net/bnxt: refactor filter/flow
authorAjit Khaparde <ajit.khaparde@broadcom.com>
Thu, 28 Jun 2018 20:15:37 +0000 (13:15 -0700)
committerFerruh Yigit <ferruh.yigit@intel.com>
Mon, 2 Jul 2018 23:35:58 +0000 (01:35 +0200)
In preparation of more rte_flow support it has been decided to
separate out filter and flow into their own files. Functionally the
same.

Signed-off-by: Michael Wildt <michael.wildt@broadcom.com>
Signed-off-by: Scott Branden <scott.branden@broadcom.com>
Signed-off-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
drivers/net/bnxt/Makefile
drivers/net/bnxt/bnxt_filter.c
drivers/net/bnxt/bnxt_flow.c [new file with mode: 0644]
drivers/net/bnxt/meson.build

index 80db03e..8be3cb0 100644 (file)
@@ -29,6 +29,7 @@ EXPORT_MAP := rte_pmd_bnxt_version.map
 SRCS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += bnxt_cpr.c
 SRCS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += bnxt_ethdev.c
 SRCS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += bnxt_filter.c
+SRCS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += bnxt_flow.c
 SRCS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += bnxt_hwrm.c
 SRCS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += bnxt_ring.c
 SRCS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += bnxt_rxq.c
index c83b910..31757d3 100644 (file)
@@ -15,7 +15,6 @@
 #include "bnxt.h"
 #include "bnxt_filter.h"
 #include "bnxt_hwrm.h"
-#include "bnxt_util.h"
 #include "bnxt_vnic.h"
 #include "hsi_struct_def_dpdk.h"
 
@@ -181,1063 +180,3 @@ void bnxt_free_filter(struct bnxt *bp, struct bnxt_filter_info *filter)
 {
        STAILQ_INSERT_TAIL(&bp->free_filter_list, filter, next);
 }
-
-static int
-bnxt_flow_agrs_validate(const struct rte_flow_attr *attr,
-                       const struct rte_flow_item pattern[],
-                       const struct rte_flow_action actions[],
-                       struct rte_flow_error *error)
-{
-       if (!pattern) {
-               rte_flow_error_set(error, EINVAL,
-                       RTE_FLOW_ERROR_TYPE_ITEM_NUM,
-                       NULL, "NULL pattern.");
-               return -rte_errno;
-       }
-
-       if (!actions) {
-               rte_flow_error_set(error, EINVAL,
-                                  RTE_FLOW_ERROR_TYPE_ACTION_NUM,
-                                  NULL, "NULL action.");
-               return -rte_errno;
-       }
-
-       if (!attr) {
-               rte_flow_error_set(error, EINVAL,
-                                  RTE_FLOW_ERROR_TYPE_ATTR,
-                                  NULL, "NULL attribute.");
-               return -rte_errno;
-       }
-
-       return 0;
-}
-
-static const struct rte_flow_item *
-nxt_non_void_pattern(const struct rte_flow_item *cur)
-{
-       while (1) {
-               if (cur->type != RTE_FLOW_ITEM_TYPE_VOID)
-                       return cur;
-               cur++;
-       }
-}
-
-static const struct rte_flow_action *
-nxt_non_void_action(const struct rte_flow_action *cur)
-{
-       while (1) {
-               if (cur->type != RTE_FLOW_ACTION_TYPE_VOID)
-                       return cur;
-               cur++;
-       }
-}
-
-static int
-bnxt_filter_type_check(const struct rte_flow_item pattern[],
-                      struct rte_flow_error *error __rte_unused)
-{
-       const struct rte_flow_item *item = nxt_non_void_pattern(pattern);
-       int use_ntuple = 1;
-
-       while (item->type != RTE_FLOW_ITEM_TYPE_END) {
-               switch (item->type) {
-               case RTE_FLOW_ITEM_TYPE_ETH:
-                       use_ntuple = 1;
-                       break;
-               case RTE_FLOW_ITEM_TYPE_VLAN:
-                       use_ntuple = 0;
-                       break;
-               case RTE_FLOW_ITEM_TYPE_IPV4:
-               case RTE_FLOW_ITEM_TYPE_IPV6:
-               case RTE_FLOW_ITEM_TYPE_TCP:
-               case RTE_FLOW_ITEM_TYPE_UDP:
-                       /* FALLTHROUGH */
-                       /* need ntuple match, reset exact match */
-                       if (!use_ntuple) {
-                               PMD_DRV_LOG(ERR,
-                                       "VLAN flow cannot use NTUPLE filter\n");
-                               rte_flow_error_set(error, EINVAL,
-                                                  RTE_FLOW_ERROR_TYPE_ITEM,
-                                                  item,
-                                                  "Cannot use VLAN with NTUPLE");
-                               return -rte_errno;
-                       }
-                       use_ntuple |= 1;
-                       break;
-               default:
-                       PMD_DRV_LOG(ERR, "Unknown Flow type");
