Re: [PATCH] perf jevents: Check the JSON keys that jevents.py never reads

[sashiko-bot@kernel.org]

Thank you for your contribution! Sashiko AI review found 2 potential
issues to consider:
- [Low] perf jevents: crash in _read_json_events_impl() when MetricExpr
  is mistyped
- [Low] perf jevents: preprocess_arch_std_files() drops parse error in
  --check mode
--

commit b85ceb11b01de607fd5251f8da36bf8900130730
Author: TAKANO Mitsuhiro <takano32@gmail.com>

perf jevents: Check the JSON keys that jevents.py never reads

This commit adds a --check mode to jevents.py (and a jevents-check make
target) that tracks unread JSON keys in JsonEvent and reports keys that
closely match field names read by jevents.py.

> diff --git a/tools/perf/pmu-events/jevents.py b/tools/perf/pmu-events/jevents.py
> index 080e04b51b8c5..56859e983a4b0 100755
> --- a/tools/perf/pmu-events/jevents.py
> +++ b/tools/perf/pmu-events/jevents.py

[ ... ]

> @@ -493,7 +653,10 @@ def preprocess_arch_std_files(archpath: str) -> None:
>          if event.metric_name:
>            _arch_std_events[event.metric_name.lower()] = event
>      except Exception as e:
> -        raise RuntimeError(f'Failure processing \'{item.name}\' in \'{archpath}\'') from e
> +        failure = RuntimeError(f'Failure processing \'{item.name}\' in \'{archpath}\'')
> +        if failures is None:
> +          raise failure from e
> +        failures.append((item.path, failure))

[Severity: Low]
Does appending failure instead of e here drop the underlying parse error when
--check reports failures from preprocess_arch_std_files()?

In main(), each entry in failed is printed with str(error), which does not
follow __cause__ and only prints "Failure processing '<name>' in '<archpath>'"
alongside the relative path, hiding why the JSON file failed to load.

[ ... ]

> @@ -1490,12 +1660,111 @@ such as "arm/cortex-a34".''',

[ ... ]

> +    with concurrent.futures.ProcessPoolExecutor(max_workers=_parallel_worker_count(len(tasks)),
> +                                                initializer=_init_worker,
> +                                                initargs=(_arch_std_events, _args.check)) as executor:
> +      for path, topic, events in executor.map(_parallel_read_json_events, tasks):
> +        if isinstance(events, Exception):
> +          if not _args.check:
> +            raise RuntimeError(f'Failure processing \'{path}\'') from events
> +          failed.append((path, events))
> +        else:
> +          _json_cache[(path, topic)] = events

[Severity: Low]
What happens in --check mode when a metric object has a valid MetricName
(without '-') but misspells MetricExpr (for example as MetricExp)?

JsonEvent.__init__() leaves self.metric_expr as None and records the mistyped
key in self.mistyped, but _read_json_events_impl() then passes that None
expression into metric.RewriteMetricsInTermsOfOthers():

tools/perf/pmu-events/jevents.py:_read_json_events_impl() {
    ...
  for event in events:
    event.topic = topic
    if event.metric_name and '-' not in event.metric_name:
      metrics.append((event.pmu, event.metric_name, event.metric_expr))
  updates = metric.RewriteMetricsInTermsOfOthers(metrics)
    ...
}

RewriteMetricsInTermsOfOthers() raises an AttributeError on None before
_read_json_events_impl() can return events, so main() records the file in
failed instead of _json_cache and --check reports an AttributeError (exit
status 2) instead of reporting the misspelled MetricExpr key.  Checking that
event.metric_expr is not None before appending to metrics would let --check
report the misspelling.

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20261110193239.1741014-1-takano32@gmail.com?part=1