DPDK patches and discussions
 help / color / mirror / Atom feed
From: Raslan Darawsheh <rasland@nvidia.com>
To: Rongwei Liu <rongweil@nvidia.com>, Matan Azrad <matan@nvidia.com>,
	Slava Ovsiienko <viacheslavo@nvidia.com>,
	Ori Kam <orika@nvidia.com>,
	NBU-Contact-Thomas Monjalon <thomas@monjalon.net>,
	Shahaf Shuler <shahafs@nvidia.com>
Cc: "dev@dpdk.org" <dev@dpdk.org>
Subject: Re: [dpdk-dev] [PATCH v5 1/2] net/mlx5: add VXLAN header the last 8-bits matching support
Date: Tue, 13 Jul 2021 10:27:46 +0000	[thread overview]
Message-ID: <DM4PR12MB5054F9281434978B648F6C70CF149@DM4PR12MB5054.namprd12.prod.outlook.com> (raw)
In-Reply-To: <20210713095517.54259-2-rongweil@nvidia.com>

Hi,


> -----Original Message-----
> From: Rongwei Liu <rongweil@nvidia.com>
> Sent: Tuesday, July 13, 2021 12:55 PM
> To: Matan Azrad <matan@nvidia.com>; Slava Ovsiienko
> <viacheslavo@nvidia.com>; Ori Kam <orika@nvidia.com>; NBU-Contact-
> Thomas Monjalon <thomas@monjalon.net>; Shahaf Shuler
> <shahafs@nvidia.com>
> Cc: dev@dpdk.org; Raslan Darawsheh <rasland@nvidia.com>
> Subject: [PATCH v5 1/2] net/mlx5: add VXLAN header the last 8-bits matching
> support
Title can be improved:
How about:
"net/mlx5: support matching on reserved field of VXLAN"
> 
> This update adds support for the VXLAN header last 8-bits
> matching when creating steering rules. At the PCIe probe
> stage, we create a dummy VXLAN matcher using misc5 to check
> rdma-core library's capability.
This adds matching on reserved field of VXLAN header (the last 8-bits).

The capability from both rdma-core and FW is detected by creating
a dummy matcher using misc5 when the device is probed.

> 
> The logic is, group 0 depends on HCA_CAP to enable misc or misc5
> for VXLAN matching while group non zero depends on the rdma-core
> capability.
> 
for none-zero groups the capability is detected from rdma-core,
meanwhile for group zero it's relying on the HCA_CAP from FW.

> Signed-off-by: Rongwei Liu <rongweil@nvidia.com>
> Acked-by: Viacheslav Ovsiienko <viacheslavo@nvidia.com>
> ---
>  doc/guides/nics/mlx5.rst             |  11 +-
>  drivers/common/mlx5/mlx5_devx_cmds.c |   3 +
>  drivers/common/mlx5/mlx5_devx_cmds.h |   6 +
>  drivers/common/mlx5/mlx5_prm.h       |  41 +++++--
>  drivers/net/mlx5/linux/mlx5_os.c     |  77 +++++++++++++
>  drivers/net/mlx5/mlx5.h              |   2 +
>  drivers/net/mlx5/mlx5_flow.c         |  26 ++++-
>  drivers/net/mlx5/mlx5_flow.h         |   4 +-
>  drivers/net/mlx5/mlx5_flow_dv.c      | 160 +++++++++++++++++++--------
>  drivers/net/mlx5/mlx5_flow_verbs.c   |   3 +-
>  drivers/vdpa/mlx5/mlx5_vdpa_steer.c  |   6 +-
>  11 files changed, 274 insertions(+), 65 deletions(-)
> 
> diff --git a/doc/guides/nics/mlx5.rst b/doc/guides/nics/mlx5.rst
> index 8253b96e92..5842991d5d 100644
> --- a/doc/guides/nics/mlx5.rst
> +++ b/doc/guides/nics/mlx5.rst
> @@ -195,8 +195,15 @@ Limitations
>    size and ``txq_inline_min`` settings and may be from 2 (worst case forced
> by maximal
>    inline settings) to 58.
> 
> -- Flows with a VXLAN Network Identifier equal (or ends to be equal)
> -  to 0 are not supported.
> +- Match on VXLAN supports the following fields only:
> +
> +     - VNI
> +     - Last reserved 8-bits
> +
> +  Last reserved 8-bits matching is only supported When using DV flow
> +  engine (``dv_flow_en`` = 1).
> +  Group zero's behavior may differ which depends on FW.
> +  Matching value equals 0 (value & mask) is not supported.
> 
>  - L3 VXLAN and VXLAN-GPE tunnels cannot be supported together with
> MPLSoGRE and MPLSoUDP.
