Re: [PATCH v8 6/6] block: loop: support DIO & AIO

[Christoph Hellwig <hch@infradead.org>]

On Thu, Jul 30, 2015 at 07:36:24AM -0400, Ming Lei wrote:
> +	/*
> +	 * When working at direct I/O, under very unusual cases,
> +	 * such as unaligned direct I/O from application and
> +	 * access to loop block device with 'unaligned' offset & size,
> +	 * we have to fallback to non-dio mode.
> +	 *
> +	 * During the switch between dio and non-dio, page cache
> +	 * has to be flushed to the backing file.
> +	 */
> +	if (unlikely(lo->use_dio && lo->last_use_dio != cmd->use_aio))
> +		vfs_fsync(lo->lo_backing_file, 0);

Filesystems do the cache flushing for you.

> +static inline bool req_dio_aligned(struct loop_device *lo,
> +		const struct request *rq)
> +{
> +	return !((blk_rq_pos(rq) << 9) & lo->dio_align) &&
> +		!(blk_rq_bytes(rq) & lo->dio_align);
> +}
> +
>  static int loop_queue_rq(struct blk_mq_hw_ctx *hctx,
>  		const struct blk_mq_queue_data *bd)
>  {
> @@ -1554,6 +1658,13 @@ static int loop_queue_rq(struct blk_mq_hw_ctx *hctx,
>  	if (lo->lo_state != Lo_bound)
>  		return -EIO;
>  
> +	if (lo->use_dio && !lo->transfer &&
> +			req_dio_aligned(lo, bd->rq) &&
> +			!(cmd->rq->cmd_flags & (REQ_FLUSH | REQ_DISCARD)))
> +		cmd->use_aio = true;
> +	else
> +		cmd->use_aio = false;

But honestly run time switching between buffered I/O and direct I/O from
the same I/O stream is almost asking for triggering every possible
race in the dio vs buffered I/O synchronization.  And there have been
a lot of those..

I'd feel much more comfortable with a setup time check.