Re: Q: SEGSEGV && uc_mcontext->ip (Was: Signal delivery order)

From: Gábor Melis
Date: Tue Mar 17 2009 - 06:20:59 EST


On Martes 17 Marzo 2009, Oleg Nesterov wrote:
> On 03/17, Gábor Melis wrote:
> > On Martes 17 Marzo 2009, Oleg Nesterov wrote:
> > > But this doesn't look very nice. So, perhaps we can do another
> > > change?
> > >
> > > --- arch/x86/mm/fault.c
> > > +++ arch/x86/mm/fault.c
> > > @@ -177,6 +177,13 @@ static void force_sig_info_fault(int si_
> > > {
> > > siginfo_t info;
> > >
> > > + current->saved_sigmask = current->blocked;
> > > + spin_lock_irq(&current->sighand->siglock);
> > > + siginitsetinv(&current->blocked, sigmask(si_signo) |
> > > + sigmask(SIGKILL) | sigmask(SIGSTOP));
> > > + spin_unlock_irq(&current->sighand->siglock);
> > > + set_restore_sigmask();
> > > +
> > > info.si_signo = si_signo;
> > > info.si_errno = 0;
> > > info.si_code = si_code;
> > >
> > > But this is a user-visible change, all signals will be blocked
> > > until sigsegv_handler() returns. But with this change
> > > sigsegv_handler() always has the "correct" rt_sigframe.
> >
> > As an application developer what I'd like to have is this:
> > synchronously generated signals are delivered before asynchronously
> > generated ones. That is, if a number of signals are generated but
> > not yet delivered then the synchronously generated ones are
> > delivered first. I guess, in the kernel this would mean that the
> > private/non-private distinction is not enough.
>
> With the change like above, no other signal (except SIGKILL) can be
> delivered until the signal handler returns.

Surely, you don't mean the above literally: it would violate the
standard to prevent all other signals from being delivered until the
sigsegv handler returns.

> Probably it is better to just change force_sig_info(), in this case
> SIGFPE/etc will have the same behaviour.

Indeed, uniformity seems preferable to me.

While we are at it, an interesting case is when a synchronously
generated signal and an asynchronously generated signal - that is also
of the type that can be synchronously generated - are to be delivered.
Say we have a fault and a sigsegv generated but some misguided soul
pthread_kill()s with sigtrap. In this case the sigsegv shall be
delivered first, and the async sigtrap later.

> > The only thing that
> > worries me is this remark from Oleg
> > (http://marc.info/?l=linux-kernel&m=123711058421913&w=2):
> >
> > "But please note that it is still possible to hit
> > is_signal_blocked() even with test_with_kill(), but the probability
> > is very low."
>
> Sorry for confusion. Initially I misread test_with_kill() case, and
> then forgot to remove this part. I think this is not possible.

Thanks for the clarification.

> Oleg.
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/