On Thu, 12 Aug 2021 10:27:35 +0900 Masami Hiramatsu <mhiramat@xxxxxxxxxx> wrote: > I like to prohibit latter one. It is my feeling, but I think it is > natural that the eprobe is only for the static events, and I also think > dereferencing a pointer-type field in raw-event is more reliable than > dereferencing a digit value passed to the synthetic event. Although I believe we need to attach eprobes to synthetic events, for the reasons I stated in my previous email. I'm perfectly happy to forbid them from attaching to kprobe or uprobe events. Because, honestly, eprobes do not give you anything that a kprobe nor uprobe can give you. -- Steve
|