENDED_MASK bit.

In evsel__config, /* tools/perf/util/evsel.c */

if (opts->sample_intr_regs) {
        attr->sample_regs_intr = opts->sample_intr_regs;
}

If we use -IXMM0, the attr>sample_regs_intr will be set with PERF_REG_EXTENDED_MASK bit.

It doesn't make sense to set attr->sample_regs_intr for a software dummy event.

This patch adds dummy event checking before setting attr->sample_regs_intr and attr->sample_regs_user.

After:

  # ./perf record -e cycles:p -IXMM0 -a -- sleep 1
  [ perf record: Woken up 1 times to write data ]
  [ perf record: Captured and wrote 0.413 MB perf.data (45 samples) ]

v2:
---
Rebase to perf/core

Fixes 0a892c1c9472 ("perf record: Add dummy event during system wide synthesis")
---
 tools/perf/util/evsel.c | 6 ++++--
 1 file changed, 4 insertions(+), 2 deletions(-)
tools/perf/util/evsel.c +0 -0
index 9aa51a65593d..11794d3b7879 100644

tools/perf/util/evsel.c void evsel__config(struct evsel *evsel, struct record_opts *opts, · line 1018

Quoted 6 lines
tools/perf/util/evsel.c +2 -1
@@ -1014,12 +1014,14 @@ void evsel__config(struct evsel *evsel, struct record_opts *opts,
 	if (callchain && callchain->enabled && !evsel->no_aux_samples)
 		evsel__config_callchain(evsel, opts, callchain);
 
-	if (opts->sample_intr_regs && !evsel->no_aux_samples) {
+	if (opts->sample_intr_regs && !evsel->no_aux_samples &&
+	    !evsel__is_dummy_event(evsel)) {

hum, I thought it'd look something like this:

  if (opts->sample_intr_regs && (!evsel->no_aux_samples || !evsel__is_dummy_event(evsel)) 

but I'm not sure how no_aux_samples flag works exactly.. so it might be correct.. just making sure ;-)

cc-ing Adrian

jirka

Quoted 13 lines
tools/perf/util/evsel.c +2 -1
 		attr->sample_regs_intr = opts->sample_intr_regs;
 		evsel__set_sample_bit(evsel, REGS_INTR);
 	}
 
-	if (opts->sample_user_regs && !evsel->no_aux_samples) {
+	if (opts->sample_user_regs && !evsel->no_aux_samples &&
+	    !evsel__is_dummy_event(evsel)) {
 		attr->sample_regs_user |= opts->sample_user_regs;
 		evsel__set_sample_bit(evsel, REGS_USER);
 	}
Signature · 1 line
2.17.1
author
#2 lore raw

Hi Jiri,

Quoted 75 lines · Jiri Olsa
Quoted 65 lines · Jin Yao

Since commit 0a892c1c9472 ("perf record: Add dummy event during system wide synthesis"), a dummy event is added to capture mmaps.

But if we run perf-record as,

  # perf record -e cycles:p -IXMM0 -a -- sleep 1
  Error:
  dummy:HG: PMU Hardware doesn't support sampling/overflow-interrupts. Try 'perf stat'

The issue is, if we enable the extended regs (-IXMM0), but the pmu->capabilities is not set with PERF_PMU_CAP_EXTENDED_REGS, the kernel will return -EOPNOTSUPP error.

See following code:

/* in kernel/events/core.c */
static int perf_try_init_event(struct pmu *pmu, struct perf_event *event)

{
         ....
         if (!(pmu->capabilities & PERF_PMU_CAP_EXTENDED_REGS) &&
             has_extended_regs(event))
                 ret = -EOPNOTSUPP;
         ....
}

For software dummy event, the PMU should not be set with PERF_PMU_CAP_EXTENDED_REGS. But unfortunately now, the dummy event has possibility to be set with PERF_REG_EXTENDED_MASK bit.

In evsel__config, /* tools/perf/util/evsel.c */

if (opts->sample_intr_regs) {
         attr->sample_regs_intr = opts->sample_intr_regs;
}

If we use -IXMM0, the attr>sample_regs_intr will be set with PERF_REG_EXTENDED_MASK bit.

