[PATCH v2 2/2] crypto: caam: caamrng: dma map descriptors

[Lucas Stach]

Am Freitag, dem 08.03.2024 um 08:35 +0100 schrieb Rouven Czerwinski:
> With DMA API debugging Barebox complains that the descriptors are never
> mapped before a sync. Add the map and unmap function calls.
> 
> Signed-off-by: Rouven Czerwinski <r.czerwinski at pengutronix.de>
> ---
> v2:
> - remove dma_map_sync_single_for_device() calls
> - remove dma_map_single for dma coherent memory
> 
>  drivers/crypto/caam/caamrng.c | 7 +++----
>  1 file changed, 3 insertions(+), 4 deletions(-)
> 
> diff --git a/drivers/crypto/caam/caamrng.c b/drivers/crypto/caam/caamrng.c
> index ea154913ca..03e75d0b7b 100644
> --- a/drivers/crypto/caam/caamrng.c
> +++ b/drivers/crypto/caam/caamrng.c
> @@ -91,6 +91,7 @@ static void rng_done(struct device *jrdev, u32 *desc, u32 err, void *context)
>  
>  	/* Buffer refilled, invalidate cache */
>  	dma_sync_single_for_cpu(jrdev, bd->addr, RN_BUF_SIZE, DMA_FROM_DEVICE);
> +	dma_unmap_single(jrdev, (unsigned long)desc, desc_bytes(desc), DMA_TO_DEVICE);
>  }
>  
>  static inline int submit_job(struct caam_rng_ctx *ctx, int to_current)
> @@ -102,8 +103,7 @@ static inline int submit_job(struct caam_rng_ctx *ctx, int to_current)
>  
>  	dev_dbg(jrdev, "submitting job %d\n", !(to_current ^ ctx->current_buf));
>  
> -	dma_sync_single_for_device(jrdev, (unsigned long)desc, desc_bytes(desc),
> -				   DMA_TO_DEVICE);
> +	dma_map_single(jrdev, (void *)desc, desc_bytes(desc), DMA_TO_DEVICE);
>  
>  	err = caam_jr_enqueue(jrdev, desc, rng_done, ctx);
>  	if (!err)
> @@ -180,8 +180,7 @@ static inline int rng_create_sh_desc(struct caam_rng_ctx *ctx)
>  
>  	ctx->sh_desc_dma = (dma_addr_t)desc;
>  
> -	dma_sync_single_for_device(ctx->jrdev, (unsigned long)desc, desc_bytes(desc),
> -				   DMA_TO_DEVICE);
> +	dma_map_single(ctx->jrdev, desc, desc_bytes(desc), DMA_TO_DEVICE);

dma_map_single returns the DMA address that should be stored in
ctx->sh_desc_dma instead of the simple cast from virtual to DMA address
in the line above.

The other two dma_map/unmap calls in this patch look like a layering
violation. I would argue that those should move down into
caam_jr_enqueue and caam_jr_dequeue, to avoid the equally silly cast
from virt to DMA in those functions.

Regards,
Lucas

>  
>  	print_hex_dump_debug("rng shdesc@: ", DUMP_PREFIX_OFFSET, 16, 4,
>  		       desc, desc_bytes(desc), 1);