| Message ID | 20230302010051.044209550@goodmis.org (mailing list archive) |
|---|---|
| Headers | show |
| Series | tracing: Fix adding some modifiers to histogram values | expand |
I forgot to add Tom Zanussi on this series. -- Steve On Wed, 01 Mar 2023 20:00:51 -0500 Steven Rostedt <rostedt@goodmis.org> wrote: > Mark Rutland reported to me on IRC that he accidentally added the > ".buckets=8" modifier to hitcount. This should not be allowed, but it > did not error. Worse yet, when reading the hist file, it would crash > as there was a NULL pointer dereference due to the values not having > fields assigned to them. > > The first fix is to make sure that histogram values do not get assigned > modifiers that they can't use. > > The the second patch is to not crash if a NULL pointer is passed to > hist_field_name() (which is what happens if you allow some of these > modifiers to be used by values). > > Steven Rostedt (Google) (2): > tracing: Do not let histogram values have some modifiers > tracing: Check field value in hist_field_name() > > ---- > kernel/trace/trace_events_hist.c | 12 ++++++++++++ > 1 file changed, 12 insertions(+)
Mark Rutland reported to me on IRC that he accidentally added the ".buckets=8" modifier to hitcount. This should not be allowed, but it did not error. Worse yet, when reading the hist file, it would crash as there was a NULL pointer dereference due to the values not having fields assigned to them. The first fix is to make sure that histogram values do not get assigned modifiers that they can't use. The the second patch is to not crash if a NULL pointer is passed to hist_field_name() (which is what happens if you allow some of these modifiers to be used by values). Steven Rostedt (Google) (2): tracing: Do not let histogram values have some modifiers tracing: Check field value in hist_field_name() ---- kernel/trace/trace_events_hist.c | 12 ++++++++++++ 1 file changed, 12 insertions(+)