[PATCH v3 1/2] dmadev: offload to free source buffer

fengchengwen fengchengwen at huawei.com
Sat Oct 7 11:00:46 CEST 2023


Hi Amit,

On 2023/9/28 19:50, Amit Prakash Shukla wrote:
> This changeset adds support in DMA library to free source DMA buffer by
> hardware. On a supported hardware, application can pass on the mempool
> information as part of vchan config when the DMA transfer direction is
> configured as RTE_DMA_DIR_MEM_TO_DEV.
> 
> Signed-off-by: Amit Prakash Shukla <amitprakashs at marvell.com>
> Acked-by: Morten Brørup <mb at smartsharesystems.com>
> Acked-by: Anoob Joseph <anoobj at marvell.com>
> ---
>  lib/dmadev/rte_dmadev.h | 27 +++++++++++++++++++++++++++
>  1 file changed, 27 insertions(+)
> 
> diff --git a/lib/dmadev/rte_dmadev.h b/lib/dmadev/rte_dmadev.h
> index b157ab7600..f7a6af2528 100644
> --- a/lib/dmadev/rte_dmadev.h
> +++ b/lib/dmadev/rte_dmadev.h
> @@ -278,6 +278,13 @@ int16_t rte_dma_next_dev(int16_t start_dev_id);
>  #define RTE_DMA_CAPA_OPS_COPY_SG	RTE_BIT64(33)
>  /** Support fill operation. */
>  #define RTE_DMA_CAPA_OPS_FILL		RTE_BIT64(34)
> +/** Support for source buffer free for mem to dev transfer.

Support auto free source buffer once the M2D (memory-to-device) transfer completed.

> + *
> + * @note Even though the DMA driver has this capability, it may not support all
> + * mempool drivers. If the mempool is not supported by the DMA driver,
> + * rte_dma_vchan_setup() will fail.

In addition to hardware support, there are requirements for buffer attribute (e.g. only specific mempool support it).

> + **/
> +#define RTE_DMA_CAPA_MEM_TO_DEV_SOURCE_BUFFER_FREE	RTE_BIT64(35)

1) this should follow RTE_DMA_CAPA_HANDLES_ERRORS, because it not new OPS.
2) the name is too long. how abort RTE_DMA_CAPA_AUTO_FREE_M2D_SBUF? I am not sure, suggest get better name.

>  /**@}*/
>  
>  /**
> @@ -581,6 +588,19 @@ struct rte_dma_vchan_conf {
>  	 * @see struct rte_dma_port_param
>  	 */
>  	struct rte_dma_port_param dst_port;
> +	/** Mempool from which source buffer is allocated. Mempool info is used
> +	 * for freeing source buffer by hardware when configured direction is
> +	 * RTE_DMA_DIR_MEM_TO_DEV. To free the source buffer by hardware,
> +	 * RTE_DMA_OP_FLAG_FREE_SBUF must be set while calling rte_dma_copy and
> +	 * rte_dma_copy_sg().
> +	 *
> +	 * @note If the mempool is not supported by the DMA device,
> +	 * rte_dma_vchan_setup() will fail.
> +	 *
> +	 * @see RTE_DMA_OP_FLAG_FREE_SBUF
> +	 */
> +	struct rte_mempool *mem_to_dev_src_buf_pool;
> +
>  };

Suggest add one extra struct e.g.
struct rte_dma_auto_free_param {
    union {
        struct rte_mempool *pool;
    }
    uint64_t reserved[2]; /**< Reserved for future fields. */
};

In the above conf, we could add a new field: struct rte_dma_auto_free_param  m2d_buf

>  
>  /**
> @@ -818,6 +838,13 @@ struct rte_dma_sge {
>   * capability bit for this, driver should not return error if this flag was set.
>   */
>  #define RTE_DMA_OP_FLAG_LLC     RTE_BIT64(2)
> +/** Mem to dev source buffer free flag.
> + * Used for freeing source DMA buffer by hardware when the transfer direction is
> + * configured as RTE_DMA_DIR_MEM_TO_DEV.
> + *
> + * @see struct rte_dma_vchan_conf::mem_to_dev_src_buf_pool
> + */
> +#define RTE_DMA_OP_FLAG_FREE_SBUF	RTE_BIT64(3)

Suggest RTE_DMA_OP_FLAG_AUTO_FREE_SBUF

>  /**@}*/
>  
>  /**
> 

The S in SBUF seem useless, because it should not auto free dstbuf in logically.

Maybe we should direct use auto-free (just like silent)


More information about the dev mailing list