Re: [PATCH v2 4/6] debugobjects: Don't start fill if there are remaining nodes locally

From: Leizhen (ThunderTown)
Date: Tue Sep 10 2024 - 11:29:58 EST




On 2024/9/9 21:51, Leizhen (ThunderTown) wrote:
> +static void fill_pool(void)
> +{
> + gfp_t gfp = __GFP_HIGH | __GFP_NOWARN;
> + static atomic_t cpus_allocating;
>
> if (unlikely(!obj_cache))
> return;
>
> + /*
> + * Avoid allocation and lock contention when
> + *
> + * - the CPU local pool has at least 2 objects left
> + * - another CPU is already in the allocation path
> + * - the global pool has not reached the critical level yet
> + */
> + if (this_cpu_read(percpu_obj_pool.obj_free) > 1 && atomic_read(&cpus_allocating) &&

I rethink that 'cpus_allocating' and 'percpu_obj_pool.obj_free > 1' should be
contradictory. We can only choose one of them. If 'cpus_allocating' can work,
there's no need for another. Since I had no way to prove that 'cpus_allocating'
would work, I removed it in V3.

> + READ_ONCE(obj_pool_free) > (debug_objects_pool_min_level / 2))
> + return;
> +
> + atomic_inc(&cpus_allocating);
> while (READ_ONCE(obj_pool_free) < debug_objects_pool_min_level) {
> struct debug_obj *new[ODEBUG_BATCH_SIZE];
> int cnt;

--
Regards,
Zhen Lei