On 24/09/2024 10:01 pm, Guenter Roeck wrote:
Since upstream commit 334304ac2bac ("dma-mapping: don't return errors
from dma_set_max_seg_size") calling dma_set_max_seg_size() on a device
not supporting DMA results in a warning traceback. This is seen when
booting the sifive_u machine from SD. The underlying SPI controller
(sifive,spi0 compatible) explicitly sets dma_mask to NULL.
Hmm, that driver probably shouldn't be doing that, but either way it's
not actually relevant to this check - what's significant is that when
the MMC host device is an "mmc-spi-slot" instance, its mmc_dev() ends up
being the SPI *device* itself, not the grandparent SPI controller.
Thus the patch itself is appropriate on the basis that MMC hosts may
exist on non-DMA-capable buses, so the subsystem needs to consider that.
Reviewed-by: Robin Murphy <robin.murphy@xxxxxxx>
Avoid the backtrace by only calling dma_set_max_seg_size() if DMA is
supported.
Cc: Christoph Hellwig <hch@xxxxxx>
Cc: Robin Murphy <robin.murphy@xxxxxxx>
Cc: Ulf Hansson <ulf.hansson@xxxxxxxxxx>
Signed-off-by: Guenter Roeck <linux@xxxxxxxxxxxx>
---
drivers/mmc/core/queue.c | 3 ++-
1 file changed, 2 insertions(+), 1 deletion(-)
diff --git a/drivers/mmc/core/queue.c b/drivers/mmc/core/queue.c
index d0b3ca8a11f0..4d6844261912 100644
--- a/drivers/mmc/core/queue.c
+++ b/drivers/mmc/core/queue.c
@@ -388,7 +388,8 @@ static struct gendisk *mmc_alloc_disk(struct mmc_queue *mq,
blk_queue_rq_timeout(mq->queue, 60 * HZ);
- dma_set_max_seg_size(mmc_dev(host), queue_max_segment_size(mq->queue));
+ if (mmc_dev(host)->dma_parms)
+ dma_set_max_seg_size(mmc_dev(host), 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);
