On Mon, Jul 26, 2021, at 13:12, Andrii Nakryiko wrote: > Wire through user_ctx for all attach APIs that use perf_event_open under the > hood: > - for kprobes, extend existing bpf_kprobe_opts with user_ctx field; > - for perf_event, uprobe, and tracepoint APIs, add their _opts variants and > pass user_ctx through opts. > > For kernel that don't support BPF_LINK_CREATE for perf_events, and thus > user_ctx is not supported either, return error and log warning for user. > > Cc: Rafael David Tinoco <rafaeldtinoco@xxxxxxxxx> > Signed-off-by: Andrii Nakryiko <andrii@xxxxxxxxxx> I think this one is fuzzy in v2. Checking them now for my purposes. Thanks for CC'ing.