-                       use_ntuple |= 1;
-               }
-               item++;
-       }
-       return use_ntuple;
-}
-
-static int
-bnxt_validate_and_parse_flow_type(struct bnxt *bp,
-                                 const struct rte_flow_attr *attr,
-                                 const struct rte_flow_item pattern[],
-                                 struct rte_flow_error *error,
-                                 struct bnxt_filter_info *filter)
-{
-       const struct rte_flow_item *item = nxt_non_void_pattern(pattern);
-       const struct rte_flow_item_vlan *vlan_spec, *vlan_mask;
-       const struct rte_flow_item_ipv4 *ipv4_spec, *ipv4_mask;
-       const struct rte_flow_item_ipv6 *ipv6_spec, *ipv6_mask;
-       const struct rte_flow_item_tcp *tcp_spec, *tcp_mask;
-       const struct rte_flow_item_udp *udp_spec, *udp_mask;
-       const struct rte_flow_item_eth *eth_spec, *eth_mask;
-       const struct rte_flow_item_nvgre *nvgre_spec;
-       const struct rte_flow_item_nvgre *nvgre_mask;
-       const struct rte_flow_item_vxlan *vxlan_spec;
-       const struct rte_flow_item_vxlan *vxlan_mask;
-       uint8_t vni_mask[] = {0xFF, 0xFF, 0xFF};
-       uint8_t tni_mask[] = {0xFF, 0xFF, 0xFF};
-       const struct rte_flow_item_vf *vf_spec;
-       uint32_t tenant_id_be = 0;
-       bool vni_masked = 0;
-       bool tni_masked = 0;
-       uint32_t vf = 0;
-       int use_ntuple;
-       uint32_t en = 0;
-       uint32_t en_ethertype;
-       int dflt_vnic;
-
-       use_ntuple = bnxt_filter_type_check(pattern, error);
-       PMD_DRV_LOG(DEBUG, "Use NTUPLE %d\n", use_ntuple);
-       if (use_ntuple < 0)
-               return use_ntuple;
-
-       filter->filter_type = use_ntuple ?
-               HWRM_CFA_NTUPLE_FILTER : HWRM_CFA_EM_FILTER;
-       en_ethertype = use_ntuple ?
-               NTUPLE_FLTR_ALLOC_INPUT_EN_ETHERTYPE :
-               EM_FLOW_ALLOC_INPUT_EN_ETHERTYPE;
-
-       while (item->type != RTE_FLOW_ITEM_TYPE_END) {
-               if (item->last) {
-                       /* last or range is NOT supported as match criteria */
-                       rte_flow_error_set(error, EINVAL,
-                                          RTE_FLOW_ERROR_TYPE_ITEM,
-                                          item,
-                                          "No support for range");
-                       return -rte_errno;
-               }
-               if (!item->spec || !item->mask) {
-                       rte_flow_error_set(error, EINVAL,
-                                          RTE_FLOW_ERROR_TYPE_ITEM,
-                                          item,
-                                          "spec/mask is NULL");
-                       return -rte_errno;
-               }
-               switch (item->type) {
-               case RTE_FLOW_ITEM_TYPE_ETH:
-                       eth_spec = item->spec;
-                       eth_mask = item->mask;
-
-                       /* Source MAC address mask cannot be partially set.
-                        * Should be All 0's or all 1's.
-                        * Destination MAC address mask must not be partially
-                        * set. Should be all 1's or all 0's.
-                        */
-                       if ((!is_zero_ether_addr(&eth_mask->src) &&
-                            !is_broadcast_ether_addr(&eth_mask->src)) ||
-                           (!is_zero_ether_addr(&eth_mask->dst) &&
-                            !is_broadcast_ether_addr(&eth_mask->dst))) {
-                               rte_flow_error_set(error, EINVAL,
-                                                  RTE_FLOW_ERROR_TYPE_ITEM,
-                                                  item,
-                                                  "MAC_addr mask not valid");
-                               return -rte_errno;
-                       }
-
-                       /* Mask is not allowed. Only exact matches are */
-                       if (eth_mask->type &&
-                           eth_mask->type != RTE_BE16(0xffff)) {
-                               rte_flow_error_set(error, EINVAL,
-                                                  RTE_FLOW_ERROR_TYPE_ITEM,
-                                                  item,
-                                                  "ethertype mask not valid");
-                               return -rte_errno;
-                       }
-
-                       if (is_broadcast_ether_addr(&eth_mask->dst)) {
-                               rte_memcpy(filter->dst_macaddr,
-                                          &eth_spec->dst, 6);
-                               en |= use_ntuple ?
