Re: [RFC patch 4/7] futex: Add support for attached futexes

[Peter Zijlstra]

On Sat, Apr 02, 2016 at 11:09:18AM -0000, Thomas Gleixner wrote:
> +/**
> + * futex_detach_task - Detach task from global state
> + * @slot:	Slot number in the task local cache
> + *
> + * If the global state refcount drops to zero, the global state is destroyed.
> + */
> +static void futex_detach_task(int slot)
> +{
> +	struct futex_cache *tc = current->futex_cache;
> +	struct futex_state *fs = tc->slots[slot].fs;
> +	struct futex_hash_bucket *hb = fs->global_hb;
> +	struct futex_q *q = &fs->q;
> +
> +	/* Remove it from the task local cache */
> +	__clear_bit(slot, tc->cache_map);
> +	tc->slots[slot].uaddr = NULL;
> +	tc->slots[slot].fs = NULL;
> +
> +	/*
> +	 * Lock the global hash bucket. Decrement global state refcount. If 0
> +	 * remove it from the global hash and free it.
> +	 */
> +	spin_lock(&hb->lock);
> +	if (--fs->refcount == 0)
> +		hb_remove_q(q, hb);
> +	else
> +		fs = NULL;
> +	spin_unlock(&hb->lock);

So you could play funny games like:

	if (atomic_add_unless(&fs->recount, -1, 1))
		return;

	spin_lock(&hb->lock);
	if (atomic_dec_return(&fs->refcount) == 0)
		hb_remove_q(q, hb);
	else
		fs = NULL;
	spin_unlock(&hb->lock);

To avoid taking that lock entirely in the 'fast' path, but I'm not sure
how performance critical this path is.

> +	kfree(fs);
> +}