Re: [PATCH 1/2] protect /sbin/init from unwanted signals more

[Oleg Nesterov <oleg@redhat.com>]

On 11/19, Roland McGrath wrote:
>
> The effect is fine, but that seems like a kludgey way to do it.

Agreed, that is why I did the next patch to kill the ugliness.

> I really don't think the sigaction case matters--certainly it will never
> come up with SIGKILL.

Yes. This patch doesn't affect sigaction, the next one adds a very
minor side effect: init drops pending !sig_kernel_ignore() signals
if it does sigaction(SIG_IGN). But this has nothing to do with SIGKILL
of course.

> What about just this instead?
>
> +	if (unlikely(signal->flags & SIGNAL_UNKILLABLE) &&
> +	    (handler == SIG_IGN || handler == SIG_DFL))
> +		return 1;
> +
>  	if (!sig_handler_ignored(handler, sig))
>  		return 0;

Yes, this is the same, but anyway this is killed by the next patch.



> For consistency, change tracehook_consider_fatal_signal to match.

Yes, will do.

Oleg.