It doesn't make sense to set attr->sample_regs_intr for a software dummy event.

This patch adds dummy event checking before setting attr->sample_regs_intr and attr->sample_regs_user.

After:

   # ./perf record -e cycles:p -IXMM0 -a -- sleep 1
   [ perf record: Woken up 1 times to write data ]
   [ perf record: Captured and wrote 0.413 MB perf.data (45 samples) ]

v2:
---
Rebase to perf/core

Fixes 0a892c1c9472 ("perf record: Add dummy event during system wide synthesis")
---
  tools/perf/util/evsel.c | 6 ++++--
  1 file changed, 4 insertions(+), 2 deletions(-)
tools/perf/util/evsel.c +0 -0
index 9aa51a65593d..11794d3b7879 100644

tools/perf/util/evsel.c void evsel__config(struct evsel *evsel, struct record_opts *opts, · line 1018

Quoted 6 lines
tools/perf/util/evsel.c +2 -1
@@ -1014,12 +1014,14 @@ void evsel__config(struct evsel *evsel, struct record_opts *opts,
  	if (callchain && callchain->enabled && !evsel->no_aux_samples)
  		evsel__config_callchain(evsel, opts, callchain);
 
-	if (opts->sample_intr_regs && !evsel->no_aux_samples) {
+	if (opts->sample_intr_regs && !evsel->no_aux_samples &&
+	    !evsel__is_dummy_event(evsel)) {

hum, I thought it'd look something like this:

Quoted 10 lines
   if (opts->sample_intr_regs && (!evsel->no_aux_samples || !evsel__is_dummy_event(evsel))

but I'm not sure how no_aux_samples flag works exactly.. so it might be correct.. just making sure ;-)

cc-ing Adrian

jirka

no_aux_samples is set to false by default and it's only set to true by pt, right?

So most of the time, !evsel->no_aux_samples is always true.

if (opts->sample_intr_regs && (!evsel->no_aux_samples || !evsel__is_dummy_event(evsel)) {
	attr->sample_regs_intr = opts->sample_intr_regs;
	evsel__set_sample_bit(evsel, REGS_INTR);
}

So even if the evsel is dummy event, the condition check is true. :(

Or maybe I misunderstand anything?

Thanks
Jin Yao

Quoted 14 lines
Quoted 13 lines
tools/perf/util/evsel.c +2 -1
  		attr->sample_regs_intr = opts->sample_intr_regs;
  		evsel__set_sample_bit(evsel, REGS_INTR);
  	}
 
-	if (opts->sample_user_regs && !evsel->no_aux_samples) {
+	if (opts->sample_user_regs && !evsel->no_aux_samples &&
+	    !evsel__is_dummy_event(evsel)) {
  		attr->sample_regs_user |= opts->sample_user_regs;
  		evsel__set_sample_bit(evsel, REGS_USER);
  	}
Signature · 1 line
2.17.1
#3 lore raw

On Wed, Jul 22, 2020 at 01:00:03PM +0800, Jin, Yao wrote:

SNIP

Quoted 53 lines
Quoted 38 lines
Quoted 30 lines

If we use -IXMM0, the attr>sample_regs_intr will be set with PERF_REG_EXTENDED_MASK bit.

It doesn't make sense to set attr->sample_regs_intr for a software dummy event.

This patch adds dummy event checking before setting attr->sample_regs_intr and attr->sample_regs_user.

