Re: [RFC][PATCH] signal: Store pending signal exit in tsk.jobctl not in tsk.pending

[Oleg Nesterov]

Eric, at al,

Sorry, I am on on vacation, can't even read this thread right now,
so I am not sure I understand the problem correctly...

On 02/05, Eric W. Biederman wrote:
>
> @@ -2393,6 +2393,11 @@ bool get_signal(struct ksignal *ksig)
>  		goto relock;
>  	}
>
> +	/* Has this task already been flagged for death? */
> +	ksig->info.si_signo = signr = SIGKILL;
> +	if (current->jobctl & JOBCTL_TASK_EXIT)
> +		goto fatal;
> +

Can't we simply change, say, next_signal() to return SIGKILL if it is
pending?

In any case, I am not sure we need JOBCTL_TASK_EXIT. Can't we rely on
signal_group_exit() ?

Oleg.