Re: [PATCH] fs: Optimized fget to improve performance

[Matthew Wilcox <willy@infradead.org>]

On Thu, Aug 27, 2020 at 06:19:44PM +0800, Shaokun Zhang wrote:
> From: Yuqi Jin <jinyuqi@huawei.com>
> 
> It is well known that the performance of atomic_add is better than that of
> atomic_cmpxchg.

I don't think that's well-known at all.

> +static inline bool get_file_unless_negative(atomic_long_t *v, long a)
> +{
> +	long c = atomic_long_read(v);
> +
> +	if (c <= 0)
> +		return 0;
> +
> +	return atomic_long_add_return(a, v) - 1;
> +}
> +
>  #define get_file_rcu_many(x, cnt)	\
> -	atomic_long_add_unless(&(x)->f_count, (cnt), 0)
> +	get_file_unless_negative(&(x)->f_count, (cnt))
>  #define get_file_rcu(x) get_file_rcu_many((x), 1)
>  #define file_count(x)	atomic_long_read(&(x)->f_count)

I think you should be proposing a patch to fix atomic_long_add_unless()
on arm64 instead.