> 
> diff --git a/drivers/common/mlx5/mlx5_devx_cmds.c
> b/drivers/common/mlx5/mlx5_devx_cmds.c
> index f5914bce32..63ae95832d 100644
> --- a/drivers/common/mlx5/mlx5_devx_cmds.c
> +++ b/drivers/common/mlx5/mlx5_devx_cmds.c
> @@ -947,6 +947,9 @@ mlx5_devx_cmd_query_hca_attr(void *ctx,
>  	attr->log_max_ft_sampler_num = MLX5_GET
>  		(flow_table_nic_cap, hcattr,
> 
> flow_table_properties_nic_receive.log_max_ft_sampler_num);
> +	attr->flow.tunnel_header_0_1 = MLX5_GET
> +		(flow_table_nic_cap, hcattr,
> +		 ft_field_support_2_nic_receive.tunnel_header_0_1);
>  	attr->pkt_integrity_match =
> mlx5_devx_query_pkt_integrity_match(hcattr);
>  	/* Query HCA offloads for Ethernet protocol. */
>  	memset(in, 0, sizeof(in));
> diff --git a/drivers/common/mlx5/mlx5_devx_cmds.h
> b/drivers/common/mlx5/mlx5_devx_cmds.h
> index f8a17b886b..124f43e852 100644
> --- a/drivers/common/mlx5/mlx5_devx_cmds.h
> +++ b/drivers/common/mlx5/mlx5_devx_cmds.h
> @@ -89,6 +89,11 @@ struct mlx5_hca_vdpa_attr {
>  	uint64_t doorbell_bar_offset;
>  };
> 
> +struct mlx5_hca_flow_attr {
> +	uint32_t tunnel_header_0_1;
> +	uint32_t tunnel_header_2_3;
> +};
> +
>  /* HCA supports this number of time periods for LRO. */
>  #define MLX5_LRO_NUM_SUPP_PERIODS 4
> 
> @@ -155,6 +160,7 @@ struct mlx5_hca_attr {
>  	uint32_t pkt_integrity_match:1; /* 1 if HW supports integrity item */
>  	struct mlx5_hca_qos_attr qos;
>  	struct mlx5_hca_vdpa_attr vdpa;
> +	struct mlx5_hca_flow_attr flow;
>  	int log_max_qp_sz;
>  	int log_max_cq_sz;
>  	int log_max_qp;
> diff --git a/drivers/common/mlx5/mlx5_prm.h
> b/drivers/common/mlx5/mlx5_prm.h
> index 26761f5bd3..7950070976 100644
> --- a/drivers/common/mlx5/mlx5_prm.h
> +++ b/drivers/common/mlx5/mlx5_prm.h
> @@ -977,6 +977,18 @@ struct mlx5_ifc_fte_match_set_misc4_bits {
>  	u8 reserved_at_100[0x100];
>  };
> 
> +struct mlx5_ifc_fte_match_set_misc5_bits {
> +	u8 macsec_tag_0[0x20];
> +	u8 macsec_tag_1[0x20];
> +	u8 macsec_tag_2[0x20];
> +	u8 macsec_tag_3[0x20];
> +	u8 tunnel_header_0[0x20];
> +	u8 tunnel_header_1[0x20];
> +	u8 tunnel_header_2[0x20];
> +	u8 tunnel_header_3[0x20];
> +	u8 reserved[0x100];
> +};
> +
>  /* Flow matcher. */
>  struct mlx5_ifc_fte_match_param_bits {
>  	struct mlx5_ifc_fte_match_set_lyr_2_4_bits outer_headers;
> @@ -985,12 +997,13 @@ struct mlx5_ifc_fte_match_param_bits {
>  	struct mlx5_ifc_fte_match_set_misc2_bits misc_parameters_2;
>  	struct mlx5_ifc_fte_match_set_misc3_bits misc_parameters_3;
>  	struct mlx5_ifc_fte_match_set_misc4_bits misc_parameters_4;
> +	struct mlx5_ifc_fte_match_set_misc5_bits misc_parameters_5;
>  /*
>   * Add reserved bit to match the struct size with the size defined in PRM.
>   * This extension is not required in Linux.
>   */
>  #ifndef HAVE_INFINIBAND_VERBS_H
> -	u8 reserved_0[0x400];
> +	u8 reserved_0[0x200];
>  #endif
>  };
> 
> @@ -1007,6 +1020,7 @@ enum {
>  	MLX5_MATCH_CRITERIA_ENABLE_MISC2_BIT,
>  	MLX5_MATCH_CRITERIA_ENABLE_MISC3_BIT,
>  	MLX5_MATCH_CRITERIA_ENABLE_MISC4_BIT,
> +	MLX5_MATCH_CRITERIA_ENABLE_MISC5_BIT,
>  };
> 
>  enum {
> @@ -1784,7 +1798,12 @@ struct mlx5_ifc_roce_caps_bits {
>   * Table 1872 - Flow Table Fields Supported 2 Format
>   */
>  struct mlx5_ifc_ft_fields_support_2_bits {
> -	u8 reserved_at_0[0x14];
> +	u8 reserved_at_0[0xf];
> +	u8 tunnel_header_2_3[0x1];
> +	u8 tunnel_header_0_1[0x1];
> +	u8 macsec_syndrome[0x1];
> +	u8 macsec_tag[0x1];
> +	u8 outer_lrh_sl[0x1];
>  	u8 inner_ipv4_ihl[0x1];
>  	u8 outer_ipv4_ihl[0x1];
>  	u8 psp_syndrome[0x1];
> @@ -1797,18 +1816,26 @@ struct mlx5_ifc_ft_fields_support_2_bits {
>  	u8 inner_l4_checksum_ok[0x1];
>  	u8 outer_ipv4_checksum_ok[0x1];
>  	u8 outer_l4_checksum_ok[0x1];
> +	u8 reserved_at_20[0x60];
>  };
> 
>  struct mlx5_ifc_flow_table_nic_cap_bits {
>  	u8 reserved_at_0[0x200];
>  	struct mlx5_ifc_flow_table_prop_layout_bits
> -	       flow_table_properties_nic_receive;
> +		flow_table_properties_nic_receive;
> +	struct mlx5_ifc_flow_table_prop_layout_bits
> +		flow_table_properties_nic_receive_rdma;
> +	struct mlx5_ifc_flow_table_prop_layout_bits
> +		flow_table_properties_nic_receive_sniffer;
> +	struct mlx5_ifc_flow_table_prop_layout_bits
> +		flow_table_properties_nic_transmit;
> +	struct mlx5_ifc_flow_table_prop_layout_bits
> +		flow_table_properties_nic_transmit_rdma;
>  	struct mlx5_ifc_flow_table_prop_layout_bits
> -	       flow_table_properties_unused[5];
> -	u8 reserved_at_1C0[0x200];
> -	u8 header_modify_nic_receive[0x400];
> +		flow_table_properties_nic_transmit_sniffer;
