Re: [PATCH v2] memcg: replace memcg ID idr with xarray

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

On Thu, Aug 15, 2024 at 08:54:02AM -0700, Shakeel Butt wrote:
> At the moment memcg IDs are managed through IDR which requires external
> synchronization mechanisms and makes the allocation code a bit awkward.
> Let's switch to xarray and make the code simpler.
> 
> Signed-off-by: Shakeel Butt <shakeel.butt@linux.dev>
> Suggested-by: Matthew Wilcox <willy@infradead.org>
> Reviewed-by: Roman Gushchin <roman.gushchin@linux.dev>
> Reviewed-by: Matthew Wilcox (Oracle) <willy@infradead.org>
> Acked-by: Johannes Weiner <hannes@cmpxchg.org>
> Reviewed-by: Muchun Song <muchun.song@linux.dev>
> Acked-by: Michal Hocko <mhocko@suse.com>
> Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
> ---
> 
> Changes since v1:
> - Fix error path in mem_cgroup_alloc (Dan Carpenter)
> 
>  mm/memcontrol.c | 39 ++++++++++-----------------------------
>  1 file changed, 10 insertions(+), 29 deletions(-)
> 
> diff --git a/mm/memcontrol.c b/mm/memcontrol.c
> index df84683a0e1c..e8e03a5e1e5e 100644
> --- a/mm/memcontrol.c
> +++ b/mm/memcontrol.c
> @@ -3408,29 +3408,12 @@ static void memcg_wb_domain_size_changed(struct mem_cgroup *memcg)
>   */
>  
>  #define MEM_CGROUP_ID_MAX	((1UL << MEM_CGROUP_ID_SHIFT) - 1)
> -static DEFINE_IDR(mem_cgroup_idr);
> -static DEFINE_SPINLOCK(memcg_idr_lock);
> -
> -static int mem_cgroup_alloc_id(void)
> -{
> -	int ret;
> -
> -	idr_preload(GFP_KERNEL);
> -	spin_lock(&memcg_idr_lock);
> -	ret = idr_alloc(&mem_cgroup_idr, NULL, 1, MEM_CGROUP_ID_MAX + 1,
> -			GFP_NOWAIT);
> -	spin_unlock(&memcg_idr_lock);
> -	idr_preload_end();
> -	return ret;
> -}
> +static DEFINE_XARRAY_ALLOC1(mem_cgroup_ids);
>  
>  static void mem_cgroup_id_remove(struct mem_cgroup *memcg)
>  {
>  	if (memcg->id.id > 0) {
> -		spin_lock(&memcg_idr_lock);
> -		idr_remove(&mem_cgroup_idr, memcg->id.id);
> -		spin_unlock(&memcg_idr_lock);
> -
> +		xa_erase(&mem_cgroup_ids, memcg->id.id);
>  		memcg->id.id = 0;
>  	}
>  }
> @@ -3465,7 +3448,7 @@ static inline void mem_cgroup_id_put(struct mem_cgroup *memcg)
>  struct mem_cgroup *mem_cgroup_from_id(unsigned short id)
>  {
>  	WARN_ON_ONCE(!rcu_read_lock_held());
> -	return idr_find(&mem_cgroup_idr, id);
> +	return xa_load(&mem_cgroup_ids, id);
>  }
>  
>  #ifdef CONFIG_SHRINKER_DEBUG
> @@ -3558,17 +3541,17 @@ static struct mem_cgroup *mem_cgroup_alloc(struct mem_cgroup *parent)
>  	struct mem_cgroup *memcg;
>  	int node, cpu;
>  	int __maybe_unused i;
> -	long error = -ENOMEM;
> +	long error;
>  
>  	memcg = kzalloc(struct_size(memcg, nodeinfo, nr_node_ids), GFP_KERNEL);
>  	if (!memcg)
> -		return ERR_PTR(error);
> +		return ERR_PTR(-ENOMEM);
>  
> -	memcg->id.id = mem_cgroup_alloc_id();
> -	if (memcg->id.id < 0) {
> -		error = memcg->id.id;
> +	error = xa_alloc(&mem_cgroup_ids, &memcg->id.id, NULL,
> +			 XA_LIMIT(1, MEM_CGROUP_ID_MAX), GFP_KERNEL);
> +	if (error)
>  		goto fail;
> -	}
> +	error = -ENOMEM;

There is another subtle change here: xa_alloc() returns -EBUSY in the case
of the address space exhaustion, while the old code returned -ENOSPC.
It's unlikely a big practical problem.