[PATCH v4 3/3] net/sfc: support VLAN stripping offload

[Andrew Rybchenko]

On 6/1/23 18:30, Artemii Morozov wrote:
> Extract vlan tci provided by the HW in the prefix and put it to mbuf.

vlan -> VLAN, tci -> TCI

> VLAN stripping is supported for ef100 datapath only.

It should be highlighted that it is device level offload.

> 
> Signed-off-by: Artemii Morozov <artemii.morozov at arknetworks.am>
> Reviewed-by: Viacheslav Galaktionov <viacheslav.galaktionov at arknetworks.am>
> Reviewed-by: Ivan Malov <ivan.malov at arknetworks.am>
> Reviewed-by: Andy Moreton <amoreton at xilinx.com>
> ---
>   doc/guides/nics/sfc_efx.rst    |  4 ++--
>   drivers/net/sfc/sfc_ef100_rx.c | 21 ++++++++++++++++++++-
>   drivers/net/sfc/sfc_rx.c       | 19 +++++++++++++++++++

Release notes should be updated to advertise the feature.

>   3 files changed, 41 insertions(+), 3 deletions(-)
> 
> diff --git a/doc/guides/nics/sfc_efx.rst b/doc/guides/nics/sfc_efx.rst
> index de0656876b..44fa24e1ba 100644
> --- a/doc/guides/nics/sfc_efx.rst
> +++ b/doc/guides/nics/sfc_efx.rst
> @@ -118,6 +118,8 @@ SFC EFX PMD has support for:
>   
>   - Port representors (see :ref: switch_representation)
>   
> +- VLAN stripping (if running firmware variant supports it)
> +
>   
>   Non-supported Features
>   ----------------------
> @@ -132,8 +134,6 @@ The features not yet supported include:
>   
>   - VLAN filtering
>   
> -- VLAN stripping
> -
>   - LRO
>   
>   
> diff --git a/drivers/net/sfc/sfc_ef100_rx.c b/drivers/net/sfc/sfc_ef100_rx.c
> index 37b754fa33..e323156a26 100644
> --- a/drivers/net/sfc/sfc_ef100_rx.c
> +++ b/drivers/net/sfc/sfc_ef100_rx.c
> @@ -68,6 +68,7 @@ struct sfc_ef100_rxq {
>   #define SFC_EF100_RXQ_INGRESS_MPORT	0x80
>   #define SFC_EF100_RXQ_USER_FLAG		0x100
>   #define SFC_EF100_RXQ_NIC_DMA_MAP	0x200
> +#define SFC_EF100_RXQ_VLAN_STRIP	0x400
>   	unsigned int			ptr_mask;
>   	unsigned int			evq_phase_bit_shift;
>   	unsigned int			ready_pkts;
> @@ -392,6 +393,7 @@ static const efx_rx_prefix_layout_t sfc_ef100_rx_prefix_layout = {
>   		SFC_EF100_RX_PREFIX_FIELD(RSS_HASH, B_FALSE),
>   		SFC_EF100_RX_PREFIX_FIELD(USER_FLAG, B_FALSE),
>   		SFC_EF100_RX_PREFIX_FIELD(USER_MARK, B_FALSE),
> +		SFC_EF100_RX_PREFIX_FIELD(VLAN_STRIP_TCI, B_FALSE),
>   
>   #undef	SFC_EF100_RX_PREFIX_FIELD
>   	}
> @@ -472,6 +474,17 @@ sfc_ef100_rx_prefix_to_offloads(const struct sfc_ef100_rxq *rxq,
>   						ESF_GZ_RX_PREFIX_INGRESS_MPORT);
>   	}
>   
> +	if (rxq->flags & SFC_EF100_RXQ_VLAN_STRIP) {
> +		uint32_t vlan_stripped;

Please, add empty line after variable declaration.
IMHO, bool type should be used here.

> +		vlan_stripped = EFX_XWORD_FIELD(rx_prefix[0], ESF_GZ_RX_PREFIX_VLAN_STRIPPED);
> +
> +		if (vlan_stripped != 0) {

No comparison if bool is used.