> +	u8 reserved_at_e00[0x600];
>  	struct mlx5_ifc_ft_fields_support_2_bits
> -	       ft_field_support_2_nic_receive;
> +		ft_field_support_2_nic_receive;
>  };
> 
>  struct mlx5_ifc_cmd_hca_cap_2_bits {
> diff --git a/drivers/net/mlx5/linux/mlx5_os.c
> b/drivers/net/mlx5/linux/mlx5_os.c
> index be22d9cbd2..55bb71c170 100644
> --- a/drivers/net/mlx5/linux/mlx5_os.c
> +++ b/drivers/net/mlx5/linux/mlx5_os.c
> @@ -193,6 +193,79 @@ mlx5_alloc_verbs_buf(size_t size, void *data)
>  	return ret;
>  }
> 
> +/**
> + * Detect misc5 support or not
> + *
> + * @param[in] priv
> + *   Device private data pointer
> + */
> +#ifdef HAVE_MLX5DV_DR
> +static void
> +__mlx5_discovery_misc5_cap(struct mlx5_priv *priv)
> +{
> +#ifdef HAVE_IBV_FLOW_DV_SUPPORT
> +	/* Dummy VxLAN matcher to detect rdma-core misc5 cap
> +	 * Case: IPv4--->UDP--->VxLAN--->vni
> +	 */
> +	void *tbl;
> +	struct mlx5_flow_dv_match_params matcher_mask;
> +	void *match_m;
> +	void *matcher;
> +	void *headers_m;
> +	void *misc5_m;
> +	uint32_t *tunnel_header_m;
> +	struct mlx5dv_flow_matcher_attr dv_attr;
> +
> +	memset(&matcher_mask, 0, sizeof(matcher_mask));
> +	matcher_mask.size = sizeof(matcher_mask.buf);
> +	match_m = matcher_mask.buf;
> +	headers_m = MLX5_ADDR_OF(fte_match_param, match_m,
> outer_headers);
> +	misc5_m = MLX5_ADDR_OF(fte_match_param,
> +			       match_m, misc_parameters_5);
> +	tunnel_header_m = (uint32_t *)
> +				MLX5_ADDR_OF(fte_match_set_misc5,
> +				misc5_m, tunnel_header_1);
> +	MLX5_SET(fte_match_set_lyr_2_4, headers_m, ip_protocol, 0xff);
> +	MLX5_SET(fte_match_set_lyr_2_4, headers_m, ip_version, 4);
> +	MLX5_SET(fte_match_set_lyr_2_4, headers_m, udp_dport, 0xffff);
> +	*tunnel_header_m = 0xffffff;
> +
> +	tbl = mlx5_glue->dr_create_flow_tbl(priv->sh->rx_domain, 1);
> +	if (!tbl) {
> +		DRV_LOG(INFO, "No SW steering support");
> +		return;
> +	}
> +	dv_attr.type = IBV_FLOW_ATTR_NORMAL,
> +	dv_attr.match_mask = (void *)&matcher_mask,
> +	dv_attr.match_criteria_enable =
> +			(1 << MLX5_MATCH_CRITERIA_ENABLE_OUTER_BIT)
> |
> +			(1 << MLX5_MATCH_CRITERIA_ENABLE_MISC5_BIT);
> +	dv_attr.priority = 3;
> +#ifdef HAVE_MLX5DV_DR_ESWITCH
> +	void *misc2_m;
> +	if (priv->config.dv_esw_en) {
> +		/* FDB enabled reg_c_0 */
> +		dv_attr.match_criteria_enable |=
> +				(1 <<
> MLX5_MATCH_CRITERIA_ENABLE_MISC2_BIT);
> +		misc2_m = MLX5_ADDR_OF(fte_match_param,
> +				       match_m, misc_parameters_2);
> +		MLX5_SET(fte_match_set_misc2, misc2_m,
> +			 metadata_reg_c_0, 0xffff);
> +	}
> +#endif
> +	matcher = mlx5_glue->dv_create_flow_matcher(priv->sh->ctx,
> +						    &dv_attr, tbl);
> +	if (matcher) {
> +		priv->sh->misc5_cap = 1;
> +		mlx5_glue->dv_destroy_flow_matcher(matcher);
> +	}
> +	mlx5_glue->dr_destroy_flow_tbl(tbl);
> +#else
> +	RTE_SET_USED(priv);
> +#endif
> +}
> +#endif
> +
>  /**
>   * Verbs callback to free a memory.
>   *
> @@ -364,6 +437,8 @@ mlx5_alloc_shared_dr(struct mlx5_priv *priv)
>  		if (sh->fdb_domain)
>  			mlx5_glue->dr_allow_duplicate_rules(sh-
> >fdb_domain, 0);
>  	}
> +
> +	__mlx5_discovery_misc5_cap(priv);
>  #endif /* HAVE_MLX5DV_DR */
>  	sh->default_miss_action =
>  			mlx5_glue->dr_create_flow_action_default_miss();
> @@ -1313,6 +1388,8 @@ mlx5_dev_spawn(struct rte_device *dpdk_dev,
>  				goto error;
>  			}
>  		}
> +		if (config->hca_attr.flow.tunnel_header_0_1)
> +			sh->tunnel_header_0_1 = 1;
>  #endif
>  #ifdef HAVE_MLX5_DR_CREATE_ACTION_ASO
>  		if (config->hca_attr.flow_hit_aso &&
> diff --git a/drivers/net/mlx5/mlx5.h b/drivers/net/mlx5/mlx5.h
> index f864c1d701..75a0e04ea0 100644
> --- a/drivers/net/mlx5/mlx5.h
> +++ b/drivers/net/mlx5/mlx5.h
> @@ -1094,6 +1094,8 @@ struct mlx5_dev_ctx_shared {
>  	uint32_t qp_ts_format:2; /* QP timestamp formats supported. */
>  	uint32_t meter_aso_en:1; /* Flow Meter ASO is supported. */
>  	uint32_t ct_aso_en:1; /* Connection Tracking ASO is supported. */
> +	uint32_t tunnel_header_0_1:1; /* tunnel_header_0_1 is supported.
> */
> +	uint32_t misc5_cap:1; /* misc5 matcher parameter is supported. */
>  	uint32_t max_port; /* Maximal IB device port index. */
>  	struct mlx5_bond_info bond; /* Bonding information. */
>  	void *ctx; /* Verbs/DV/DevX context. */
> diff --git a/drivers/net/mlx5/mlx5_flow.c b/drivers/net/mlx5/mlx5_flow.c
> index 2feddb0254..f3f5752dbe 100644
> --- a/drivers/net/mlx5/mlx5_flow.c
> +++ b/drivers/net/mlx5/mlx5_flow.c
> @@ -2410,12 +2410,14 @@ mlx5_flow_validate_item_tcp(const struct
> rte_flow_item *item,
>  /**
>   * Validate VXLAN item.
