2022-10-26 07:56:32

by Christian Loehle

Subject: [PATCH 2/3] mmc: block: Limit queue size for consecutive requeues

In case mmcblk relies on requeueing to fulfill block layer requests due to
hardware or driver restricting the max_blk_count to 1 also limit the
hardware queue size to 1 to get consecutive accesses instead of a
round-robin on all queued reads.

See SDHCI_QUIRK_NO_MULTIBLOCK for such a restriction.

Signed-off-by: Christian Loehle <[email protected]>
---
drivers/mmc/core/queue.c | 2 ++
1 file changed, 2 insertions(+)

diff --git a/drivers/mmc/core/queue.c b/drivers/mmc/core/queue.c
index fefaa901b50f..3a20f3933502 100644
--- a/drivers/mmc/core/queue.c
+++ b/drivers/mmc/core/queue.c
@@ -425,6 +425,8 @@ struct gendisk *mmc_init_queue(struct mmc_queue *mq, struct mmc_card *card)
if (host->cqe_enabled && !host->hsq_enabled)
mq->tag_set.queue_depth =
min_t(int, card->ext_csd.cmdq_depth, host->cqe_qdepth);
+ else if (host->max_blk_count == 1)
+ mq->tag_set.queue_depth = 1;
else
mq->tag_set.queue_depth = MMC_QUEUE_DEPTH;
mq->tag_set.numa_node = NUMA_NO_NODE;
--
2.37.3

Hyperstone GmbH | Reichenaustr. 39a | 78467 Konstanz
Managing Director: Dr. Jan Peter Berns.
Commercial register of local courts: Freiburg HRB381782

2022-11-18 11:49:10

by Adrian Hunter

Subject: Re: [PATCH 2/3] mmc: block: Limit queue size for consecutive requeues

On 26/10/22 10:30, Christian Löhle wrote:
> In case mmcblk relies on requeueing to fulfill block layer requests due to
> hardware or driver restricting the max_blk_count to 1 also limit the
> hardware queue size to 1 to get consecutive accesses instead of a
> round-robin on all queued reads.
>
> See SDHCI_QUIRK_NO_MULTIBLOCK for such a restriction.
>
> Signed-off-by: Christian Loehle <[email protected]>
> ---
> drivers/mmc/core/queue.c | 2 ++
> 1 file changed, 2 insertions(+)
>
> diff --git a/drivers/mmc/core/queue.c b/drivers/mmc/core/queue.c
> index fefaa901b50f..3a20f3933502 100644
> --- a/drivers/mmc/core/queue.c
> +++ b/drivers/mmc/core/queue.c
> @@ -425,6 +425,8 @@ struct gendisk *mmc_init_queue(struct mmc_queue *mq, struct mmc_card *card)
> if (host->cqe_enabled && !host->hsq_enabled)
> mq->tag_set.queue_depth =
> min_t(int, card->ext_csd.cmdq_depth, host->cqe_qdepth);
> + else if (host->max_blk_count == 1)

I suggest adding a big comment here about why this is needed.
Perhaps like the information in your cover letter email.

> + mq->tag_set.queue_depth = 1;
> else
> mq->tag_set.queue_depth = MMC_QUEUE_DEPTH;
> mq->tag_set.numa_node = NUMA_NO_NODE;