Re: 5.8-rc*: kernel BUG at kernel/signal.c:1917

[Oleg Nesterov]

Hi Hillf,

On 07/19, Hillf Danton wrote:
>
> Dunno if the wheel prior to JOBCTL_TASK_WORK helps debug the warnings.
>
> --- a/kernel/signal.c
> +++ b/kernel/signal.c
> @@ -2541,7 +2541,7 @@ bool get_signal(struct ksignal *ksig)
>
>  relock:
>  	spin_lock_irq(&sighand->siglock);
> -	current->jobctl &= ~JOBCTL_TASK_WORK;
> +	task_clear_jobctl_pending(current, JOBCTL_TASK_WORK);
>  	if (unlikely(current->task_works)) {
>  		spin_unlock_irq(&sighand->siglock);
>  		task_work_run();
> --- a/kernel/task_work.c
> +++ b/kernel/task_work.c
> @@ -43,8 +43,8 @@ task_work_add(struct task_struct *task,
>  		break;
>  	case TWA_SIGNAL:
>  		if (lock_task_sighand(task, &flags)) {
> -			task->jobctl |= JOBCTL_TASK_WORK;
> -			signal_wake_up(task, 0);
> +			if (task_set_jobctl_pending(task, JOBCTL_TASK_WORK))
> +				recalc_sigpending_and_wake(task);
>  			unlock_task_sighand(task, &flags);

To be be honest I don't understand why do you think this makes any sense...

But this doesn't matter, please note that the Jiri tested the kernel with
e91b48162332480f5 (which added JOBCTL_TASK_WORK) reverted.

Oleg.