>   *
> + * @param[in] dev
> + *   Pointer to the Ethernet device structure.
>   * @param[in] item
>   *   Item specification.
>   * @param[in] item_flags
>   *   Bit-fields that holds the items detected until now.
> - * @param[in] target_protocol
> - *   The next protocol in the previous item.
> + * @param[in] attr
> + *   Flow rule attributes.
>   * @param[out] error
>   *   Pointer to error structure.
>   *
> @@ -2423,24 +2425,32 @@ mlx5_flow_validate_item_tcp(const struct
> rte_flow_item *item,
>   *   0 on success, a negative errno value otherwise and rte_errno is set.
>   */
>  int
> -mlx5_flow_validate_item_vxlan(const struct rte_flow_item *item,
> +mlx5_flow_validate_item_vxlan(struct rte_eth_dev *dev,
> +			      const struct rte_flow_item *item,
>  			      uint64_t item_flags,
> +			      const struct rte_flow_attr *attr,
>  			      struct rte_flow_error *error)
>  {
>  	const struct rte_flow_item_vxlan *spec = item->spec;
>  	const struct rte_flow_item_vxlan *mask = item->mask;
>  	int ret;
> +	struct mlx5_priv *priv = dev->data->dev_private;
>  	union vni {
>  		uint32_t vlan_id;
>  		uint8_t vni[4];
>  	} id = { .vlan_id = 0, };
> -
> +	const struct rte_flow_item_vxlan nic_mask = {
> +		.vni = "\xff\xff\xff",
> +		.rsvd1 = 0xff,
> +	};
> +	const struct rte_flow_item_vxlan *valid_mask;
> 
>  	if (item_flags & MLX5_FLOW_LAYER_TUNNEL)
>  		return rte_flow_error_set(error, ENOTSUP,
>  					  RTE_FLOW_ERROR_TYPE_ITEM,
> item,
>  					  "multiple tunnel layers not"
>  					  " supported");
> +	valid_mask = &rte_flow_item_vxlan_mask;
>  	/*
>  	 * Verify only UDPv4 is present as defined in
>  	 * https://tools.ietf.org/html/rfc7348
> @@ -2451,9 +2461,15 @@ mlx5_flow_validate_item_vxlan(const struct
> rte_flow_item *item,
>  					  "no outer UDP layer found");
>  	if (!mask)
>  		mask = &rte_flow_item_vxlan_mask;
> +	/* FDB domain & NIC domain non-zero group */
> +	if ((attr->transfer || attr->group) && priv->sh->misc5_cap)
> +		valid_mask = &nic_mask;
> +	/* Group zero in NIC domain */
> +	if (!attr->group && !attr->transfer && priv->sh-
> >tunnel_header_0_1)
> +		valid_mask = &nic_mask;
>  	ret = mlx5_flow_item_acceptable
>  		(item, (const uint8_t *)mask,
> -		 (const uint8_t *)&rte_flow_item_vxlan_mask,
> +		 (const uint8_t *)valid_mask,
>  		 sizeof(struct rte_flow_item_vxlan),
>  		 MLX5_ITEM_RANGE_NOT_ACCEPTED, error);
>  	if (ret < 0)
> diff --git a/drivers/net/mlx5/mlx5_flow.h b/drivers/net/mlx5/mlx5_flow.h
> index 7d97c5880f..66a38c3630 100644
> --- a/drivers/net/mlx5/mlx5_flow.h
> +++ b/drivers/net/mlx5/mlx5_flow.h
> @@ -1533,8 +1533,10 @@ int mlx5_flow_validate_item_vlan(const struct
> rte_flow_item *item,
>  				 uint64_t item_flags,
>  				 struct rte_eth_dev *dev,
>  				 struct rte_flow_error *error);
> -int mlx5_flow_validate_item_vxlan(const struct rte_flow_item *item,
> +int mlx5_flow_validate_item_vxlan(struct rte_eth_dev *dev,
> +				  const struct rte_flow_item *item,
>  				  uint64_t item_flags,
> +				  const struct rte_flow_attr *attr,
>  				  struct rte_flow_error *error);
>  int mlx5_flow_validate_item_vxlan_gpe(const struct rte_flow_item *item,
>  				      uint64_t item_flags,
> diff --git a/drivers/net/mlx5/mlx5_flow_dv.c
> b/drivers/net/mlx5/mlx5_flow_dv.c
> index 2f4c0eeb5b..6c3715a5e8 100644
> --- a/drivers/net/mlx5/mlx5_flow_dv.c
> +++ b/drivers/net/mlx5/mlx5_flow_dv.c
> @@ -6930,7 +6930,8 @@ flow_dv_validate(struct rte_eth_dev *dev, const
> struct rte_flow_attr *attr,
>  			last_item = MLX5_FLOW_LAYER_GRE_KEY;
>  			break;
>  		case RTE_FLOW_ITEM_TYPE_VXLAN:
> -			ret = mlx5_flow_validate_item_vxlan(items,
> item_flags,
> +			ret = mlx5_flow_validate_item_vxlan(dev, items,
> +							    item_flags, attr,
>  							    error);
>  			if (ret < 0)
>  				return ret;
> @@ -7892,15 +7893,7 @@ flow_dv_prepare(struct rte_eth_dev *dev,
>  	memset(dev_flow, 0, sizeof(*dev_flow));
>  	dev_flow->handle = dev_handle;
>  	dev_flow->handle_idx = handle_idx;
> -	/*
> -	 * In some old rdma-core releases, before continuing, a check of the
> -	 * length of matching parameter will be done at first. It needs to use
> -	 * the length without misc4 param. If the flow has misc4 support,
> then
> -	 * the length needs to be adjusted accordingly. Each param member
> is
> -	 * aligned with a 64B boundary naturally.
