Re: [patch V3 50/64] locking/rtmutex: Extend the rtmutex core to support ww_mutex

[Peter Zijlstra]

On Thu, Aug 05, 2021 at 05:13:50PM +0200, Thomas Gleixner wrote:
>  static __always_inline bool __waiter_less(struct rb_node *a, const struct rb_node *b)
>  {
> -	return rt_mutex_waiter_less(__node_2_waiter(a), __node_2_waiter(b));

Given ^

> +	struct rt_mutex_waiter *aw = __node_2_waiter(a);
> +	struct rt_mutex_waiter *bw = __node_2_waiter(b);
> +
> +	if (rt_mutex_waiter_less(aw, bw))
> +		return 1;

We can, with this new build_ww_mutex(), do:

	if (!build_ww_mutex())
		return 0;

here, to preserve the old behaviour.

> +	if (rt_mutex_waiter_less(bw, aw))
> +		return 0;
> +
> +	/* NOTE: relies on waiter->ww_ctx being set before insertion */
> +	if (build_ww_mutex() && aw->ww_ctx) {

Then it can go away here.

> +		if (!bw->ww_ctx)
> +			return 1;
> +
> +		return (signed long)(aw->ww_ctx->stamp -
> +				     bw->ww_ctx->stamp) < 0;
> +	}
> +
> +	return 0;
>  }

Also, I found a note that said I had to check what this code does for
!RT tasks, lemme go do that now.