[PATCHv2 1/5] block: support polling through blk_execute_rq

Christoph Hellwig hch at lst.de
Mon Apr 26 15:35:43 BST 2021


On Fri, Apr 23, 2021 at 03:05:54PM -0700, Keith Busch wrote:
> Poll for completions if the request's hctx is a polling type.
> 
> Signed-off-by: Keith Busch <kbusch at kernel.org>
> ---
>  block/blk-exec.c | 12 +++++++++++-
>  1 file changed, 11 insertions(+), 1 deletion(-)
> 
> diff --git a/block/blk-exec.c b/block/blk-exec.c
> index beae70a0e5e5..b960ad187ba5 100644
> --- a/block/blk-exec.c
> +++ b/block/blk-exec.c
> @@ -63,6 +63,11 @@ void blk_execute_rq_nowait(struct gendisk *bd_disk, struct request *rq,
>  }
>  EXPORT_SYMBOL_GPL(blk_execute_rq_nowait);
>  
> +static bool blk_rq_is_poll(struct request *rq)
> +{
> +	return rq->mq_hctx && rq->mq_hctx->type == HCTX_TYPE_POLL;
> +}
> +
>  /**
>   * blk_execute_rq - insert a request into queue for execution
>   * @bd_disk:	matching gendisk
> @@ -83,7 +88,12 @@ void blk_execute_rq(struct gendisk *bd_disk, struct request *rq, int at_head)
>  
>  	/* Prevent hang_check timer from firing at us during very long I/O */
>  	hang_check = sysctl_hung_task_timeout_secs;
> -	if (hang_check)
> +	if (blk_rq_is_poll(rq)) {
> +		do {
> +			blk_poll(rq->q, request_to_qc_t(rq->mq_hctx, rq), true);
> +			cond_resched();
> +		} while (!completion_done(&wait));

I think it would be nice to split this into a little helper.

> +	} else if (hang_check)
>  		while (!wait_for_completion_io_timeout(&wait, hang_check * (HZ/2)));

And while we're at it, it would be nice to split this out as well and
document this mess..



More information about the Linux-nvme mailing list