> -	 */
> -	dev_flow->dv.value.size = MLX5_ST_SZ_BYTES(fte_match_param) -
> -				  MLX5_ST_SZ_BYTES(fte_match_set_misc4);
> +	dev_flow->dv.value.size = MLX5_ST_SZ_BYTES(fte_match_param);
>  	dev_flow->ingress = attr->ingress;
>  	dev_flow->dv.transfer = attr->transfer;
>  	return dev_flow;
> @@ -8681,6 +8674,10 @@ flow_dv_translate_item_nvgre(void *matcher,
> void *key,
>  /**
>   * Add VXLAN item to matcher and to the value.
>   *
> + * @param[in] dev
> + *   Pointer to the Ethernet device structure.
> + * @param[in] attr
> + *   Flow rule attributes.
>   * @param[in, out] matcher
>   *   Flow matcher.
>   * @param[in, out] key
> @@ -8691,7 +8688,9 @@ flow_dv_translate_item_nvgre(void *matcher,
> void *key,
>   *   Item is inner pattern.
>   */
>  static void
> -flow_dv_translate_item_vxlan(void *matcher, void *key,
> +flow_dv_translate_item_vxlan(struct rte_eth_dev *dev,
> +			     const struct rte_flow_attr *attr,
> +			     void *matcher, void *key,
>  			     const struct rte_flow_item *item,
>  			     int inner)
>  {
> @@ -8699,13 +8698,16 @@ flow_dv_translate_item_vxlan(void *matcher,
> void *key,
>  	const struct rte_flow_item_vxlan *vxlan_v = item->spec;
>  	void *headers_m;
>  	void *headers_v;
> -	void *misc_m = MLX5_ADDR_OF(fte_match_param, matcher,
> misc_parameters);
> -	void *misc_v = MLX5_ADDR_OF(fte_match_param, key,
> misc_parameters);
> -	char *vni_m;
> -	char *vni_v;
> +	void *misc5_m;
> +	void *misc5_v;
> +	uint32_t *tunnel_header_v;
> +	uint32_t *tunnel_header_m;
>  	uint16_t dport;
> -	int size;
> -	int i;
> +	struct mlx5_priv *priv = dev->data->dev_private;
> +	const struct rte_flow_item_vxlan nic_mask = {
> +		.vni = "\xff\xff\xff",
> +		.rsvd1 = 0xff,
> +	};
> 
>  	if (inner) {
>  		headers_m = MLX5_ADDR_OF(fte_match_param, matcher,
> @@ -8724,14 +8726,52 @@ flow_dv_translate_item_vxlan(void *matcher,
> void *key,
>  	}
>  	if (!vxlan_v)
>  		return;
> -	if (!vxlan_m)
> -		vxlan_m = &rte_flow_item_vxlan_mask;
> -	size = sizeof(vxlan_m->vni);
> -	vni_m = MLX5_ADDR_OF(fte_match_set_misc, misc_m, vxlan_vni);
> -	vni_v = MLX5_ADDR_OF(fte_match_set_misc, misc_v, vxlan_vni);
> -	memcpy(vni_m, vxlan_m->vni, size);
> -	for (i = 0; i < size; ++i)
> -		vni_v[i] = vni_m[i] & vxlan_v->vni[i];
> +	if (!vxlan_m) {
> +		if ((!attr->group && !priv->sh->tunnel_header_0_1) ||
> +		    (attr->group && !priv->sh->misc5_cap))
> +			vxlan_m = &rte_flow_item_vxlan_mask;
> +		else
> +			vxlan_m = &nic_mask;
> +	}
> +	if ((!attr->group && !attr->transfer && !priv->sh-
> >tunnel_header_0_1) ||
> +	    ((attr->group || attr->transfer) && !priv->sh->misc5_cap)) {
> +		void *misc_m;
> +		void *misc_v;
> +		char *vni_m;
> +		char *vni_v;
> +		int size;
> +		int i;
> +		misc_m = MLX5_ADDR_OF(fte_match_param,
> +				      matcher, misc_parameters);
> +		misc_v = MLX5_ADDR_OF(fte_match_param, key,
> misc_parameters);
> +		size = sizeof(vxlan_m->vni);
> +		vni_m = MLX5_ADDR_OF(fte_match_set_misc, misc_m,
> vxlan_vni);
> +		vni_v = MLX5_ADDR_OF(fte_match_set_misc, misc_v,
> vxlan_vni);
> +		memcpy(vni_m, vxlan_m->vni, size);
> +		for (i = 0; i < size; ++i)
> +			vni_v[i] = vni_m[i] & vxlan_v->vni[i];
> +		return;
> +	}
> +	misc5_m = MLX5_ADDR_OF(fte_match_param, matcher,
> misc_parameters_5);
> +	misc5_v = MLX5_ADDR_OF(fte_match_param, key,
> misc_parameters_5);
> +	tunnel_header_v = (uint32_t
> *)MLX5_ADDR_OF(fte_match_set_misc5,
> +						   misc5_v,
> +						   tunnel_header_1);
> +	tunnel_header_m = (uint32_t
> *)MLX5_ADDR_OF(fte_match_set_misc5,
> +						   misc5_m,
> +						   tunnel_header_1);
> +	*tunnel_header_v = (vxlan_v->vni[0] & vxlan_m->vni[0]) |
> +			   (vxlan_v->vni[1] & vxlan_m->vni[1]) << 8 |
> +			   (vxlan_v->vni[2] & vxlan_m->vni[2]) << 16;
> +	if (*tunnel_header_v)
> +		*tunnel_header_m = vxlan_m->vni[0] |
> +			vxlan_m->vni[1] << 8 |
> +			vxlan_m->vni[2] << 16;
> +	else
> +		*tunnel_header_m = 0x0;
> +	*tunnel_header_v |= (vxlan_v->rsvd1 & vxlan_m->rsvd1) << 24;
> +	if (vxlan_v->rsvd1 & vxlan_m->rsvd1)
> +		*tunnel_header_m |= vxlan_m->rsvd1 << 24;
>  }
> 
>  /**
> @@ -9892,9 +9932,32 @@ flow_dv_matcher_enable(uint32_t
> *match_criteria)
>  	match_criteria_enable |=
>  		(!HEADER_IS_ZERO(match_criteria, misc_parameters_4)) <<
>  		MLX5_MATCH_CRITERIA_ENABLE_MISC4_BIT;
> +	match_criteria_enable |=
> +		(!HEADER_IS_ZERO(match_criteria, misc_parameters_5)) <<
> +		MLX5_MATCH_CRITERIA_ENABLE_MISC5_BIT;
>  	return match_criteria_enable;
>  }
> 
> +static void
> +__flow_dv_adjust_buf_size(size_t *size, uint8_t match_criteria)
> +{
> +	/*
> +	 * Check flow matching criteria first, subtract misc5/4 length if flow
> +	 * doesn't own misc5/4 parameters. In some old rdma-core releases,
> +	 * misc5/4 are not supported, and matcher creation failure is
> expected
> +	 * w/o subtration. If misc5 is provided, misc4 must be counted in
> since
> +	 * misc5 is right after misc4.