-                                       NTUPLE_FLTR_ALLOC_INPUT_EN_DST_MACADDR :
-                                       EM_FLOW_ALLOC_INPUT_EN_DST_MACADDR;
-                       }
-                       if (is_broadcast_ether_addr(&eth_mask->src)) {
-                               rte_memcpy(filter->src_macaddr,
-                                          &eth_spec->src, 6);
-                               en |= use_ntuple ?
-                                       NTUPLE_FLTR_ALLOC_INPUT_EN_SRC_MACADDR :
-                                       EM_FLOW_ALLOC_INPUT_EN_SRC_MACADDR;
-                       } /*
-                          * else {
-                          *  RTE_LOG(ERR, PMD, "Handle this condition\n");
-                          * }
-                          */
-                       if (eth_mask->type) {
-                               filter->ethertype =
-                                       rte_be_to_cpu_16(eth_spec->type);
-                               en |= en_ethertype;
-                       }
-
-                       break;
-               case RTE_FLOW_ITEM_TYPE_VLAN:
-                       vlan_spec = item->spec;
-                       vlan_mask = item->mask;
-                       if (en & en_ethertype) {
-                               rte_flow_error_set(error, EINVAL,
-                                                  RTE_FLOW_ERROR_TYPE_ITEM,
-                                                  item,
-                                                  "VLAN TPID matching is not"
-                                                  " supported");
-                               return -rte_errno;
-                       }
-                       if (vlan_mask->tci &&
-                           vlan_mask->tci == RTE_BE16(0x0fff)) {
-                               /* Only the VLAN ID can be matched. */
-                               filter->l2_ovlan =
-                                       rte_be_to_cpu_16(vlan_spec->tci &
-                                                        RTE_BE16(0x0fff));
-                               en |= EM_FLOW_ALLOC_INPUT_EN_OVLAN_VID;
-                       } else if (vlan_mask->tci) {
-                               rte_flow_error_set(error, EINVAL,
-                                                  RTE_FLOW_ERROR_TYPE_ITEM,
-                                                  item,
-                                                  "VLAN mask is invalid");
-                               return -rte_errno;
-                       }
-                       if (vlan_mask->inner_type &&
-                           vlan_mask->inner_type != RTE_BE16(0xffff)) {
-                               rte_flow_error_set(error, EINVAL,
-                                                  RTE_FLOW_ERROR_TYPE_ITEM,
-                                                  item,
-                                                  "inner ethertype mask not"
-                                                  " valid");
-                               return -rte_errno;
-                       }
-                       if (vlan_mask->inner_type) {
-                               filter->ethertype =
-                                       rte_be_to_cpu_16(vlan_spec->inner_type);
-                               en |= en_ethertype;
-                       }
-
-                       break;
-               case RTE_FLOW_ITEM_TYPE_IPV4:
-                       /* If mask is not involved, we could use EM filters. */
-                       ipv4_spec = item->spec;
-                       ipv4_mask = item->mask;
-                       /* Only IP DST and SRC fields are maskable. */
-                       if (ipv4_mask->hdr.version_ihl ||
-                           ipv4_mask->hdr.type_of_service ||
-                           ipv4_mask->hdr.total_length ||
-                           ipv4_mask->hdr.packet_id ||
-                           ipv4_mask->hdr.fragment_offset ||
-                           ipv4_mask->hdr.time_to_live ||
-                           ipv4_mask->hdr.next_proto_id ||
-                           ipv4_mask->hdr.hdr_checksum) {
-                               rte_flow_error_set(error, EINVAL,
-                                          RTE_FLOW_ERROR_TYPE_ITEM,
-                                          item,
-                                          "Invalid IPv4 mask.");
-                               return -rte_errno;
-                       }
-                       filter->dst_ipaddr[0] = ipv4_spec->hdr.dst_addr;
-                       filter->src_ipaddr[0] = ipv4_spec->hdr.src_addr;
-                       if (use_ntuple)
-                               en |= NTUPLE_FLTR_ALLOC_INPUT_EN_SRC_IPADDR |
-                                       NTUPLE_FLTR_ALLOC_INPUT_EN_DST_IPADDR;
-                       else
-                               en |= EM_FLOW_ALLOC_INPUT_EN_SRC_IPADDR |
-                                       EM_FLOW_ALLOC_INPUT_EN_DST_IPADDR;
-                       if (ipv4_mask->hdr.src_addr) {
-                               filter->src_ipaddr_mask[0] =
-                                       ipv4_mask->hdr.src_addr;
-                               en |= !use_ntuple ? 0 :
-                                    NTUPLE_FLTR_ALLOC_INPUT_EN_SRC_IPADDR_MASK;
-                       }
-                       if (ipv4_mask->hdr.dst_addr) {
-                               filter->dst_ipaddr_mask[0] =
-                                       ipv4_mask->hdr.dst_addr;
-                               en |= !use_ntuple ? 0 :
-                                    NTUPLE_FLTR_ALLOC_INPUT_EN_DST_IPADDR_MASK;
-                       }
-                       filter->ip_addr_type = use_ntuple ?
