Re: [PATCH] perf inject: correct event attribute sizes

Em Wed, Dec 23, 2020 at 11:11:05PM +0100, Jiri Olsa escreveu:

On Tue, Dec 15, 2020 at 10:58:16PM -0800, Denis Nikitin wrote:

quoted

When perf inject reads a perf.data file from an older version of perf,
it writes event attributes into the output with the original size field,
but lays them out as if they had the size currently used. Readers see
a corrupt file. Update the size field to match the layout.

Signed-off-by: Al Grant <redacted>
Signed-off-by: Denis Nikitin <redacted>
---
 tools/perf/util/header.c | 8 ++++++++
 1 file changed, 8 insertions(+)

diff --git a/tools/perf/util/header.c b/tools/perf/util/header.c
index be850e9f8852..0d95981df8dd 100644
--- a/tools/perf/util/header.c
+++ b/tools/perf/util/header.c

@@ -3331,6 +3331,14 @@ int perf_session__write_header(struct perf_session *session,
 	attr_offset = lseek(ff.fd, 0, SEEK_CUR);
 
 	evlist__for_each_entry(evlist, evsel) {
+		if (evsel->core.attr.size < sizeof(evsel->core.attr)) {
+			/*
+			 * We are likely in "perf inject" and have read
+			 * from an older file. Update attr size so that
+			 * reader gets the right offset to the ids.
+			 */
+			evsel->core.attr.size = sizeof(evsel->core.attr);
+		}

seems ok, just wondering if it would be better
to fix it in perf_event__process_attr instead

where we know where the data is coming from,
here we could cover for some unexpected case
of having different attr.size?

anyway both would be probably equal, I'm ok
with either way

Acked-by: Jiri Olsa <redacted>

Thanks, applied.

- Arnaldo