> +	 */
> +	if (!(match_criteria & (1 <<
> MLX5_MATCH_CRITERIA_ENABLE_MISC5_BIT))) {
> +		*size = MLX5_ST_SZ_BYTES(fte_match_param) -
> +			MLX5_ST_SZ_BYTES(fte_match_set_misc5);
> +		if (!(match_criteria & (1 <<
> +			MLX5_MATCH_CRITERIA_ENABLE_MISC4_BIT))) {
> +			*size -= MLX5_ST_SZ_BYTES(fte_match_set_misc4);
> +		}
> +	}
> +}
> +
>  struct mlx5_hlist_entry *
>  flow_dv_tbl_create_cb(struct mlx5_hlist *list, uint64_t key64, void *cb_ctx)
>  {
> @@ -10161,6 +10224,8 @@ flow_dv_matcher_create_cb(struct
> mlx5_cache_list *list,
>  	*cache = *ref;
>  	dv_attr.match_criteria_enable =
>  		flow_dv_matcher_enable(cache->mask.buf);
> +	__flow_dv_adjust_buf_size(&ref->mask.size,
> +				  dv_attr.match_criteria_enable);
>  	dv_attr.priority = ref->priority;
>  	if (tbl->is_egress)
>  		dv_attr.flags |= IBV_FLOW_ATTR_FLAGS_EGRESS;
> @@ -10210,7 +10275,6 @@ flow_dv_matcher_register(struct rte_eth_dev
> *dev,
>  		.error = error,
>  		.data = ref,
>  	};
> -
>  	/**
>  	 * tunnel offload API requires this registration for cases when
>  	 * tunnel match rule was inserted before tunnel set rule.
> @@ -12069,8 +12133,7 @@ flow_dv_translate(struct rte_eth_dev *dev,
>  	uint64_t action_flags = 0;
>  	struct mlx5_flow_dv_matcher matcher = {
>  		.mask = {
> -			.size = sizeof(matcher.mask.buf) -
> -				MLX5_ST_SZ_BYTES(fte_match_set_misc4),
> +			.size = sizeof(matcher.mask.buf),
>  		},
>  	};
>  	int actions_n = 0;
> @@ -12877,7 +12940,8 @@ flow_dv_translate(struct rte_eth_dev *dev,
>  			last_item = MLX5_FLOW_LAYER_GRE;
>  			break;
>  		case RTE_FLOW_ITEM_TYPE_VXLAN:
> -			flow_dv_translate_item_vxlan(match_mask,
> match_value,
> +			flow_dv_translate_item_vxlan(dev, attr,
> +						     match_mask,
> match_value,
>  						     items, tunnel);
>  			matcher.priority =
> MLX5_TUNNEL_PRIO_GET(rss_desc);
>  			last_item = MLX5_FLOW_LAYER_VXLAN;
> @@ -12975,10 +13039,6 @@ flow_dv_translate(struct rte_eth_dev *dev,
>  						NULL,
>  						"cannot create eCPRI
> parser");
>  			}
> -			/* Adjust the length matcher and device flow value.
> */
> -			matcher.mask.size =
> MLX5_ST_SZ_BYTES(fte_match_param);
> -			dev_flow->dv.value.size =
> -
> 	MLX5_ST_SZ_BYTES(fte_match_param);
>  			flow_dv_translate_item_ecpri(dev, match_mask,
>  						     match_value, items);
>  			/* No other protocol should follow eCPRI layer. */
> @@ -13288,6 +13348,7 @@ flow_dv_apply(struct rte_eth_dev *dev, struct
> rte_flow *flow,
>  	int idx;
>  	struct mlx5_flow_workspace *wks =
> mlx5_flow_get_thread_workspace();
>  	struct mlx5_flow_rss_desc *rss_desc = &wks->rss_desc;
> +	uint8_t misc_mask;
> 
>  	MLX5_ASSERT(wks);
>  	for (idx = wks->flow_idx - 1; idx >= 0; idx--) {
> @@ -13358,6 +13419,8 @@ flow_dv_apply(struct rte_eth_dev *dev, struct
> rte_flow *flow,
>  			}
>  			dv->actions[n++] = priv->sh->default_miss_action;
>  		}
> +		misc_mask = flow_dv_matcher_enable(dv->value.buf);
> +		__flow_dv_adjust_buf_size(&dv->value.size, misc_mask);
>  		err = mlx5_flow_os_create_flow(dv_h->matcher-
> >matcher_object,
>  					       (void *)&dv->value, n,
>  					       dv->actions, &dh->drv_flow);
> @@ -15476,14 +15539,13 @@ __flow_dv_create_policy_flow(struct
> rte_eth_dev *dev,
>  {
>  	int ret;
>  	struct mlx5_flow_dv_match_params value = {
> -		.size = sizeof(value.buf) -
> -			MLX5_ST_SZ_BYTES(fte_match_set_misc4),
> +		.size = sizeof(value.buf),
>  	};
>  	struct mlx5_flow_dv_match_params matcher = {
> -		.size = sizeof(matcher.buf) -
> -			MLX5_ST_SZ_BYTES(fte_match_set_misc4),
> +		.size = sizeof(matcher.buf),
>  	};
>  	struct mlx5_priv *priv = dev->data->dev_private;
> +	uint8_t misc_mask;
> 
>  	if (match_src_port && (priv->representor || priv->master)) {
>  		if (flow_dv_translate_item_port_id(dev, matcher.buf,
> @@ -15497,6 +15559,8 @@ __flow_dv_create_policy_flow(struct
> rte_eth_dev *dev,
>  				(enum modify_reg)color_reg_c_idx,
>  				rte_col_2_mlx5_col(color),
>  				UINT32_MAX);
> +	misc_mask = flow_dv_matcher_enable(value.buf);
> +	__flow_dv_adjust_buf_size(&value.size, misc_mask);
>  	ret = mlx5_flow_os_create_flow(matcher_object,
>  			(void *)&value, actions_n, actions, rule);
>  	if (ret) {