-                        HWRM_CFA_NTUPLE_FILTER_ALLOC_INPUT_IP_ADDR_TYPE_IPV4 :
-                        HWRM_CFA_EM_FLOW_ALLOC_INPUT_IP_ADDR_TYPE_IPV4;
-                       if (ipv4_spec->hdr.next_proto_id) {
-                               filter->ip_protocol =
-                                       ipv4_spec->hdr.next_proto_id;
-                               if (use_ntuple)
-                                       en |= NTUPLE_FLTR_ALLOC_IN_EN_IP_PROTO;
-                               else
-                                       en |= EM_FLOW_ALLOC_INPUT_EN_IP_PROTO;
-                       }
-                       break;
-               case RTE_FLOW_ITEM_TYPE_IPV6:
-                       ipv6_spec = item->spec;
-                       ipv6_mask = item->mask;
-
-                       /* Only IP DST and SRC fields are maskable. */
-                       if (ipv6_mask->hdr.vtc_flow ||
-                           ipv6_mask->hdr.payload_len ||
-                           ipv6_mask->hdr.proto ||
-                           ipv6_mask->hdr.hop_limits) {
-                               rte_flow_error_set(error, EINVAL,
-                                          RTE_FLOW_ERROR_TYPE_ITEM,
-                                          item,
-                                          "Invalid IPv6 mask.");
-                               return -rte_errno;
-                       }
-
-                       if (use_ntuple)
-                               en |= NTUPLE_FLTR_ALLOC_INPUT_EN_SRC_IPADDR |
-                                       NTUPLE_FLTR_ALLOC_INPUT_EN_DST_IPADDR;
-                       else
-                               en |= EM_FLOW_ALLOC_INPUT_EN_SRC_IPADDR |
-                                       EM_FLOW_ALLOC_INPUT_EN_DST_IPADDR;
-                       rte_memcpy(filter->src_ipaddr,
-                                  ipv6_spec->hdr.src_addr, 16);
-                       rte_memcpy(filter->dst_ipaddr,
-                                  ipv6_spec->hdr.dst_addr, 16);
-                       if (!bnxt_check_zero_bytes(ipv6_mask->hdr.src_addr,
-                                                  16)) {
-                               rte_memcpy(filter->src_ipaddr_mask,
-                                          ipv6_mask->hdr.src_addr, 16);
-                               en |= !use_ntuple ? 0 :
-                                   NTUPLE_FLTR_ALLOC_INPUT_EN_SRC_IPADDR_MASK;
-                       }
-                       if (!bnxt_check_zero_bytes(ipv6_mask->hdr.dst_addr,
-                                                  16)) {
-                               rte_memcpy(filter->dst_ipaddr_mask,
-                                          ipv6_mask->hdr.dst_addr, 16);
-                               en |= !use_ntuple ? 0 :
-                                    NTUPLE_FLTR_ALLOC_INPUT_EN_DST_IPADDR_MASK;
-                       }
-                       filter->ip_addr_type = use_ntuple ?
-                               NTUPLE_FLTR_ALLOC_INPUT_IP_ADDR_TYPE_IPV6 :
-                               EM_FLOW_ALLOC_INPUT_IP_ADDR_TYPE_IPV6;
-                       break;
-               case RTE_FLOW_ITEM_TYPE_TCP:
-                       tcp_spec = item->spec;
-                       tcp_mask = item->mask;
-
-                       /* Check TCP mask. Only DST & SRC ports are maskable */
-                       if (tcp_mask->hdr.sent_seq ||
-                           tcp_mask->hdr.recv_ack ||
-                           tcp_mask->hdr.data_off ||
-                           tcp_mask->hdr.tcp_flags ||
-                           tcp_mask->hdr.rx_win ||
-                           tcp_mask->hdr.cksum ||
-                           tcp_mask->hdr.tcp_urp) {
-                               rte_flow_error_set(error, EINVAL,
-                                          RTE_FLOW_ERROR_TYPE_ITEM,
-                                          item,
-                                          "Invalid TCP mask");
-                               return -rte_errno;
-                       }
-                       filter->src_port = tcp_spec->hdr.src_port;
-                       filter->dst_port = tcp_spec->hdr.dst_port;
-                       if (use_ntuple)
-                               en |= NTUPLE_FLTR_ALLOC_INPUT_EN_SRC_PORT |
-                                       NTUPLE_FLTR_ALLOC_INPUT_EN_DST_PORT;
-                       else
-                               en |= EM_FLOW_ALLOC_INPUT_EN_SRC_PORT |
-                                       EM_FLOW_ALLOC_INPUT_EN_DST_PORT;
-                       if (tcp_mask->hdr.dst_port) {
-                               filter->dst_port_mask = tcp_mask->hdr.dst_port;
-                               en |= !use_ntuple ? 0 :
-                                 NTUPLE_FLTR_ALLOC_INPUT_EN_DST_PORT_MASK;
-                       }
-                       if (tcp_mask->hdr.src_port) {
