Re: [PATCH] x86: do_signal: simplify the TS_RESTORE_SIGMASK logic

[Oleg Nesterov <oleg@redhat.com>]

On 07/10, Al Viro wrote:
>
> On Sun, Jul 10, 2011 at 08:22:03PM +0200, Oleg Nesterov wrote:
>
> > 2. do_signal() also clears TS_RESTORE_SIGMASK if handle_signal()
> >    succeeds.
> >
> >    We can move this to setup_rt_frame() as well, this avoids the
> >    unnecessary checks and makes the logic more clear.
>
>
> > +	current_thread_info()->status &= ~TS_RESTORE_SIGMASK;
> >  	return ret;
>
> This is broken.

You know, I was also going to change this "return ret", "return 0"
looks more clear to me. But then I should have renamed "ret", and
I decided to leave  it alone.

> If setup_rt_frame() fails, you don't want to do that.

Sure. Please look at the code, it does

	if (ret) {
		force_sigsegv(sig, current);
		return -EFAULT;
	}

	current_thread_info()->status &= ~TS_RESTORE_SIGMASK;
	return ret;

We clear TS_RESTORE_SIGMASK only if we return 0.

Oleg.