After:

   # ./perf record -e cycles:p -IXMM0 -a -- sleep 1
   [ perf record: Woken up 1 times to write data ]
   [ perf record: Captured and wrote 0.413 MB perf.data (45 samples) ]

v2:
---
Rebase to perf/core

Fixes 0a892c1c9472 ("perf record: Add dummy event during system wide synthesis")
---
  tools/perf/util/evsel.c | 6 ++++--
  1 file changed, 4 insertions(+), 2 deletions(-)
tools/perf/util/evsel.c +0 -0
index 9aa51a65593d..11794d3b7879 100644

tools/perf/util/evsel.c void evsel__config(struct evsel *evsel, struct record_opts *opts, · line 1017

Quoted 5 lines
tools/perf/util/evsel.c +2 -1
@@ -1014,12 +1014,14 @@ void evsel__config(struct evsel *evsel, struct record_opts *opts,
  	if (callchain && callchain->enabled && !evsel->no_aux_samples)
  		evsel__config_callchain(evsel, opts, callchain);
-	if (opts->sample_intr_regs && !evsel->no_aux_samples) {
+	if (opts->sample_intr_regs && !evsel->no_aux_samples &&
+	    !evsel__is_dummy_event(evsel)) {

hum, I thought it'd look something like this:

Quoted 10 lines
   if (opts->sample_intr_regs && (!evsel->no_aux_samples || !evsel__is_dummy_event(evsel))

but I'm not sure how no_aux_samples flag works exactly.. so it might be correct.. just making sure ;-)

cc-ing Adrian

jirka

no_aux_samples is set to false by default and it's only set to true by pt, right?

So most of the time, !evsel->no_aux_samples is always true.

Quoted 8 lines
if (opts->sample_intr_regs && (!evsel->no_aux_samples || !evsel__is_dummy_event(evsel)) {
	attr->sample_regs_intr = opts->sample_intr_regs;
	evsel__set_sample_bit(evsel, REGS_INTR);
}

So even if the evsel is dummy event, the condition check is true. :(

Or maybe I misunderstand anything?

I was just curious, because I did not follow the no_aux_samples usage in detail.. so how about a case where:

evsel->no_aux_samples == true and evsel__is_dummy_event(evsel) = false

then the original condition will be false for non dummy event

  (opts->sample_intr_regs && !evsel->no_aux_samples && !evsel__is_dummy_event(evsel))

is that ok?

jirka

author
#4 lore raw

Hi Jiri, Adrian,

Quoted 69 lines · Jiri Olsa

On Wed, Jul 22, 2020 at 01:00:03PM +0800, Jin, Yao wrote:

SNIP

Quoted 53 lines
Quoted 38 lines
Quoted 30 lines

If we use -IXMM0, the attr>sample_regs_intr will be set with PERF_REG_EXTENDED_MASK bit.

It doesn't make sense to set attr->sample_regs_intr for a software dummy event.

This patch adds dummy event checking before setting attr->sample_regs_intr and attr->sample_regs_user.

After:

    # ./perf record -e cycles:p -IXMM0 -a -- sleep 1
    [ perf record: Woken up 1 times to write data ]
    [ perf record: Captured and wrote 0.413 MB perf.data (45 samples) ]
   v2:
   ---
   Rebase to perf/core
Fixes 0a892c1c9472 ("perf record: Add dummy event during system wide synthesis")
---
   tools/perf/util/evsel.c | 6 ++++--
   1 file changed, 4 insertions(+), 2 deletions(-)
tools/perf/util/evsel.c +0 -0
index 9aa51a65593d..11794d3b7879 100644

tools/perf/util/evsel.c void evsel__config(struct evsel *evsel, struct record_opts *opts, · line 1017

Quoted 5 lines
tools/perf/util/evsel.c +2 -1
@@ -1014,12 +1014,14 @@ void evsel__config(struct evsel *evsel, struct record_opts *opts,
   	if (callchain && callchain->enabled && !evsel->no_aux_samples)
   		evsel__config_callchain(evsel, opts, callchain);
-	if (opts->sample_intr_regs && !evsel->no_aux_samples) {
+	if (opts->sample_intr_regs && !evsel->no_aux_samples &&
+	    !evsel__is_dummy_event(evsel)) {

hum, I thought it'd look something like this:

Quoted 10 lines
    if (opts->sample_intr_regs && (!evsel->no_aux_samples || !evsel__is_dummy_event(evsel))

but I'm not sure how no_aux_samples flag works exactly.. so it might be correct.. just making sure ;-)

cc-ing Adrian

jirka

no_aux_samples is set to false by default and it's only set to true by pt, right?

