Re: [RFC PATCH bpf-next 03/10] mm, memcg: Add new helper obj_cgroup_from_current()

[Roman Gushchin <roman.gushchin@linux.dev>]

On Sun, Jun 19, 2022 at 03:50:25PM +0000, Yafang Shao wrote:
> The difference between get_obj_cgroup_from_current() and obj_cgroup_from_current()
> is that the later one doesn't add objcg's refcnt.
> 
> Signed-off-by: Yafang Shao <laoar.shao@gmail.com>
> ---
>  include/linux/memcontrol.h |  1 +
>  mm/memcontrol.c            | 24 ++++++++++++++++++++++++
>  2 files changed, 25 insertions(+)
> 
> diff --git a/include/linux/memcontrol.h b/include/linux/memcontrol.h
> index cf074156c6ac..402b42670bcd 100644
> --- a/include/linux/memcontrol.h
> +++ b/include/linux/memcontrol.h
> @@ -1703,6 +1703,7 @@ bool mem_cgroup_kmem_disabled(void);
>  int __memcg_kmem_charge_page(struct page *page, gfp_t gfp, int order);
>  void __memcg_kmem_uncharge_page(struct page *page, int order);
>  
> +struct obj_cgroup *obj_cgroup_from_current(void);
>  struct obj_cgroup *get_obj_cgroup_from_current(void);
>  struct obj_cgroup *get_obj_cgroup_from_page(struct page *page);
>  
> diff --git a/mm/memcontrol.c b/mm/memcontrol.c
> index abec50f31fe6..350a7849dac3 100644
> --- a/mm/memcontrol.c
> +++ b/mm/memcontrol.c
> @@ -2950,6 +2950,30 @@ struct obj_cgroup *get_obj_cgroup_from_page(struct page *page)
>  	return objcg;
>  }
>  
> +__always_inline struct obj_cgroup *obj_cgroup_from_current(void)
> +{
> +	struct obj_cgroup *objcg = NULL;
> +	struct mem_cgroup *memcg;
> +
> +	if (memcg_kmem_bypass())
> +		return NULL;
> +
> +	rcu_read_lock();
> +	if (unlikely(active_memcg()))
> +		memcg = active_memcg();
> +	else
> +		memcg = mem_cgroup_from_task(current);
> +
> +	for (; memcg != root_mem_cgroup; memcg = parent_mem_cgroup(memcg)) {
> +		objcg = rcu_dereference(memcg->objcg);
> +		if (objcg)
> +			break;
> +	}
> +	rcu_read_unlock();

Hm, what prevents the objcg from being released here? Under which conditions
it's safe to call it?