-                               filter->src_port_mask = tcp_mask->hdr.src_port;
-                               en |= !use_ntuple ? 0 :
-                                 NTUPLE_FLTR_ALLOC_INPUT_EN_SRC_PORT_MASK;
-                       }
-                       break;
-               case RTE_FLOW_ITEM_TYPE_UDP:
-                       udp_spec = item->spec;
-                       udp_mask = item->mask;
-
-                       if (udp_mask->hdr.dgram_len ||
-                           udp_mask->hdr.dgram_cksum) {
-                               rte_flow_error_set(error, EINVAL,
-                                          RTE_FLOW_ERROR_TYPE_ITEM,
-                                          item,
-                                          "Invalid UDP mask");
-                               return -rte_errno;
-                       }
-
-                       filter->src_port = udp_spec->hdr.src_port;
-                       filter->dst_port = udp_spec->hdr.dst_port;
-                       if (use_ntuple)
-                               en |= NTUPLE_FLTR_ALLOC_INPUT_EN_SRC_PORT |
-                                       NTUPLE_FLTR_ALLOC_INPUT_EN_DST_PORT;
-                       else
-                               en |= EM_FLOW_ALLOC_INPUT_EN_SRC_PORT |
-                                       EM_FLOW_ALLOC_INPUT_EN_DST_PORT;
-
-                       if (udp_mask->hdr.dst_port) {
-                               filter->dst_port_mask = udp_mask->hdr.dst_port;
-                               en |= !use_ntuple ? 0 :
-                                 NTUPLE_FLTR_ALLOC_INPUT_EN_DST_PORT_MASK;
-                       }
-                       if (udp_mask->hdr.src_port) {
-                               filter->src_port_mask = udp_mask->hdr.src_port;
-                               en |= !use_ntuple ? 0 :
-                                 NTUPLE_FLTR_ALLOC_INPUT_EN_SRC_PORT_MASK;
-                       }
-                       break;
-               case RTE_FLOW_ITEM_TYPE_VXLAN:
-                       vxlan_spec = item->spec;
-                       vxlan_mask = item->mask;
-                       /* Check if VXLAN item is used to describe protocol.
-                        * If yes, both spec and mask should be NULL.
-                        * If no, both spec and mask shouldn't be NULL.
-                        */
-                       if ((!vxlan_spec && vxlan_mask) ||
-                           (vxlan_spec && !vxlan_mask)) {
-                               rte_flow_error_set(error, EINVAL,
-                                          RTE_FLOW_ERROR_TYPE_ITEM,
-                                          item,
-                                          "Invalid VXLAN item");
-                               return -rte_errno;
-                       }
-
-                       if (vxlan_spec->rsvd1 || vxlan_spec->rsvd0[0] ||
-                           vxlan_spec->rsvd0[1] || vxlan_spec->rsvd0[2] ||
-                           vxlan_spec->flags != 0x8) {
-                               rte_flow_error_set(error, EINVAL,
-                                          RTE_FLOW_ERROR_TYPE_ITEM,
-                                          item,
-                                          "Invalid VXLAN item");
-                               return -rte_errno;
-                       }
-
-                       /* Check if VNI is masked. */
-                       if (vxlan_spec && vxlan_mask) {
-                               vni_masked =
-                                       !!memcmp(vxlan_mask->vni, vni_mask,
-                                                RTE_DIM(vni_mask));
-                               if (vni_masked) {
-                                       rte_flow_error_set(error, EINVAL,
-                                                  RTE_FLOW_ERROR_TYPE_ITEM,
-                                                  item,
-                                                  "Invalid VNI mask");
-                                       return -rte_errno;
-                               }
-
-                               rte_memcpy(((uint8_t *)&tenant_id_be + 1),
-                                          vxlan_spec->vni, 3);
-                               filter->vni =
-                                       rte_be_to_cpu_32(tenant_id_be);
-                               filter->tunnel_type =
-                                CFA_NTUPLE_FILTER_ALLOC_REQ_TUNNEL_TYPE_VXLAN;
-                       }
-                       break;
-               case RTE_FLOW_ITEM_TYPE_NVGRE:
-                       nvgre_spec = item->spec;
-                       nvgre_mask = item->mask;
-                       /* Check if NVGRE item is used to describe protocol.
-                        * If yes, both spec and mask should be NULL.
-                        * If no, both spec and mask shouldn't be NULL.
