Re: [PATCH v5 2/2] slub: Introduce CONFIG_SLUB_RCU_DEBUG

[Marco Elver]

On Tue, Jul 30, 2024 at 01:06PM +0200, Jann Horn wrote:
[...]
> +#ifdef CONFIG_SLUB_RCU_DEBUG
> +	if ((s->flags & SLAB_TYPESAFE_BY_RCU) && !after_rcu_delay) {
> +		struct rcu_delayed_free *delayed_free;
> +
> +		delayed_free = kmalloc(sizeof(*delayed_free), GFP_NOWAIT);

This may well be allocated by KFENCE.

[...]
> +#ifdef CONFIG_SLUB_RCU_DEBUG
> +static void slab_free_after_rcu_debug(struct rcu_head *rcu_head)
> +{
> +	struct rcu_delayed_free *delayed_free =
> +			container_of(rcu_head, struct rcu_delayed_free, head);
> +	void *object = delayed_free->object;
> +	struct slab *slab = virt_to_slab(object);
> +	struct kmem_cache *s;
> +
> +	if (WARN_ON(is_kfence_address(rcu_head)))
> +		return;

syzbot found this warning to trigger (because see above comment):
https://lore.kernel.org/all/00000000000052aa15061eaeb1fd@xxxxxxxxxx/

Should this have been `is_kfence_address(object)`?