Re: [dpdk-dev] [PATCH v2 05/15] net/mlx5: support tunnel inner checksum offloads

["Nélio Laranjeiro" <nelio.laranjeiro@6wind.com>]

On Tue, Apr 10, 2018 at 09:34:05PM +0800, Xueming Li wrote:
> This patch support tunnel inner checksum offloads. By creating tunnel
> flow, once tunnel packet type(RTE_PTYPE_TUNNEL_xxx) identified,

Where is the code creating the tunnel flow?

> PKT_RX_IP_CKSUM_XXX and PKT_RX_L4_CKSUM_XXX represent checksum result of
> inner headers, outer L3 and L4 header checksum are always valid as soon
> as tunnel identified. If no tunnel identified, PKT_RX_IP_CKSUM_XXX and
> PKT_RX_L4_CKSUM_XXX represent checksum result of outer L3 and L4
> headers.
> 
> Signed-off-by: Xueming Li <xuemingl@mellanox.com>
> ---
>  drivers/net/mlx5/mlx5_flow.c |  7 +++++--
>  drivers/net/mlx5/mlx5_rxq.c  |  2 --
>  drivers/net/mlx5/mlx5_rxtx.c | 18 ++++--------------
>  drivers/net/mlx5/mlx5_rxtx.h |  1 -
>  4 files changed, 9 insertions(+), 19 deletions(-)
> 
> diff --git a/drivers/net/mlx5/mlx5_flow.c b/drivers/net/mlx5/mlx5_flow.c
> index 65d7a9b62..b3ad6dc85 100644
> --- a/drivers/net/mlx5/mlx5_flow.c
> +++ b/drivers/net/mlx5/mlx5_flow.c
> @@ -829,6 +829,8 @@ mlx5_flow_convert_actions(struct rte_eth_dev *dev,
>  /**
>   * Validate items.
>   *
> + * @param dev
> + *   Pointer to Ethernet device.
>   * @param[in] items
>   *   Pattern specification (list terminated by the END pattern item).
>   * @param[out] error
> @@ -840,7 +842,8 @@ mlx5_flow_convert_actions(struct rte_eth_dev *dev,
>   *   0 on success, a negative errno value otherwise and rte_errno is set.
>   */
>  static int
> -mlx5_flow_convert_items_validate(const struct rte_flow_item items[],
> +mlx5_flow_convert_items_validate(struct rte_eth_dev *dev __rte_unused,
> +				 const struct rte_flow_item items[],
>  				 struct rte_flow_error *error,
>  				 struct mlx5_flow_parse *parser)
>  {
> @@ -1146,7 +1149,7 @@ mlx5_flow_convert(struct rte_eth_dev *dev,
>  	ret = mlx5_flow_convert_actions(dev, actions, error, parser);
>  	if (ret)
>  		return ret;
> -	ret = mlx5_flow_convert_items_validate(items, error, parser);
> +	ret = mlx5_flow_convert_items_validate(dev, items, error, parser);
>  	if (ret)
>  		return ret;
>  	mlx5_flow_convert_finalise(parser);

I don't understand the necessity of the two hunks above.

