Skip to contentSkip to conversation

perf script: Fix Cannot print 'iregs' field for hybrid systems

PATCH Applied [email protected] 3 participants 3 messages Sep 8, 2022 perf events

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 perf script: Fix Cannot print 'iregs' field for hybrid systems

From: Zhengjun Xing <[email protected]>

Commit b91e5492f9d7 ("perf record: Add a dummy event on hybrid systems to collect metadata records") adds a dummy event on hybrid systems to fix the symbol "unknown" issue when the workload is created in a P-core but runs on an E-core. The added dummy event will cause "perf script -F iregs" to fail. Dummy events do not have "iregs" attribute set, so when we do evsel__check_attr, the "iregs" attribute check will fail, so the issue happened.

The following commit [1] has fixed a similar issue by skipping the attr check for the dummy event because it does not have any samples anyway. It works okay for the normal mode, but the issue still happened when running the test in the pipe mode. In the pipe mode, it calls process_attr() which still checks the attr for the dummy event. This commit fixed the issue by skipping the attr check for the dummy event in the API evsel__check_attr, Otherwise, we have to patch everywhere when evsel__check_attr() is called.

Before:

#./perf record -o - --intr-regs=di,r8,dx,cx -e br_inst_retired.near_call:p -c 1000 --per-thread true 2>/dev/null|./perf script -F iregs |head -5
Samples for 'dummy:HG' event do not have IREGS attribute set. Cannot print 'iregs' field.
0x120 [0x90]: failed to process type: 64

After:

 # ./perf record -o - --intr-regs=di,r8,dx,cx -e br_inst_retired.near_call:p -c 1000 --per-thread true 2>/dev/null|./perf script -F iregs |head -5
 ABI:2    CX:0x55b8efa87000    DX:0x55b8efa7e000    DI:0xffffba5e625efbb0    R8:0xffff90e51f8ae100
 ABI:2    CX:0x7f1dae1e4000    DX:0xd0    DI:0xffff90e18c675ac0    R8:0x71
 ABI:2    CX:0xcc0    DX:0x1    DI:0xffff90e199880240    R8:0x0
 ABI:2    CX:0xffff90e180dd7500    DX:0xffff90e180dd7500    DI:0xffff90e180043500    R8:0x1
 ABI:2    CX:0x50    DX:0xffff90e18c583bd0    DI:0xffff90e1998803c0    R8:0x58

[1]https://lore.kernel.org/lkml/[email protected]/

Fixes b91e5492f9d7 ("perf record: Add a dummy event on hybrid systems to collect metadata records")
---
 tools/perf/builtin-script.c | 3 +++
 1 file changed, 3 insertions(+)
tools/perf/builtin-script.c +3 -0
index 585171479876..1a8790385eba 100644
@@ -445,6 +445,9 @@ static int evsel__check_attr(struct evsel *evsel, struct perf_session *session)
 	struct perf_event_attr *attr = &evsel->core.attr;
 	bool allow_user_set;
 
+	if (evsel__is_dummy_event(evsel))
+		return 0;
+
 	if (perf_header__has_feat(&session->header, HEADER_STAT))
 		return 0;
 
Signature · 1 line
2.25.1
Acked-by on v1 1/1
#1 lore raw
Quoted 33 lines · On Thu, Sep 08, 2022 at 03:00:30PM +0800, [email protected] wrote:

From: Zhengjun Xing <[email protected]>

Commit b91e5492f9d7 ("perf record: Add a dummy event on hybrid systems to collect metadata records") adds a dummy event on hybrid systems to fix the symbol "unknown" issue when the workload is created in a P-core but runs on an E-core. The added dummy event will cause "perf script -F iregs" to fail. Dummy events do not have "iregs" attribute set, so when we do evsel__check_attr, the "iregs" attribute check will fail, so the issue happened.

The following commit [1] has fixed a similar issue by skipping the attr check for the dummy event because it does not have any samples anyway. It works okay for the normal mode, but the issue still happened when running the test in the pipe mode. In the pipe mode, it calls process_attr() which still checks the attr for the dummy event. This commit fixed the issue by skipping the attr check for the dummy event in the API evsel__check_attr, Otherwise, we have to patch everywhere when evsel__check_attr() is called.

