Re: [PATCH] signals: Generate warning when flush_signals() is called from non-kthread context

[Ingo Molnar <mingo2.kernel.org@gmail.com>]

* Oleg Nesterov <oleg@redhat.com> wrote:

> > --- a/kernel/signal.c
> > +++ b/kernel/signal.c
> > @@ -427,6 +427,10 @@ void flush_signals(struct task_struct *t)
> >  {
> >  	unsigned long flags;
> >
> > +	/* Only kthreads are allowed to destroy signals: */
> > +	if (WARN_ON_ONCE(!(current->flags & PF_KTHREAD)))
> > +		return;
> > +
> 
> But I am not sure this can't make some buggy driver even more buggy.
> Just suppose it does something
> 
> 	do {
> 		if (signal_pending())
> 			flush_signals();
> 	} while (wait_event_interruptible(...));
> 
> and this change will turn this into busy-wait loop.
> 
> So perhaps another change which just adds WARN_ON_RATELIMIT() 
> without "return" will be safer?

Yeah, absolutely.

Thanks,

	Ingo