[PATCH] perf python: Accept events larger than union perf_event

[PATCH] perf python: Accept events larger than union perf_event

[Lalit Maganti]

Samples recorded with --call-graph dwarf carry a copy of the user stack
and are larger than union perf_event, so session.process_events()
fails on the first one:

  $ perf record --call-graph dwarf -- ./workload
  $ python3 tools/perf/python/stackcollapse.py
  ...
  TypeError: Unexpected event size: 4168 < 8424

This makes all the standalone scripts unusable on such recordings.

pyrf_event__new() rejects events larger than union perf_event, which
guarded the copy into the union that pyrf_event used to embed. Since
commit 2c672ffdad4b ("perf python: Lazily copy events and samples from
process_events") the event is borrowed, or copied to a buffer of
header.size bytes when a callback keeps it, so nothing depends on the
event fitting in the union any more. Remove the check.

With this, stackcollapse.py and flamegraph.py process a DWARF recording
of a two-thread workload, with the same per-symbol sample counts as
perf script, also when the callback keeps the events after returning.

Assisted-by: LLM
Signed-off-by: Lalit Maganti <lalitmaganti@gmail.com>
---
 tools/perf/util/python.c | 5 -----
 1 file changed, 5 deletions(-)

diff --git a/tools/perf/util/python.c b/tools/perf/util/python.c
index 60dbe2dad5eb..8b4b2f3d0926 100644
--- a/tools/perf/util/python.c
+++ b/tools/perf/util/python.c
@@ -1792,11 +1792,6 @@ static PyObject *pyrf_event__new(const union perf_event *event, struct evsel *ev
 				    event->header.size, event->header.type);
 	}
 
-	if (event->header.size > sizeof(union perf_event)) {
-		return PyErr_Format(PyExc_TypeError, "Unexpected event size: %zu < %u",
-				    sizeof(union perf_event), event->header.size);
-	}
-
 	pevent = PyObject_New(struct pyrf_event, pyrf_event__type[event->header.type]);
 	if (pevent == NULL)
 		return PyErr_NoMemory();

base-commit: 1c90f0a3c8a1c4e38f9edd97af9e20848805a7fb
-- 
2.56.0

Re: [PATCH] perf python: Accept events larger than union perf_event

[sashiko-bot]

> Samples recorded with --call-graph dwarf carry a copy of the user stack
> and are larger than union perf_event, so session.process_events()
> fails on the first one:
> 
>   $ perf record --call-graph dwarf -- ./workload
>   $ python3 tools/perf/python/stackcollapse.py
>   ...
>   TypeError: Unexpected event size: 4168 < 8424
> 
> This makes all the standalone scripts unusable on such recordings.
> 
> pyrf_event__new() rejects events larger than union perf_event, which
> guarded the copy into the union that pyrf_event used to embed. Since
> commit 2c672ffdad4b ("perf python: Lazily copy events and samples from
> process_events") the event is borrowed, or copied to a buffer of
> header.size bytes when a callback keeps it, so nothing depends on the
> event fitting in the union any more. Remove the check.
> 
> With this, stackcollapse.py and flamegraph.py process a DWARF recording
> of a two-thread workload, with the same per-symbol sample counts as
> perf script, also when the callback keeps the events after returning.
> 
> Assisted-by: LLM
> Signed-off-by: Lalit Maganti <lalitmaganti@gmail.com>

Sashiko has reviewed this patch and found no issues.

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20261018074528.974953-1-lalitmaganti@gmail.com?part=1