> @@ -15521,14 +15585,12 @@ __flow_dv_create_policy_matcher(struct
> rte_eth_dev *dev,
>  	struct mlx5_flow_tbl_resource *tbl_rsc = sub_policy->tbl_rsc;
>  	struct mlx5_flow_dv_matcher matcher = {
>  		.mask = {
> -			.size = sizeof(matcher.mask.buf) -
> -				MLX5_ST_SZ_BYTES(fte_match_set_misc4),
> +			.size = sizeof(matcher.mask.buf),
>  		},
>  		.tbl = tbl_rsc,
>  	};
>  	struct mlx5_flow_dv_match_params value = {
> -		.size = sizeof(value.buf) -
> -			MLX5_ST_SZ_BYTES(fte_match_set_misc4),
> +		.size = sizeof(value.buf),
>  	};
>  	struct mlx5_flow_cb_ctx ctx = {
>  		.error = error,
> @@ -16002,12 +16064,10 @@ flow_dv_create_mtr_tbls(struct rte_eth_dev
> *dev,
>  	int domain, ret, i;
>  	struct mlx5_flow_counter *cnt;
>  	struct mlx5_flow_dv_match_params value = {
> -		.size = sizeof(value.buf) -
> -		MLX5_ST_SZ_BYTES(fte_match_set_misc4),
> +		.size = sizeof(value.buf),
>  	};
>  	struct mlx5_flow_dv_match_params matcher_para = {
> -		.size = sizeof(matcher_para.buf) -
> -		MLX5_ST_SZ_BYTES(fte_match_set_misc4),
> +		.size = sizeof(matcher_para.buf),
>  	};
>  	int mtr_id_reg_c = mlx5_flow_get_reg_id(dev, MLX5_MTR_ID,
>  						     0, &error);
> @@ -16016,8 +16076,7 @@ flow_dv_create_mtr_tbls(struct rte_eth_dev
> *dev,
>  	struct mlx5_cache_entry *entry;
>  	struct mlx5_flow_dv_matcher matcher = {
>  		.mask = {
> -			.size = sizeof(matcher.mask.buf) -
> -			MLX5_ST_SZ_BYTES(fte_match_set_misc4),
> +			.size = sizeof(matcher.mask.buf),
>  		},
>  	};
>  	struct mlx5_flow_dv_matcher *drop_matcher;
> @@ -16025,6 +16084,7 @@ flow_dv_create_mtr_tbls(struct rte_eth_dev
> *dev,
>  		.error = &error,
>  		.data = &matcher,
>  	};
> +	uint8_t misc_mask;
> 
>  	if (!priv->mtr_en || mtr_id_reg_c < 0) {
>  		rte_errno = ENOTSUP;
> @@ -16074,6 +16134,8 @@ flow_dv_create_mtr_tbls(struct rte_eth_dev
> *dev,
>  			actions[i++] = priv->sh->dr_drop_action;
>  			flow_dv_match_meta_reg(matcher_para.buf,
> value.buf,
>  				(enum modify_reg)mtr_id_reg_c, 0, 0);
> +			misc_mask = flow_dv_matcher_enable(value.buf);
> +			__flow_dv_adjust_buf_size(&value.size,
> misc_mask);
>  			ret = mlx5_flow_os_create_flow
>  				(mtrmng->def_matcher[domain]-
> >matcher_object,
>  				(void *)&value, i, actions,
> @@ -16117,6 +16179,8 @@ flow_dv_create_mtr_tbls(struct rte_eth_dev
> *dev,
>  					fm->drop_cnt, NULL);
>  		actions[i++] = cnt->action;
>  		actions[i++] = priv->sh->dr_drop_action;
> +		misc_mask = flow_dv_matcher_enable(value.buf);
> +		__flow_dv_adjust_buf_size(&value.size, misc_mask);
>  		ret = mlx5_flow_os_create_flow(drop_matcher-
> >matcher_object,
>  					       (void *)&value, i, actions,
>  					       &fm->drop_rule[domain]);
> @@ -16637,10 +16701,12 @@
> mlx5_flow_dv_discover_counter_offset_support(struct rte_eth_dev *dev)
>  	if (ret)
>  		goto err;
>  	dv_attr.match_criteria_enable =
> flow_dv_matcher_enable(mask.buf);
> +	__flow_dv_adjust_buf_size(&mask.size,
> dv_attr.match_criteria_enable);
>  	ret = mlx5_flow_os_create_flow_matcher(sh->ctx, &dv_attr, tbl-
> >obj,
>  					       &matcher);
>  	if (ret)
>  		goto err;
> +	__flow_dv_adjust_buf_size(&value.size,
> dv_attr.match_criteria_enable);
>  	ret = mlx5_flow_os_create_flow(matcher, (void *)&value, 1,
>  				       actions, &flow);
>  err:
> diff --git a/drivers/net/mlx5/mlx5_flow_verbs.c
> b/drivers/net/mlx5/mlx5_flow_verbs.c
> index fe9673310a..7b3d0b320d 100644
> --- a/drivers/net/mlx5/mlx5_flow_verbs.c
> +++ b/drivers/net/mlx5/mlx5_flow_verbs.c
> @@ -1381,7 +1381,8 @@ flow_verbs_validate(struct rte_eth_dev *dev,
> 
> MLX5_FLOW_LAYER_OUTER_L4_TCP;
>  			break;
>  		case RTE_FLOW_ITEM_TYPE_VXLAN:
> -			ret = mlx5_flow_validate_item_vxlan(items,
> item_flags,
> +			ret = mlx5_flow_validate_item_vxlan(dev, items,
> +							    item_flags, attr,
>  							    error);
>  			if (ret < 0)
>  				return ret;
> diff --git a/drivers/vdpa/mlx5/mlx5_vdpa_steer.c
> b/drivers/vdpa/mlx5/mlx5_vdpa_steer.c
> index 1fcd24c002..383f003966 100644
> --- a/drivers/vdpa/mlx5/mlx5_vdpa_steer.c
> +++ b/drivers/vdpa/mlx5/mlx5_vdpa_steer.c
> @@ -140,11 +140,13 @@ mlx5_vdpa_rss_flows_create(struct
> mlx5_vdpa_priv *priv)
>  		/**< Matcher value. This value is used as the mask or a key.