-                        */
-                       if ((!nvgre_spec && nvgre_mask) ||
-                           (nvgre_spec && !nvgre_mask)) {
-                               rte_flow_error_set(error, EINVAL,
-                                          RTE_FLOW_ERROR_TYPE_ITEM,
-                                          item,
-                                          "Invalid NVGRE item");
-                               return -rte_errno;
-                       }
-
-                       if (nvgre_spec->c_k_s_rsvd0_ver != 0x2000 ||
-                           nvgre_spec->protocol != 0x6558) {
-                               rte_flow_error_set(error, EINVAL,
-                                          RTE_FLOW_ERROR_TYPE_ITEM,
-                                          item,
-                                          "Invalid NVGRE item");
-                               return -rte_errno;
-                       }
-
-                       if (nvgre_spec && nvgre_mask) {
-                               tni_masked =
-                                       !!memcmp(nvgre_mask->tni, tni_mask,
-                                                RTE_DIM(tni_mask));
-                               if (tni_masked) {
-                                       rte_flow_error_set(error, EINVAL,
-                                                      RTE_FLOW_ERROR_TYPE_ITEM,
-                                                      item,
-                                                      "Invalid TNI mask");
-                                       return -rte_errno;
-                               }
-                               rte_memcpy(((uint8_t *)&tenant_id_be + 1),
-                                          nvgre_spec->tni, 3);
-                               filter->vni =
-                                       rte_be_to_cpu_32(tenant_id_be);
-                               filter->tunnel_type =
-                                CFA_NTUPLE_FILTER_ALLOC_REQ_TUNNEL_TYPE_NVGRE;
-                       }
-                       break;
-               case RTE_FLOW_ITEM_TYPE_VF:
-                       vf_spec = item->spec;
-                       vf = vf_spec->id;
-                       if (!BNXT_PF(bp)) {
-                               rte_flow_error_set(error, EINVAL,
-                                          RTE_FLOW_ERROR_TYPE_ITEM,
-                                          item,
-                                          "Configuring on a VF!");
-                               return -rte_errno;
-                       }
-
-                       if (vf >= bp->pdev->max_vfs) {
-                               rte_flow_error_set(error, EINVAL,
-                                          RTE_FLOW_ERROR_TYPE_ITEM,
-                                          item,
-                                          "Incorrect VF id!");
-                               return -rte_errno;
-                       }
-
-                       if (!attr->transfer) {
-                               rte_flow_error_set(error, ENOTSUP,
-                                          RTE_FLOW_ERROR_TYPE_ITEM,
-                                          item,
-                                          "Matching VF traffic without"
-                                          " affecting it (transfer attribute)"
-                                          " is unsupported");
-                               return -rte_errno;
-                       }
-
-                       filter->mirror_vnic_id =
-                       dflt_vnic = bnxt_hwrm_func_qcfg_vf_dflt_vnic_id(bp, vf);
-                       if (dflt_vnic < 0) {
-                               /* This simply indicates there's no driver
-                                * loaded. This is not an error.
-                                */
-                               rte_flow_error_set(error, EINVAL,
-                                          RTE_FLOW_ERROR_TYPE_ITEM,
-                                          item,
-                                          "Unable to get default VNIC for VF");
-                               return -rte_errno;
-                       }
-                       filter->mirror_vnic_id = dflt_vnic;
-                       en |= NTUPLE_FLTR_ALLOC_INPUT_EN_MIRROR_VNIC_ID;
-                       break;
-               default:
-                       break;
-               }
-               item++;
-       }
-       filter->enables = en;
-
-       return 0;
-}
-
-/* Parse attributes */
-static int
-bnxt_flow_parse_attr(const struct rte_flow_attr *attr,
-                    struct rte_flow_error *error)
-{
-       /* Must be input direction */
-       if (!attr->ingress) {
-               rte_flow_error_set(error, EINVAL,
-                                  RTE_FLOW_ERROR_TYPE_ATTR_INGRESS,
-                                  attr, "Only support ingress.");
-               return -rte_errno;
-       }
-
-       /* Not supported */
-       if (attr->egress) {
-               rte_flow_error_set(error, EINVAL,
-                                  RTE_FLOW_ERROR_TYPE_ATTR_EGRESS,
-                                  attr, "No support for egress.");
-               return -rte_errno;
-       }
-
-       /* Not supported */
-       if (attr->priority) {
-               rte_flow_error_set(error, EINVAL,
-                                  RTE_FLOW_ERROR_TYPE_ATTR_PRIORITY,
-                                  attr, "No support for priority.");
-               return -rte_errno;
-       }
-
-       /* Not supported */
-       if (attr->group) {
-               rte_flow_error_set(error, EINVAL,
-                                  RTE_FLOW_ERROR_TYPE_ATTR_GROUP,
-                                  attr, "No support for group.");
-               return -rte_errno;
-       }
-
-       return 0;
-}
-
-struct bnxt_filter_info *
-bnxt_get_l2_filter(struct bnxt *bp, struct bnxt_filter_info *nf,
-                  struct bnxt_vnic_info *vnic)
-{
-       struct bnxt_filter_info *filter1, *f0;
-       struct bnxt_vnic_info *vnic0;
-       int rc;
-
-       vnic0 = STAILQ_FIRST(&bp->ff_pool[0]);
-       f0 = STAILQ_FIRST(&vnic0->filter);
-
-       //This flow has same DST MAC as the port/l2 filter.
