On Wed, Sep 03, 2025 at 01:59:13PM +0200, Oleg Nesterov wrote:
> Slightly off-topic, but
>
> On 09/02, Jiri Olsa wrote:
> >
> > @@ -11144,7 +11147,7 @@ static int perf_uprobe_event_init(struct perf_event *event)
> > {
> > int err;
> > unsigned long ref_ctr_offset;
> > - bool is_retprobe;
> > + bool is_retprobe, is_unique;
> >
> > if (event->attr.type != perf_uprobe.type)
> > return -ENOENT;
> > @@ -11159,8 +11162,9 @@ static int perf_uprobe_event_init(struct perf_event *event)
> > return -EOPNOTSUPP;
> >
> > is_retprobe = event->attr.config & PERF_PROBE_CONFIG_IS_RETPROBE;
> > + is_unique = event->attr.config & PERF_PROBE_CONFIG_IS_UNIQUE;
> > ref_ctr_offset = event->attr.config >> PERF_UPROBE_REF_CTR_OFFSET_SHIFT;
> > - err = perf_uprobe_init(event, ref_ctr_offset, is_retprobe);
> > + err = perf_uprobe_init(event, ref_ctr_offset, is_retprobe, is_unique);
>
> I am wondering why (with or without this change) perf_uprobe_init() needs
> the additional arguments besides "event". It can look at event->attr.config
> itself?
>
> Same for perf_kprobe_init()...
I think that's because we define enum perf_probe_config together
with PMU_FORMAT_ATTRs and code for attr->config parsing, which
makes sense to me
otherwise I think we could pass perf_event_attr all the way to
create_local_trace_[ku]probe
jirka