Before:

#./perf record -o - --intr-regs=di,r8,dx,cx -e br_inst_retired.near_call:p -c 1000 --per-thread true 2>/dev/null|./perf script -F iregs |head -5
Samples for 'dummy:HG' event do not have IREGS attribute set. Cannot print 'iregs' field.
0x120 [0x90]: failed to process type: 64

After:

 # ./perf record -o - --intr-regs=di,r8,dx,cx -e br_inst_retired.near_call:p -c 1000 --per-thread true 2>/dev/null|./perf script -F iregs |head -5
 ABI:2    CX:0x55b8efa87000    DX:0x55b8efa7e000    DI:0xffffba5e625efbb0    R8:0xffff90e51f8ae100
 ABI:2    CX:0x7f1dae1e4000    DX:0xd0    DI:0xffff90e18c675ac0    R8:0x71
 ABI:2    CX:0xcc0    DX:0x1    DI:0xffff90e199880240    R8:0x0
 ABI:2    CX:0xffff90e180dd7500    DX:0xffff90e180dd7500    DI:0xffff90e180043500    R8:0x1
 ABI:2    CX:0x50    DX:0xffff90e18c583bd0    DI:0xffff90e1998803c0    R8:0x58
Quoted 6 lines

[1]https://lore.kernel.org/lkml/[email protected]/

Fixes b91e5492f9d7 ("perf record: Add a dummy event on hybrid systems to collect metadata records")

thanks,
jirka

Quoted 21 lines
---
 tools/perf/builtin-script.c | 3 +++
 1 file changed, 3 insertions(+)
tools/perf/builtin-script.c +3 -0
index 585171479876..1a8790385eba 100644
@@ -445,6 +445,9 @@ static int evsel__check_attr(struct evsel *evsel, struct perf_session *session)
 	struct perf_event_attr *attr = &evsel->core.attr;
 	bool allow_user_set;
 
+	if (evsel__is_dummy_event(evsel))
+		return 0;
+
 	if (perf_header__has_feat(&session->header, HEADER_STAT))
 		return 0;
 
Signature · 1 line
2.25.1
maintainer Applied
#2 lore raw

Em Thu, Sep 08, 2022 at 10:24:27AM +0200, Jiri Olsa escreveu:

Quoted 34 lines · On Thu, Sep 08, 2022 at 03:00:30PM +0800, [email protected] wrote:
Quoted 33 lines

From: Zhengjun Xing <[email protected]>

Commit b91e5492f9d7 ("perf record: Add a dummy event on hybrid systems to collect metadata records") adds a dummy event on hybrid systems to fix the symbol "unknown" issue when the workload is created in a P-core but runs on an E-core. The added dummy event will cause "perf script -F iregs" to fail. Dummy events do not have "iregs" attribute set, so when we do evsel__check_attr, the "iregs" attribute check will fail, so the issue happened.

The following commit [1] has fixed a similar issue by skipping the attr check for the dummy event because it does not have any samples anyway. It works okay for the normal mode, but the issue still happened when running the test in the pipe mode. In the pipe mode, it calls process_attr() which still checks the attr for the dummy event. This commit fixed the issue by skipping the attr check for the dummy event in the API evsel__check_attr, Otherwise, we have to patch everywhere when evsel__check_attr() is called.

Before:

#./perf record -o - --intr-regs=di,r8,dx,cx -e br_inst_retired.near_call:p -c 1000 --per-thread true 2>/dev/null|./perf script -F iregs |head -5
Samples for 'dummy:HG' event do not have IREGS attribute set. Cannot print 'iregs' field.
0x120 [0x90]: failed to process type: 64

After:

 # ./perf record -o - --intr-regs=di,r8,dx,cx -e br_inst_retired.near_call:p -c 1000 --per-thread true 2>/dev/null|./perf script -F iregs |head -5
 ABI:2    CX:0x55b8efa87000    DX:0x55b8efa7e000    DI:0xffffba5e625efbb0    R8:0xffff90e51f8ae100
 ABI:2    CX:0x7f1dae1e4000    DX:0xd0    DI:0xffff90e18c675ac0    R8:0x71
 ABI:2    CX:0xcc0    DX:0x1    DI:0xffff90e199880240    R8:0x0
 ABI:2    CX:0xffff90e180dd7500    DX:0xffff90e180dd7500    DI:0xffff90e180043500    R8:0x1
 ABI:2    CX:0x50    DX:0xffff90e18c583bd0    DI:0xffff90e1998803c0    R8:0x58
Quoted 8 lines
Quoted 6 lines

[1]https://lore.kernel.org/lkml/[email protected]/

Fixes b91e5492f9d7 ("perf record: Add a dummy event on hybrid systems to collect metadata records")

Thanks, applied.

- Arnaldo