> +			ol_flags |= RTE_MBUF_F_RX_VLAN | RTE_MBUF_F_RX_VLAN_STRIPPED;
> +			m->vlan_tci = EFX_XWORD_FIELD(rx_prefix[0],
> +							ESF_GZ_RX_PREFIX_VLAN_STRIP_TCI);
> +		}
> +	}
> +
>   	m->ol_flags = ol_flags;
>   	return true;
>   }
> @@ -892,6 +905,12 @@ sfc_ef100_rx_qstart(struct sfc_dp_rxq *dp_rxq, unsigned int evq_read_ptr,
>   	    (rxq->flags & SFC_EF100_RXQ_INGRESS_MPORT))
>   		return ENOTSUP;
>   
> +	if ((unsup_rx_prefix_fields &
> +	     (1U << EFX_RX_PREFIX_FIELD_VLAN_STRIP_TCI)) == 0)

Shouldn't offload enable/disable be taken into account here?
If offload is not enabled, it is better to skip extra read
from Rx prefix and branching on fast path.

> +		rxq->flags |= SFC_EF100_RXQ_VLAN_STRIP;
> +	else
> +		rxq->flags &= ~SFC_EF100_RXQ_VLAN_STRIP;
> +
>   	rxq->prefix_size = pinfo->erpl_length;
>   	rxq->rearm_data = sfc_ef100_mk_mbuf_rearm_data(rxq->dp.dpq.port_id,
>   						       rxq->prefix_size);
> @@ -1004,7 +1023,7 @@ struct sfc_dp_rx sfc_ef100_rx = {
>   				  SFC_DP_RX_FEAT_FLOW_MARK |
>   				  SFC_DP_RX_FEAT_INTR |
>   				  SFC_DP_RX_FEAT_STATS,
> -	.dev_offload_capa	= 0,
> +	.dev_offload_capa	= RTE_ETH_RX_OFFLOAD_VLAN_STRIP,
>   	.queue_offload_capa	= RTE_ETH_RX_OFFLOAD_CHECKSUM |
>   				  RTE_ETH_RX_OFFLOAD_OUTER_IPV4_CKSUM |
>   				  RTE_ETH_RX_OFFLOAD_OUTER_UDP_CKSUM |
> diff --git a/drivers/net/sfc/sfc_rx.c b/drivers/net/sfc/sfc_rx.c
> index edd0f0c038..e9ef1d92ed 100644
> --- a/drivers/net/sfc/sfc_rx.c
> +++ b/drivers/net/sfc/sfc_rx.c
> @@ -938,6 +938,9 @@ sfc_rx_get_offload_mask(struct sfc_adapter *sa)
>   	if (encp->enc_tunnel_encapsulations_supported == 0)
>   		no_caps |= RTE_ETH_RX_OFFLOAD_OUTER_IPV4_CKSUM;
>   
> +	if (encp->enc_rx_vlan_stripping == 0)
> +		no_caps |= RTE_ETH_RX_OFFLOAD_VLAN_STRIP;
> +
>   	return ~no_caps;
>   }
>   
> @@ -1186,6 +1189,16 @@ sfc_rx_qinit(struct sfc_adapter *sa, sfc_sw_index_t sw_index,
>   	if (offloads & RTE_ETH_RX_OFFLOAD_RSS_HASH)
>   		rxq_info->type_flags |= EFX_RXQ_FLAG_RSS_HASH;
>   
> +

Too many empty lines

> +	if (sa->eth_dev->data->dev_conf.rxmode.offloads &
> +	    RTE_ETH_RX_OFFLOAD_VLAN_STRIP) {
> +		rxq_info->type_flags |= EFX_RXQ_FLAG_VLAN_STRIP;
> +	} else if (rx_conf->offloads & RTE_ETH_RX_OFFLOAD_VLAN_STRIP) {
> +		sfc_err(sa, "VLAN stripping must be configured during device configure");
> +		rc = EINVAL;
> +		goto fail_bad_conf;

As far as I know generic ethdev code will reject the request
earlier. So, the code is unreachable and dead.

> +	}
> +
>   	if ((sa->negotiated_rx_metadata & RTE_ETH_RX_METADATA_USER_FLAG) != 0)
>   		rxq_info->type_flags |= EFX_RXQ_FLAG_USER_FLAG;
>   
> @@ -1691,6 +1704,12 @@ sfc_rx_check_mode(struct sfc_adapter *sa, struct rte_eth_rxmode *rxmode)
>   		rxmode->offloads |= RTE_ETH_RX_OFFLOAD_OUTER_IPV4_CKSUM;
>   	}
>   
> +	if ((rxmode->offloads & RTE_ETH_RX_OFFLOAD_VLAN_STRIP) &&
> +	    (~offloads_supported & RTE_ETH_RX_OFFLOAD_VLAN_STRIP)) {
> +		sfc_err(sa, "VLAN stripping offload is requested but not supported");
> +		rc = ENOTSUP;
> +	}
> +

If I'm not mistaken generic ethdev code will reject the request
earlier and will not allow to reach the code here.

>   	return rc;
>   }
>