Skip to contentSkip to conversation

perf test: Skip test_arm_callgraph_fp.sh if unwinding isn't built in

PATCH Applied James Clark 3 participants 7 messages Jan 23, 2024 – Jan 25, 2024 perf events

lore
Share & tools
lore
#0 lore raw

cover perf test: Skip test_arm_callgraph_fp.sh if unwinding isn't built in

The first commit outputs the unwinding feature definition in perf version so that the script can gate on it. And then skip the test if it's not present.

I didn't add any fixes tags because it's modifying the perf version output, requires both commits, and it's just for a test skip change so I don't think backporting makes sense.

James Clark (2):
perf version: Display availability of HAVE_DWARF_UNWIND_SUPPORT perf test: Skip test_arm_callgraph_fp.sh if unwinding isn't built in

 tools/perf/builtin-version.c                    | 1 +
 tools/perf/tests/shell/test_arm_callgraph_fp.sh | 6 ++++++
 2 files changed, 7 insertions(+)
Signature · 1 line
2.34.1
#1 lore raw

PATCH 1/2 perf version: Display availability of HAVE_DWARF_UNWIND_SUPPORT

Even though unwinding depends on either HAVE_DWARF_SUPPORT or HAVE_LIBUNWIND, scripts testing unwinding can't just look for the existence of either of those flags. This is because NO_LIBDW_DWARF_UNWIND=1 can disable unwinding with libdw, but libdw will still be linked leaving HAVE_DWARF_SUPPORT turned on. Presumably because it is used for things other than unwinding, so I don't think this needs to be fixed.

HAVE_DWARF_UNWIND_SUPPORT already takes the combination of all those things into account, and is used to gate the built in tests like "Test dwarf unwind", so add it to the feature list output so that it can be used by the script tests too.

---
 tools/perf/builtin-version.c | 1 +
 1 file changed, 1 insertion(+)
tools/perf/builtin-version.c +1 -0
index ac20c2b9bbc2..529e9ce8c46c 100644
@@ -82,6 +82,7 @@ static void library_status(void)
 	STATUS(HAVE_LIBPFM, libpfm4);
 	STATUS(HAVE_LIBTRACEEVENT, libtraceevent);
 	STATUS(HAVE_BPF_SKEL, bpf_skeletons);
+	STATUS(HAVE_DWARF_UNWIND_SUPPORT, dwarf-unwind-support);
 }
 
 int cmd_version(int argc, const char **argv)
Signature · 1 line
2.34.1
#2 lore raw

PATCH 2/2 perf test: Skip test_arm_callgraph_fp.sh if unwinding isn't built in

In reply to James Clark

Even though this is a frame pointer unwind test, it's testing that a frame pointer stack can be augmented correctly with a partial Dwarf unwind. So add a feature check so that this test skips instead of fails if Dwarf unwinding isn't present.

---
 tools/perf/tests/shell/test_arm_callgraph_fp.sh | 6 ++++++
 1 file changed, 6 insertions(+)
tools/perf/tests/shell/test_arm_callgraph_fp.sh +6 -0
index e342e6c8aa50..83b53591b1ea 100755
@@ -8,6 +8,12 @@ shelldir=$(dirname "$0")
 
 lscpu | grep -q "aarch64" || exit 2
 
+if perf version --build-options | grep HAVE_DWARF_UNWIND_SUPPORT | grep -q OFF
+then
+  echo "Skipping, no dwarf unwind support"
+  exit 2
+fi
+
 skip_test_missing_symbol leafloop
 
 PERF_DATA=$(mktemp /tmp/__perf_test.perf.data.XXXXX)
Signature · 1 line
2.34.1
reviewer
#3 lore raw
Quoted 5 lines · James Clark

Even though this is a frame pointer unwind test, it's testing that a frame pointer stack can be augmented correctly with a partial Dwarf unwind. So add a feature check so that this test skips instead of fails if Dwarf unwinding isn't present.

Hi James,

Is there value in testing without the partial Dwarf unwind? Presumably that is covered by the existing dwarf unwind test?
https://git.kernel.org/pub/scm/linux/kernel/git/perf/perf-tools-next.git/tree/tools/perf/tests/dwarf-unwind.c?h=perf-tools-next
If the issue is inlined functions I'm surprised addr2line isn't doing the job properly. Is there an addr2line perf script issue here?

Thanks,
Ian

Quoted 25 lines
---
 tools/perf/tests/shell/test_arm_callgraph_fp.sh | 6 ++++++
 1 file changed, 6 insertions(+)
tools/perf/tests/shell/test_arm_callgraph_fp.sh +6 -1
index e342e6c8aa50..83b53591b1ea 100755
@@ -8,6 +8,12 @@ shelldir=$(dirname "$0")
 
 lscpu | grep -q "aarch64" || exit 2
 
+if perf version --build-options | grep HAVE_DWARF_UNWIND_SUPPORT | grep -q OFF
+then
+  echo "Skipping, no dwarf unwind support"
+  exit 2
+fi
+
 skip_test_missing_symbol leafloop
 
 PERF_DATA=$(mktemp /tmp/__perf_test.perf.data.XXXXX)
--

2.34.1

#4 lore raw
Quoted 10 lines · Ian Rogers
Quoted 5 lines · James Clark

Even though this is a frame pointer unwind test, it's testing that a frame pointer stack can be augmented correctly with a partial Dwarf unwind. So add a feature check so that this test skips instead of fails if Dwarf unwinding isn't present.