> diff --git a/drivers/net/mlx5/mlx5_rxq.c b/drivers/net/mlx5/mlx5_rxq.c
> index 351acfc0f..073732e16 100644
> --- a/drivers/net/mlx5/mlx5_rxq.c
> +++ b/drivers/net/mlx5/mlx5_rxq.c
> @@ -1045,8 +1045,6 @@ mlx5_rxq_new(struct rte_eth_dev *dev, uint16_t idx, uint16_t desc,
>  	}
>  	/* Toggle RX checksum offload if hardware supports it. */
>  	tmpl->rxq.csum = !!(conf->offloads & DEV_RX_OFFLOAD_CHECKSUM);
> -	tmpl->rxq.csum_l2tun = (!!(conf->offloads & DEV_RX_OFFLOAD_CHECKSUM) &&
> -				priv->config.tunnel_en);
>  	tmpl->rxq.hw_timestamp = !!(conf->offloads & DEV_RX_OFFLOAD_TIMESTAMP);
>  	/* Configure VLAN stripping. */
>  	tmpl->rxq.vlan_strip = !!(conf->offloads & DEV_RX_OFFLOAD_VLAN_STRIP);
> diff --git a/drivers/net/mlx5/mlx5_rxtx.c b/drivers/net/mlx5/mlx5_rxtx.c
> index d061dfc8a..285b2dbf0 100644
> --- a/drivers/net/mlx5/mlx5_rxtx.c
> +++ b/drivers/net/mlx5/mlx5_rxtx.c
> @@ -41,7 +41,7 @@ mlx5_rx_poll_len(struct mlx5_rxq_data *rxq, volatile struct mlx5_cqe *cqe,
>  		 uint16_t cqe_cnt, uint32_t *rss_hash);
>  
>  static __rte_always_inline uint32_t
> -rxq_cq_to_ol_flags(struct mlx5_rxq_data *rxq, volatile struct mlx5_cqe *cqe);
> +rxq_cq_to_ol_flags(volatile struct mlx5_cqe *cqe);
>  
>  uint32_t mlx5_ptype_table[] __rte_cache_aligned = {
>  	[0xff] = RTE_PTYPE_ALL_MASK, /* Last entry for errored packet. */
> @@ -1728,8 +1728,6 @@ mlx5_rx_poll_len(struct mlx5_rxq_data *rxq, volatile struct mlx5_cqe *cqe,
>  /**
>   * Translate RX completion flags to offload flags.
>   *
> - * @param[in] rxq
> - *   Pointer to RX queue structure.
>   * @param[in] cqe
>   *   Pointer to CQE.
>   *
> @@ -1737,7 +1735,7 @@ mlx5_rx_poll_len(struct mlx5_rxq_data *rxq, volatile struct mlx5_cqe *cqe,
>   *   Offload flags (ol_flags) for struct rte_mbuf.
>   */
>  static inline uint32_t
> -rxq_cq_to_ol_flags(struct mlx5_rxq_data *rxq, volatile struct mlx5_cqe *cqe)
> +rxq_cq_to_ol_flags(volatile struct mlx5_cqe *cqe)
>  {
>  	uint32_t ol_flags = 0;
>  	uint16_t flags = rte_be_to_cpu_16(cqe->hdr_type_etc);
> @@ -1749,14 +1747,6 @@ rxq_cq_to_ol_flags(struct mlx5_rxq_data *rxq, volatile struct mlx5_cqe *cqe)
>  		TRANSPOSE(flags,
>  			  MLX5_CQE_RX_L4_HDR_VALID,
>  			  PKT_RX_L4_CKSUM_GOOD);
> -	if ((cqe->pkt_info & MLX5_CQE_RX_TUNNEL_PACKET) && (rxq->csum_l2tun))
> -		ol_flags |=
> -			TRANSPOSE(flags,
> -				  MLX5_CQE_RX_L3_HDR_VALID,
> -				  PKT_RX_IP_CKSUM_GOOD) |
> -			TRANSPOSE(flags,
> -				  MLX5_CQE_RX_L4_HDR_VALID,
> -				  PKT_RX_L4_CKSUM_GOOD);
>  	return ol_flags;
>  }
>  
> @@ -1855,8 +1845,8 @@ mlx5_rx_burst(void *dpdk_rxq, struct rte_mbuf **pkts, uint16_t pkts_n)
>  						mlx5_flow_mark_get(mark);
>  				}
>  			}
> -			if (rxq->csum | rxq->csum_l2tun)
> -				pkt->ol_flags |= rxq_cq_to_ol_flags(rxq, cqe);
> +			if (rxq->csum)
> +				pkt->ol_flags |= rxq_cq_to_ol_flags(cqe);
>  			if (rxq->vlan_strip &&
>  			    (cqe->hdr_type_etc &
>  			     rte_cpu_to_be_16(MLX5_CQE_VLAN_STRIPPED))) {
> diff --git a/drivers/net/mlx5/mlx5_rxtx.h b/drivers/net/mlx5/mlx5_rxtx.h
> index 6866f6818..d35605b55 100644
> --- a/drivers/net/mlx5/mlx5_rxtx.h
> +++ b/drivers/net/mlx5/mlx5_rxtx.h
> @@ -77,7 +77,6 @@ struct rxq_zip {
>  /* RX queue descriptor. */
>  struct mlx5_rxq_data {
>  	unsigned int csum:1; /* Enable checksum offloading. */
> -	unsigned int csum_l2tun:1; /* Same for L2 tunnels. */
>  	unsigned int hw_timestamp:1; /* Enable HW timestamp. */
>  	unsigned int vlan_strip:1; /* Enable VLAN stripping. */
>  	unsigned int crc_present:1; /* CRC must be subtracted. */
> -- 
> 2.13.3

This last part seems to introduce a regression by removing the support
for the tunnel checksum offload.

Seems this patch is incomplete or wrongly explained.

-- 
Nélio Laranjeiro
6WIND