> Perhaps we need more changes here, but I think this one is most important.
Agreed, we definitely do. For the generic case it needs to complete what do_signal_stop would do. That is, do_notify_parent_cldstop and maybe also: current->exit_code = sig->group_exit_code; Though it might be now that it obsolete wrt your newer exit.c changes, and either nothing uses that ->exit_code or only ptrace does. However, we want to suppress that do_notify_parent_cldstop for ptrace, where its ->report_jctl has already arranged to deal with that. So this would have whatever the same new "inhibit-wait" check is we use inside tracehook_notify_jctl for ptrace, but we haven't decied on that yet. Hmm. Perhaps it can just call tracehook_notify_jctl here. Thanks, Roland