+ Christoph, Arnd
On Wed, 19 Feb 2020 at 09:31, <haibo.chen@xxxxxxx> wrote:
>
> From: Haibo Chen <haibo.chen@xxxxxxx>
>
> To make dma_set_max_seg_size() work, need to create dev->dma_parms.
>
> Find this issue on i.MX8QM mek board, this platform config the
> max_segment_size to 65535, but this dma_set_max_seg_size do not
> actuall work, find sometimes the segment size is 65536, exceed
> the hardware max segment limitation, trigger ADMA error.
Sounds like we want something along the lines of this to be tagged for
stable as well.
>
> Signed-off-by: Haibo Chen <haibo.chen@xxxxxxx>
> ---
> drivers/mmc/core/queue.c | 11 +++++++----
> 1 file changed, 7 insertions(+), 4 deletions(-)
>
> diff --git a/drivers/mmc/core/queue.c b/drivers/mmc/core/queue.c
> index 9edc08685e86..91094e56f7c9 100644
> --- a/drivers/mmc/core/queue.c
> +++ b/drivers/mmc/core/queue.c
> @@ -359,6 +359,7 @@ static const struct blk_mq_ops mmc_mq_ops = {
> static void mmc_setup_queue(struct mmc_queue *mq, struct mmc_card *card)
> {
> struct mmc_host *host = card->host;
> + struct device *dev = mmc_dev(host);
> unsigned block_size = 512;
>
> blk_queue_flag_set(QUEUE_FLAG_NONROT, mq->queue);
> @@ -366,13 +367,12 @@ static void mmc_setup_queue(struct mmc_queue *mq, struct mmc_card *card)
> if (mmc_can_erase(card))
> mmc_queue_setup_discard(mq->queue, card);
>
> - if (!mmc_dev(host)->dma_mask || !*mmc_dev(host)->dma_mask)
> + if (!dev->dma_mask || !*dev->dma_mask)
> blk_queue_bounce_limit(mq->queue, BLK_BOUNCE_HIGH);
> blk_queue_max_hw_sectors(mq->queue,
> min(host->max_blk_count, host->max_req_size / 512));
> if (host->can_dma_map_merge)
> - WARN(!blk_queue_can_use_dma_map_merging(mq->queue,
> - mmc_dev(host)),
> + WARN(!blk_queue_can_use_dma_map_merging(mq->queue, dev),
> "merging was advertised but not possible");
If we plan to send this to stable, I rather make all the changes above
as a separate clean-up patch.
> blk_queue_max_segments(mq->queue, mmc_get_max_segments(host));
>
> @@ -389,7 +389,10 @@ static void mmc_setup_queue(struct mmc_queue *mq, struct mmc_card *card)
> blk_queue_max_segment_size(mq->queue,
> round_down(host->max_seg_size, block_size));
>
> - dma_set_max_seg_size(mmc_dev(host), queue_max_segment_size(mq->queue));
> + if (!dev->dma_parms)
> + dev->dma_parms = devm_kzalloc(dev, sizeof(*dev->dma_parms),
> + GFP_KERNEL);
I think this normally should be managed by the host driver, typically
while it runs its ->probe() function.
On the other hand, it seems like not many host drivers does this,
which makes me wonder a bit of how this is intended to work. I guess
we can always do the allocation here if it's not been done before.
Also, I see the drivers/mmc/host/alcor.c calls dma_set_max_seg_size(),
so I am guessing that calls fails as well. Hmm.
> + dma_set_max_seg_size(dev, queue_max_segment_size(mq->queue));
>
> INIT_WORK(&mq->recovery_work, mmc_mq_recovery_handler);
> INIT_WORK(&mq->complete_work, mmc_blk_mq_complete_work);
> --
> 2.17.1
>
Kind regards
Uffe
