On Sat, 5 Oct 2024 at 12:29, Oleg Nesterov <oleg@xxxxxxxxxx> wrote: > > On 10/04, Luca Boccassi wrote: > > > > On Fri, 4 Oct 2024 at 20:30, Oleg Nesterov <oleg@xxxxxxxxxx> wrote: > > > > > > I guess Christian meant you should simply use > > > > > > info.pid = task_pid_vnr(task); > > > > > > task_pid_vnr(task) returns the task's pid in the caller's namespace. > > > > Ah I see, I didn't realize there was a difference, sent v3 with the > > suggested change just now, thanks. > > I didn't get v3, I guess I wasn't cc'ed again. > > So, just in case, let me add that task_pid_vnr(task) can return 0 if > this task exits after get_pid_task(). > > Perhaps this is fine, I do not know. But perhaps you should actually > use pid_vnr(pid). > > Oleg. I have just sent v5 CC'ing you and adding a final check before the copy to userspace, that returns ESRCH if the task has exited. This should solve that issue, and also be future-proof against potential additions that might slow down processing due to gathering more data or so.