> */
>  	} matcher_mask = {
>  				.size = sizeof(matcher_mask.buf) -
> -
> 	MLX5_ST_SZ_BYTES(fte_match_set_misc4),
> +
> 	MLX5_ST_SZ_BYTES(fte_match_set_misc4) -
> +
> 	MLX5_ST_SZ_BYTES(fte_match_set_misc5),
>  			},
>  	  matcher_value = {
>  				.size = sizeof(matcher_value.buf) -
> -
> 	MLX5_ST_SZ_BYTES(fte_match_set_misc4),
> +
> 	MLX5_ST_SZ_BYTES(fte_match_set_misc4) -
> +
> 	MLX5_ST_SZ_BYTES(fte_match_set_misc5),
>  			};
>  	struct mlx5dv_flow_matcher_attr dv_attr = {
>  		.type = IBV_FLOW_ATTR_NORMAL,
> --
> 2.27.0


  reply	other threads:[~2021-07-13 10:27 UTC|newest]

Thread overview: 34+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2021-07-05  9:50 [dpdk-dev] [PATCH v2 0/2] support VXLAN header last 8-bits reserved field matching rongwei liu
2021-07-05  9:50 ` [dpdk-dev] [PATCH v2 1/2] drivers: add VXLAN header the last 8-bits matching support rongwei liu
2021-07-06 12:35   ` Thomas Monjalon
2021-07-07  8:09     ` [dpdk-dev] [PATCH v4 0/2] support VXLAN header the last 8-bits matching Rongwei Liu
2021-07-07  8:09       ` [dpdk-dev] [PATCH v4 1/2] net/mlx5: add VXLAN header the last 8-bits matching support Rongwei Liu
2021-07-07  8:09       ` [dpdk-dev] [PATCH v4 2/2] app/testpmd: support VXLAN the last 8-bits field matching Rongwei Liu
2021-07-13  8:33       ` [dpdk-dev] [PATCH v4 0/2] support VXLAN header the last 8-bits matching Andrew Rybchenko
2021-07-13  9:55         ` [dpdk-dev] [PATCH v5 " Rongwei Liu
2021-07-13  9:55           ` [dpdk-dev] [PATCH v5 1/2] net/mlx5: add VXLAN header the last 8-bits matching support Rongwei Liu
2021-07-13 10:27             ` Raslan Darawsheh [this message]
2021-07-13 10:50               ` [dpdk-dev] [PATCH v6 0/2] support VXLAN header the last 8-bits matching Rongwei Liu
2021-07-13 10:50                 ` [dpdk-dev] [PATCH v6 1/2] net/mlx5: support matching on the reserved field of VXLAN Rongwei Liu
2021-07-13 11:40                   ` Raslan Darawsheh
2021-07-13 11:49                     ` Rongwei Liu
2021-07-13 12:09                       ` [dpdk-dev] [PATCH v7 0/2] support VXLAN header the last 8-bits matching Rongwei Liu
2021-07-13 12:09                         ` [dpdk-dev] [PATCH v7 1/2] net/mlx5: support matching on the reserved field of VXLAN Rongwei Liu
2021-07-13 12:55                           ` Raslan Darawsheh
2021-07-13 13:44                             ` Rongwei Liu
2021-07-13 12:09                         ` [dpdk-dev] [PATCH v7 2/2] app/testpmd: support matching the reserved filed for VXLAN Rongwei Liu
2021-07-13 12:54                           ` Raslan Darawsheh
2021-07-13 15:34                             ` Raslan Darawsheh
2021-07-13 15:36                               ` Rongwei Liu
2021-07-13 13:09                         ` [dpdk-dev] [PATCH v7 0/2] support VXLAN header the last 8-bits matching Andrew Rybchenko
2021-07-13 12:11                     ` [dpdk-dev] [PATCH v6 1/2] net/mlx5: support matching on the reserved field of VXLAN Rongwei Liu
2021-07-13 10:50                 ` [dpdk-dev] [PATCH v6 2/2] app/testpmd: support VXLAN header last 8-bits matching Rongwei Liu
2021-07-13 11:37                   ` Raslan Darawsheh
2021-07-13 11:39                     ` Rongwei Liu
2021-07-13 10:52               ` [dpdk-dev] [PATCH v5 1/2] net/mlx5: add VXLAN header the last 8-bits matching support Rongwei Liu
2021-07-13  9:55           ` [dpdk-dev] [PATCH v5 2/2] app/testpmd: support VXLAN the last 8-bits field matching Rongwei Liu
2021-07-13 10:02             ` Raslan Darawsheh
2021-07-13 10:06               ` Andrew Rybchenko
2021-07-13  9:56         ` [dpdk-dev] [PATCH v4 0/2] support VXLAN header the last 8-bits matching Rongwei Liu
2021-07-05  9:50 ` [dpdk-dev] [PATCH v2 2/2] app/testpmd: support VXLAN last 8-bits field matching rongwei liu
2021-07-06 12:28   ` Thomas Monjalon

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=DM4PR12MB5054F9281434978B648F6C70CF149@DM4PR12MB5054.namprd12.prod.outlook.com \
    --to=rasland@nvidia.com \
    --cc=dev@dpdk.org \
    --cc=matan@nvidia.com \
    --cc=orika@nvidia.com \
    --cc=rongweil@nvidia.com \
    --cc=shahafs@nvidia.com \
    --cc=thomas@monjalon.net \
    --cc=viacheslavo@nvidia.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).