So most of the time, !evsel->no_aux_samples is always true.

Quoted 8 lines
if (opts->sample_intr_regs && (!evsel->no_aux_samples || !evsel__is_dummy_event(evsel)) {
	attr->sample_regs_intr = opts->sample_intr_regs;
	evsel__set_sample_bit(evsel, REGS_INTR);
}

So even if the evsel is dummy event, the condition check is true. :(

Or maybe I misunderstand anything?

I was just curious, because I did not follow the no_aux_samples usage in detail.. so how about a case where:

Quoted 8 lines
    evsel->no_aux_samples == true and evsel__is_dummy_event(evsel) = false

then the original condition will be false for non dummy event

   (opts->sample_intr_regs && !evsel->no_aux_samples && !evsel__is_dummy_event(evsel))

is that ok?

I searched the perf source and found the no_aux_samples was only set to true in intel-pt.c. So I assume for the non-pt usage, the no_aux_samples is always false.

For non-pt usage,
(opts->sample_intr_regs && !evsel->no_aux_samples && !evsel__is_dummy_event(evsel)) is equal to (opts->sample_intr_regs && !evsel__is_dummy_event(evsel))

For pt usage, we need to consider the case that evsel__is_dummy_event(evsel) is true or false.

If evsel__is_dummy_event(evsel) is true:
(opts->sample_intr_regs && !evsel->no_aux_samples && !evsel__is_dummy_event(evsel)) is false. It's expected.

If evsel__is_dummy_event(evsel) is false:
(opts->sample_intr_regs && !evsel->no_aux_samples && !evsel__is_dummy_event(evsel)) is equal to (opts->sample_intr_regs && !evsel->no_aux_samples)
That's the current code logic.

So I think the condition "(opts->sample_intr_regs && !evsel->no_aux_samples && !evsel__is_dummy_event(evsel))" looks reasonable.

Adrian, please correct me if I'm wrong here.

Thanks
Jin Yao

Quoted 2 lines

jirka

author
#5 lore raw

Hi Adrian,

Could you help to check if following condition will break PT?

"(opts->sample_intr_regs && !evsel->no_aux_samples && !evsel__is_dummy_event(evsel))"

Thanks
Jin Yao

Quoted 107 lines · Yao

Hi Jiri, Adrian,

Quoted 69 lines · Jiri Olsa

On Wed, Jul 22, 2020 at 01:00:03PM +0800, Jin, Yao wrote:

SNIP

Quoted 53 lines
Quoted 38 lines
Quoted 30 lines

If we use -IXMM0, the attr>sample_regs_intr will be set with PERF_REG_EXTENDED_MASK bit.

It doesn't make sense to set attr->sample_regs_intr for a software dummy event.

This patch adds dummy event checking before setting attr->sample_regs_intr and attr->sample_regs_user.

After:

    # ./perf record -e cycles:p -IXMM0 -a -- sleep 1
    [ perf record: Woken up 1 times to write data ]
    [ perf record: Captured and wrote 0.413 MB perf.data (45 samples) ]

   v2:
   ---
   Rebase to perf/core

Fixes 0a892c1c9472 ("perf record: Add dummy event during system wide synthesis")

---
   tools/perf/util/evsel.c | 6 ++++--

   1 file changed, 4 insertions(+), 2 deletions(-)
tools/perf/util/evsel.c +0 -0
index 9aa51a65593d..11794d3b7879 100644

tools/perf/util/evsel.c void evsel__config(struct evsel *evsel, struct record_opts *opts, · line 1017

Quoted 5 lines
tools/perf/util/evsel.c +2 -1
@@ -1014,12 +1014,14 @@ void evsel__config(struct evsel *evsel, struct record_opts *opts,
        if (callchain && callchain->enabled && !evsel->no_aux_samples)
            evsel__config_callchain(evsel, opts, callchain);
-    if (opts->sample_intr_regs && !evsel->no_aux_samples) {
+    if (opts->sample_intr_regs && !evsel->no_aux_samples &&
+        !evsel__is_dummy_event(evsel)) {

hum, I thought it'd look something like this:

Quoted 10 lines

    if (opts->sample_intr_regs && (!evsel->no_aux_samples || !evsel__is_dummy_event(evsel))

but I'm not sure how no_aux_samples flag works exactly.. so it might be correct.. just making sure ;-)

cc-ing Adrian

jirka

no_aux_samples is set to false by default and it's only set to true by pt, right?