-       if (memcmp(f0->l2_addr, nf->dst_macaddr, ETHER_ADDR_LEN) == 0)
-               return f0;
-
-       //This flow needs DST MAC which is not same as port/l2
-       PMD_DRV_LOG(DEBUG, "Create L2 filter for DST MAC\n");
-       filter1 = bnxt_get_unused_filter(bp);
-       if (filter1 == NULL)
-               return NULL;
-       filter1->flags = HWRM_CFA_L2_FILTER_ALLOC_INPUT_FLAGS_PATH_RX;
-       filter1->enables = HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_L2_ADDR |
-                       L2_FILTER_ALLOC_INPUT_EN_L2_ADDR_MASK;
-       memcpy(filter1->l2_addr, nf->dst_macaddr, ETHER_ADDR_LEN);
-       memset(filter1->l2_addr_mask, 0xff, ETHER_ADDR_LEN);
-       rc = bnxt_hwrm_set_l2_filter(bp, vnic->fw_vnic_id,
-                                    filter1);
-       if (rc) {
-               bnxt_free_filter(bp, filter1);
-               return NULL;
-       }
-       return filter1;
-}
-
-static int
-bnxt_validate_and_parse_flow(struct rte_eth_dev *dev,
-                            const struct rte_flow_item pattern[],
-                            const struct rte_flow_action actions[],
-                            const struct rte_flow_attr *attr,
-                            struct rte_flow_error *error,
-                            struct bnxt_filter_info *filter)
-{
-       const struct rte_flow_action *act = nxt_non_void_action(actions);
-       struct bnxt *bp = (struct bnxt *)dev->data->dev_private;
-       const struct rte_flow_action_queue *act_q;
-       const struct rte_flow_action_vf *act_vf;
-       struct bnxt_vnic_info *vnic, *vnic0;
-       struct bnxt_filter_info *filter1;
-       uint32_t vf = 0;
-       int dflt_vnic;
-       int rc;
-
-       if (bp->eth_dev->data->dev_conf.rxmode.mq_mode & ETH_MQ_RX_RSS) {
-               PMD_DRV_LOG(ERR, "Cannot create flow on RSS queues\n");
-               rte_flow_error_set(error, EINVAL,
-                                  RTE_FLOW_ERROR_TYPE_UNSPECIFIED, NULL,
-                                  "Cannot create flow on RSS queues");
-               rc = -rte_errno;
-               goto ret;
-       }
-
-       rc = bnxt_validate_and_parse_flow_type(bp, attr, pattern, error,
-                                              filter);
-       if (rc != 0)
-               goto ret;
-
-       rc = bnxt_flow_parse_attr(attr, error);
-       if (rc != 0)
-               goto ret;
-       //Since we support ingress attribute only - right now.
-       if (filter->filter_type == HWRM_CFA_EM_FILTER)
-               filter->flags = HWRM_CFA_EM_FLOW_ALLOC_INPUT_FLAGS_PATH_RX;
-
-       switch (act->type) {
-       case RTE_FLOW_ACTION_TYPE_QUEUE:
-               /* Allow this flow. Redirect to a VNIC. */
-               act_q = (const struct rte_flow_action_queue *)act->conf;
-               if (act_q->index >= bp->rx_nr_rings) {
-                       rte_flow_error_set(error, EINVAL,
-                                          RTE_FLOW_ERROR_TYPE_ACTION, act,
-                                          "Invalid queue ID.");
-                       rc = -rte_errno;
-                       goto ret;
-               }
-               PMD_DRV_LOG(DEBUG, "Queue index %d\n", act_q->index);
-
-               vnic0 = STAILQ_FIRST(&bp->ff_pool[0]);
-               vnic = STAILQ_FIRST(&bp->ff_pool[act_q->index]);
-               if (vnic == NULL) {
-                       rte_flow_error_set(error, EINVAL,
-                                          RTE_FLOW_ERROR_TYPE_ACTION, act,
-                                          "No matching VNIC for queue ID.");
-                       rc = -rte_errno;
-                       goto ret;
-               }
-               filter->dst_id = vnic->fw_vnic_id;
-               filter1 = bnxt_get_l2_filter(bp, filter, vnic);
-               if (filter1 == NULL) {
-                       rc = -ENOSPC;
-                       goto ret;
-               }
-               filter->fw_l2_filter_id = filter1->fw_l2_filter_id;
-               PMD_DRV_LOG(DEBUG, "VNIC found\n");
-               break;
-       case RTE_FLOW_ACTION_TYPE_DROP:
-               vnic0 = STAILQ_FIRST(&bp->ff_pool[0]);
-               filter1 = bnxt_get_l2_filter(bp, filter, vnic0);