Hi James,

Is there value in testing without the partial Dwarf unwind? Presumably

Yeah I think we could add a test for just --call-graph=fp, I don't think there is one. But that would be separate to this test, and would be redundant if the tests are run with a dwarf unwinder present because this test already requires the frame pointer unwinder to be correct.

Quoted 1 line

that is covered by the existing dwarf unwind test?

There is no overlap, this test test is for --call-graph=fp, and the dwarf test is for --call-graph=dwarf

Quoted 4 lines

https://git.kernel.org/pub/scm/linux/kernel/git/perf/perf-tools-next.git/tree/tools/perf/tests/dwarf-unwind.c?h=perf-tools-next
If the issue is inlined functions I'm surprised addr2line isn't doing the job properly. Is there an addr2line perf script issue here?

The issue isn't inlined functions, it's when the leaf frame doesn't insert a frame pointer. In that case we use the link register to see what the parent function of the leaf frame was and insert it into the frame pointer stack.

Dwarf is only used in this case to confirm if the link register was valid at that instruction.

See commit b9f6fbb for more info. Long story short this test was only added for that feature and it requires a dwarf unwinder to pass despite being called test_arm_callgraph_fp

Quoted 30 lines

Thanks,
Ian

Quoted 25 lines
---
 tools/perf/tests/shell/test_arm_callgraph_fp.sh | 6 ++++++
 1 file changed, 6 insertions(+)
tools/perf/tests/shell/test_arm_callgraph_fp.sh +6 -1
index e342e6c8aa50..83b53591b1ea 100755
@@ -8,6 +8,12 @@ shelldir=$(dirname "$0")
 
 lscpu | grep -q "aarch64" || exit 2
 
+if perf version --build-options | grep HAVE_DWARF_UNWIND_SUPPORT | grep -q OFF
+then
+  echo "Skipping, no dwarf unwind support"
+  exit 2
+fi
+
 skip_test_missing_symbol leafloop
 
 PERF_DATA=$(mktemp /tmp/__perf_test.perf.data.XXXXX)
--

2.34.1

reviewer Reviewed-by
#5 lore raw
Quoted 35 lines · James Clark
Quoted 10 lines · Ian Rogers
Quoted 5 lines · James Clark

Even though this is a frame pointer unwind test, it's testing that a frame pointer stack can be augmented correctly with a partial Dwarf unwind. So add a feature check so that this test skips instead of fails if Dwarf unwinding isn't present.

Hi James,

Is there value in testing without the partial Dwarf unwind? Presumably

Yeah I think we could add a test for just --call-graph=fp, I don't think there is one. But that would be separate to this test, and would be redundant if the tests are run with a dwarf unwinder present because this test already requires the frame pointer unwinder to be correct.

Quoted 1 line

that is covered by the existing dwarf unwind test?

There is no overlap, this test test is for --call-graph=fp, and the dwarf test is for --call-graph=dwarf

Quoted 4 lines

https://git.kernel.org/pub/scm/linux/kernel/git/perf/perf-tools-next.git/tree/tools/perf/tests/dwarf-unwind.c?h=perf-tools-next
If the issue is inlined functions I'm surprised addr2line isn't doing the job properly. Is there an addr2line perf script issue here?

The issue isn't inlined functions, it's when the leaf frame doesn't insert a frame pointer. In that case we use the link register to see what the parent function of the leaf frame was and insert it into the frame pointer stack.

Quoted 6 lines

Dwarf is only used in this case to confirm if the link register was valid at that instruction.

See commit b9f6fbb for more info. Long story short this test was only added for that feature and it requires a dwarf unwinder to pass despite being called test_arm_callgraph_fp

Ok, not directly seeing b9f6fbb being dependent on HAVE_DWARF_UNWIND_SUPPORT. I'll assume I'm ignorant and the fix here is obviously a workaround.

For the series:

Thanks,
Ian

Quoted 30 lines
Quoted 30 lines

Thanks,
Ian

Quoted 25 lines
---
 tools/perf/tests/shell/test_arm_callgraph_fp.sh | 6 ++++++
 1 file changed, 6 insertions(+)
tools/perf/tests/shell/test_arm_callgraph_fp.sh +6 -1
index e342e6c8aa50..83b53591b1ea 100755
@@ -8,6 +8,12 @@ shelldir=$(dirname "$0")
 
 lscpu | grep -q "aarch64" || exit 2
 
+if perf version --build-options | grep HAVE_DWARF_UNWIND_SUPPORT | grep -q OFF
+then
+  echo "Skipping, no dwarf unwind support"
+  exit 2
+fi
+
 skip_test_missing_symbol leafloop
 
 PERF_DATA=$(mktemp /tmp/__perf_test.perf.data.XXXXX)
--

2.34.1

maintainer Applied
#6 lore raw
In reply to James Clark
Quoted 9 lines · James Clark

The first commit outputs the unwinding feature definition in perf version so that the script can gate on it. And then skip the test if it's not present.

I didn't add any fixes tags because it's modifying the perf version output, requires both commits, and it's just for a test skip change so I don't think backporting makes sense.

[...]

Applied to perf-tools-next, thanks!

Best regards,

Signature · 1 line
Namhyung Kim <[email protected]>