Re: [PATCH] avr32: use set_current_blocked() in handle_signal/sys_rt_sigreturn

[Matt Fleming <matt@console-pimps.org>]

On Wed, 2011-08-03 at 15:39 +0200, Oleg Nesterov wrote:
> On 08/03, Oleg Nesterov wrote:
> >
> > On 08/03, Matt Fleming wrote:
> > >
> > > 	 * Block the signal if we were unsuccessful.
> > > 	 */
> > > 	if (ret != 0 || !(ka->sa.sa_flags & SA_NODEFER)) {
> > > 		spin_lock_irq(&current->sighand->siglock);
> > > 		sigorsets(&current->blocked, &current->blocked,
> > > 			  &ka->sa.sa_mask);
> > > 		sigaddset(&current->blocked, sig);
> > > 		recalc_sigpending();
> > > 		spin_unlock_irq(&current->sighand->siglock);
> > > 	}
> >
> > Agreed, this looks "obviously wrong". We should block the !SA_NODEFER
> > signal it was delivered.
> 
> (I meant, it is was successfully delivered).
> 
> While at it, I'd also suggest another patch on top of Matt's.
> Uncompiled/untested.
> 
> ------------------------------------------------------------------------------
> [PATCH] avr32: use set_current_blocked() in handle_signal/sys_rt_sigreturn

Actually this was the reason that I spotted this bug. I'm going through
all the code that sets current->blocked manually and making it use
set_current_blocked() ;-)

Moving most of the recalc_pending() callers into one place
(set_current_blocked()) makes my signal scalability series a lot
smaller.