So most of the time, !evsel->no_aux_samples is always true.

Quoted 8 lines

if (opts->sample_intr_regs && (!evsel->no_aux_samples || !evsel__is_dummy_event(evsel)) { attr->sample_regs_intr = opts->sample_intr_regs; evsel__set_sample_bit(evsel, REGS_INTR);
}

So even if the evsel is dummy event, the condition check is true. :(

Or maybe I misunderstand anything?

I was just curious, because I did not follow the no_aux_samples usage in detail.. so how about a case where:

Quoted 8 lines

    evsel->no_aux_samples == true and evsel__is_dummy_event(evsel) = false

then the original condition will be false for non dummy event

   (opts->sample_intr_regs && !evsel->no_aux_samples && !evsel__is_dummy_event(evsel))

is that ok?

I searched the perf source and found the no_aux_samples was only set to true in intel-pt.c. So I assume for the non-pt usage, the no_aux_samples is always false.

For non-pt usage,
(opts->sample_intr_regs && !evsel->no_aux_samples && !evsel__is_dummy_event(evsel)) is equal to (opts->sample_intr_regs && !evsel__is_dummy_event(evsel))

For pt usage, we need to consider the case that evsel__is_dummy_event(evsel) is true or false.

If evsel__is_dummy_event(evsel) is true:
(opts->sample_intr_regs && !evsel->no_aux_samples && !evsel__is_dummy_event(evsel)) is false. It's expected.

If evsel__is_dummy_event(evsel) is false:
(opts->sample_intr_regs && !evsel->no_aux_samples && !evsel__is_dummy_event(evsel)) is equal to (opts->sample_intr_regs && !evsel->no_aux_samples)
That's the current code logic.

So I think the condition "(opts->sample_intr_regs && !evsel->no_aux_samples && !evsel__is_dummy_event(evsel))" looks reasonable.

Adrian, please correct me if I'm wrong here.

Thanks
Jin Yao

Quoted 2 lines

jirka

reviewer Acked-by
#6 lore raw
Quoted 6 lines · Yao

Hi Adrian,

Could you help to check if following condition will break PT?

"(opts->sample_intr_regs && !evsel->no_aux_samples && !evsel__is_dummy_event(evsel))"

Sorry for slow response - I've been away.

This is fine. It will not break PT.

no_aux_samples is useful for evsels that have been added by the code rather than requested by the user. For old kernels PT adds sched_switch tracepoint to track context switches (before the current context switch event was added) and having auxiliary sample information unnecessarily uses up space in the perf buffer.

Quoted 125 lines

Thanks
Jin Yao

Quoted 120 lines · Yao

Hi Jiri, Adrian,

Quoted 74 lines · Jiri Olsa

On Wed, Jul 22, 2020 at 01:00:03PM +0800, Jin, Yao wrote:

SNIP

Quoted 63 lines
Quoted 43 lines
Quoted 31 lines

If we use -IXMM0, the attr>sample_regs_intr will be set with PERF_REG_EXTENDED_MASK bit.

It doesn't make sense to set attr->sample_regs_intr for a software dummy event.

This patch adds dummy event checking before setting attr->sample_regs_intr and attr->sample_regs_user.

