Re: [PATCH 2/2] fs: Remove kiocb->ki_complete

[Jens Axboe]

On 7/8/20 4:26 PM, Matthew Wilcox (Oracle) wrote:
> @@ -341,12 +343,25 @@ struct kiocb {
>  	randomized_struct_fields_end
>  };
>  
> +static inline int kiocb_completion_id(struct kiocb *kiocb)
> +{
> +	return kiocb->ki_flags >> _IOCB_COMPLETION_SHIFT;
> +}
> +
> +static inline void kiocb_set_completion(struct kiocb *kiocb, int id)
> +{
> +	kiocb->ki_flags = (kiocb->ki_flags & (~IOCB_COMPLETION_FNS)) |
> +				(id << _IOCB_COMPLETION_SHIFT);
> +}
> +
>  static inline bool is_sync_kiocb(struct kiocb *kiocb)
>  {
> -	return kiocb->ki_complete == NULL;
> +	return kiocb_completion_id(kiocb) == 0;
>  }
>  
>  void complete_kiocb(struct kiocb *kiocb, long ret, long ret2);
> +int register_kiocb_completion(void (*)(struct kiocb *, long, long));
> +void unregister_kiocb_completion(int id);

Same here, you seem to mix and match whether you prefix with kiocb or
not. Why not make them all kiocb_*?

kiocb_register_completion() and so forth.

-- 
Jens Axboe