Re: [PATCH 2/5] sched,ptrace: Fix ptrace_check_attach() vs PREEMPT_RT

[Peter Zijlstra]

On Mon, Apr 18, 2022 at 07:01:05PM +0200, Oleg Nesterov wrote:

> diff --git a/include/linux/sched/jobctl.h b/include/linux/sched/jobctl.h
> index ec8b312f7506..1b5a57048e13 100644
> --- a/include/linux/sched/jobctl.h
> +++ b/include/linux/sched/jobctl.h
> @@ -22,7 +22,8 @@ struct task_struct;
>  
>  #define JOBCTL_STOPPED_BIT	24
>  #define JOBCTL_TRACED_BIT	25
> +#define JOBCTL_TRACED_XXX_BIT	25

26, also we must come up with a better name than tripple-x. In my head
it's started to be called TRACED_OLEG, but that can't be right either
;-)

Does something like:

#define JOBCTL_TRACED_BIT		25
#define JOBCTL_TRACED_QUIESCE_BIT	26

work?

> diff --git a/kernel/signal.c b/kernel/signal.c
> index 0aea3f0a8002..c7a89904cc4a 100644
> --- a/kernel/signal.c
> +++ b/kernel/signal.c
> @@ -2182,6 +2182,13 @@ static void do_notify_parent_cldstop(struct task_struct *tsk,
>  	spin_unlock_irqrestore(&sighand->siglock, flags);
>  }
>  
> +static void clear_traced_xxx(void)
> +{
> +	spin_lock_irq(&current->sighand->siglock);
> +	current->jobctl &= ~JOBCTL_TRACED_XXX;
> +	spin_unlock_irq(&current->sighand->siglock);
> +}
> +
>  /*
>   * This must be called with current->sighand->siglock held.
>   *
> @@ -2220,7 +2227,7 @@ static int ptrace_stop(int exit_code, int why, int clear_code,
>  	 * schedule() will not sleep if there is a pending signal that
>  	 * can awaken the task.
>  	 */
> -	current->jobctl |= JOBCTL_TRACED;
> +	current->jobctl |= JOBCTL_TRACED | JOBCTL_TRACED_XXX;
>  	set_special_state(TASK_TRACED);
>  
>  	/*
> @@ -2282,6 +2289,8 @@ static int ptrace_stop(int exit_code, int why, int clear_code,
>  		if (gstop_done && ptrace_reparented(current))
>  			do_notify_parent_cldstop(current, false, why);
>  
> +		clear_traced_xxx();
> +		wake_up_state(current->parent, TASK_KILLABLE);
>  		/*
>  		 * Don't want to allow preemption here, because
>  		 * sys_ptrace() needs this task to be inactive.
> @@ -2297,8 +2306,12 @@ static int ptrace_stop(int exit_code, int why, int clear_code,
>  	} else {
>  		/*
>  		 * By the time we got the lock, our tracer went away.
> -		 * Don't drop the lock yet, another tracer may come.
> -		 *
> +		 * Don't drop the lock yet, another tracer may come,
> +		 * tasklist protects us from ptrace_freeze_traced().
> +		 */
> +		__set_current_state(TASK_RUNNING);
> +		clear_traced_xxx();
> +		/*
>  		 * If @gstop_done, the ptracer went away between group stop
>  		 * completion and here.  During detach, it would have set
>  		 * JOBCTL_STOP_PENDING on us and we'll re-enter

This is that same else clause again; perhaps make signal_wake_up_state()
also clear TRACED_XXX instead?