On Thu, Aug 17, 2023 at 06:24:19PM -0400, Steven Rostedt wrote: > From: "Steven Rostedt (Google)" <rostedt@xxxxxxxxxxx> > > Now that the traceeval_data has a type, add checks to traceeval_insert() > and traceveal_query() as well as traceeval_stat() to make sure the keys > and vals match the types that were for the traceeval. > > Signed-off-by: Steven Rostedt (Google) <rostedt@xxxxxxxxxxx> > --- > src/histograms.c | 12 ++++++++++++ > 1 file changed, 12 insertions(+) > > diff --git a/src/histograms.c b/src/histograms.c > index 7c0eef6cf421..560046cc8d96 100644 > --- a/src/histograms.c > +++ b/src/histograms.c > @@ -479,10 +479,16 @@ static int get_entry(struct traceeval *teval, const struct traceeval_data *keys, > struct hash_item *item; > unsigned key; > int check = 0; > + int i; > > if (!teval || !keys) > return -1; > > + for (i = 0; i < teval->nr_key_types; i++) { > + if (keys[i].type != teval->key_types[i].type) > + return -1; > + } > + > key = make_hash(teval, keys, hist->bits); > > for (iter = hash_iter_bucket(hist, key); (item = hash_iter_bucket_next(iter)); ) { > @@ -928,10 +934,16 @@ int traceeval_insert(struct traceeval *teval, > { > struct entry *entry; > int check; > + int i; > > entry = NULL; > check = get_entry(teval, keys, &entry); > > + for (i = 0; i < teval->nr_val_types; i++) { > + if (vals[i].type != teval->val_types[i].type) > + return -1; > + } I think we should have a similar type check in the update path, so do all the checks in a loop like this in update_entry(), or we can check each entry individually in copy_traceeval_data(). That will make sure we don't update an entry with mismatched typed values. > + > if (check == -1) > return check; > > -- > 2.40.1 >
![]() |