On 09/12, Oleg Nesterov wrote: > > 1. I assume that, on every arch, we alway check TIF_NOTIFY_RESUME > and TIF_SIGPENDING before TIF_SYSCALL_TRACE, right? > > IOW, can I assume that syscall_trace_leave() should not be called > before utrace_resume() or utrace_get_signal() ? > > This is true (afacs) for x86, but I am not sure about other > arches.
Damn! I am stupid. No, this is NOT true even on x86. I greatly misread entry_64.S:int_very_careful, _TIF_WORK_SYSCALL_EXIT does NOT include _TIF_DO_NOTIFY_MASK. OK, makes sense, we are not going to return to the user-mode yet... Oh. So we really need the stack of stop events, and I have to change my plan... Stupid me. Oleg.