After:

    # ./perf record -e cycles:p -IXMM0 -a -- sleep 1
    [ perf record: Woken up 1 times to write data ]
    [ perf record: Captured and wrote 0.413 MB perf.data (45 samples) ]

   v2:
   ---
   Rebase to perf/core

Fixes 0a892c1c9472 ("perf record: Add dummy event during system wide

synthesis")

---
   tools/perf/util/evsel.c | 6 ++++--

   1 file changed, 4 insertions(+), 2 deletions(-)
tools/perf/util/evsel.c +0 -0
index 9aa51a65593d..11794d3b7879 100644

tools/perf/util/evsel.c void evsel__config(struct evsel *evsel, struct · line 1018

Quoted 6 lines
tools/perf/util/evsel.c +2 -1
@@ -1014,12 +1014,14 @@ void evsel__config(struct evsel *evsel, struct
 record_opts *opts,
        if (callchain && callchain->enabled && !evsel->no_aux_samples)
            evsel__config_callchain(evsel, opts, callchain);
-    if (opts->sample_intr_regs && !evsel->no_aux_samples) {
+    if (opts->sample_intr_regs && !evsel->no_aux_samples &&
+        !evsel__is_dummy_event(evsel)) {

hum, I thought it'd look something like this:

    if (opts->sample_intr_regs && (!evsel->no_aux_samples || !evsel__is_dummy_event(evsel))

Quoted 8 lines

but I'm not sure how no_aux_samples flag works exactly.. so it might be correct.. just making sure ;-)

cc-ing Adrian

jirka

no_aux_samples is set to false by default and it's only set to true by pt, right?

So most of the time, !evsel->no_aux_samples is always true.

if (opts->sample_intr_regs && (!evsel->no_aux_samples || !evsel__is_dummy_event(evsel)) { attr->sample_regs_intr = opts->sample_intr_regs; evsel__set_sample_bit(evsel, REGS_INTR);
}

Quoted 3 lines

So even if the evsel is dummy event, the condition check is true. :(

Or maybe I misunderstand anything?

I was just curious, because I did not follow the no_aux_samples usage in detail.. so how about a case where:

Quoted 9 lines

    evsel->no_aux_samples == true and evsel__is_dummy_event(evsel) = false

then the original condition will be false for non dummy event

   (opts->sample_intr_regs && !evsel->no_aux_samples && !evsel__is_dummy_event(evsel))

is that ok?

I searched the perf source and found the no_aux_samples was only set to true in intel-pt.c. So I assume for the non-pt usage, the no_aux_samples is always false.

For non-pt usage,
(opts->sample_intr_regs && !evsel->no_aux_samples && !evsel__is_dummy_event(evsel)) is equal to (opts->sample_intr_regs && !evsel__is_dummy_event(evsel))

For pt usage, we need to consider the case that
evsel__is_dummy_event(evsel) is true or false.

If evsel__is_dummy_event(evsel) is true:
(opts->sample_intr_regs && !evsel->no_aux_samples && !evsel__is_dummy_event(evsel)) is false. It's expected.

If evsel__is_dummy_event(evsel) is false:
(opts->sample_intr_regs && !evsel->no_aux_samples && !evsel__is_dummy_event(evsel)) is equal to (opts->sample_intr_regs && !evsel->no_aux_samples) That's the current code logic.

So I think the condition "(opts->sample_intr_regs && !evsel->no_aux_samples && !evsel__is_dummy_event(evsel))" looks reasonable.

Adrian, please correct me if I'm wrong here.

Thanks
Jin Yao

Quoted 2 lines

jirka

#7 lore raw

Em Tue, Aug 04, 2020 at 10:06:56AM +0300, Adrian Hunter escreveu:

Quoted 12 lines · Yao
Quoted 6 lines

Hi Adrian,

Could you help to check if following condition will break PT?

"(opts->sample_intr_regs && !evsel->no_aux_samples && !evsel__is_dummy_event(evsel))"

Sorry for slow response - I've been away.

This is fine. It will not break PT.

Quoted 7 lines

no_aux_samples is useful for evsels that have been added by the code rather than requested by the user. For old kernels PT adds sched_switch tracepoint to track context switches (before the current context switch event was added) and having auxiliary sample information unnecessarily uses up space in the perf buffer.

Thanks for checking and providing the comment, that I added as a committer note together with your Acked-by, appreciated.

- Arnaldo