-               if (filter1 == NULL) {
-                       rc = -ENOSPC;
-                       goto ret;
-               }
-               filter->fw_l2_filter_id = filter1->fw_l2_filter_id;
-               if (filter->filter_type == HWRM_CFA_EM_FILTER)
-                       filter->flags =
-                               HWRM_CFA_EM_FLOW_ALLOC_INPUT_FLAGS_DROP;
-               else
-                       filter->flags =
-                               HWRM_CFA_NTUPLE_FILTER_ALLOC_INPUT_FLAGS_DROP;
-               break;
-       case RTE_FLOW_ACTION_TYPE_COUNT:
-               vnic0 = STAILQ_FIRST(&bp->ff_pool[0]);
-               filter1 = bnxt_get_l2_filter(bp, filter, vnic0);
-               if (filter1 == NULL) {
-                       rc = -ENOSPC;
-                       goto ret;
-               }
-               filter->fw_l2_filter_id = filter1->fw_l2_filter_id;
-               filter->flags = HWRM_CFA_NTUPLE_FILTER_ALLOC_INPUT_FLAGS_METER;
-               break;
-       case RTE_FLOW_ACTION_TYPE_VF:
-               act_vf = (const struct rte_flow_action_vf *)act->conf;
-               vf = act_vf->id;
-               if (!BNXT_PF(bp)) {
-                       rte_flow_error_set(error, EINVAL,
-                                  RTE_FLOW_ERROR_TYPE_ACTION,
-                                  act,
-                                  "Configuring on a VF!");
-                       rc = -rte_errno;
-                       goto ret;
-               }
-
-               if (vf >= bp->pdev->max_vfs) {
-                       rte_flow_error_set(error, EINVAL,
-                                  RTE_FLOW_ERROR_TYPE_ACTION,
-                                  act,
-                                  "Incorrect VF id!");
-                       rc = -rte_errno;
-                       goto ret;
-               }
-
-               filter->mirror_vnic_id =
-               dflt_vnic = bnxt_hwrm_func_qcfg_vf_dflt_vnic_id(bp, vf);
-               if (dflt_vnic < 0) {
-                       /* This simply indicates there's no driver loaded.
-                        * This is not an error.
-                        */
-                       rte_flow_error_set(error, EINVAL,
-                                  RTE_FLOW_ERROR_TYPE_ACTION,
-                                  act,
-                                  "Unable to get default VNIC for VF");
-                       rc = -rte_errno;
-                       goto ret;
-               }
-               filter->mirror_vnic_id = dflt_vnic;
-               filter->enables |= NTUPLE_FLTR_ALLOC_INPUT_EN_MIRROR_VNIC_ID;
-
-               vnic0 = STAILQ_FIRST(&bp->ff_pool[0]);
-               filter1 = bnxt_get_l2_filter(bp, filter, vnic0);
-               if (filter1 == NULL) {
-                       rc = -ENOSPC;
-                       goto ret;
-               }
-               filter->fw_l2_filter_id = filter1->fw_l2_filter_id;
-               break;
-
-       default:
-               rte_flow_error_set(error, EINVAL,
-                                  RTE_FLOW_ERROR_TYPE_ACTION, act,
-                                  "Invalid action.");
-               rc = -rte_errno;
-               goto ret;
-       }
-
-       act = nxt_non_void_action(++act);
-       if (act->type != RTE_FLOW_ACTION_TYPE_END) {
-               rte_flow_error_set(error, EINVAL,
-                                  RTE_FLOW_ERROR_TYPE_ACTION,
-                                  act, "Invalid action.");
-               rc = -rte_errno;
-               goto ret;
-       }
-ret:
-       return rc;
-}
-
-static int
-bnxt_flow_validate(struct rte_eth_dev *dev,
-               const struct rte_flow_attr *attr,
-               const struct rte_flow_item pattern[],
-               const struct rte_flow_action actions[],
-               struct rte_flow_error *error)
-{
-       struct bnxt *bp = (struct bnxt *)dev->data->dev_private;
-       struct bnxt_filter_info *filter;
-       int ret = 0;
-
-       ret = bnxt_flow_agrs_validate(attr, pattern, actions, error);
-       if (ret != 0)
-               return ret;
-
-       filter = bnxt_get_unused_filter(bp);
-       if (filter == NULL) {
-               PMD_DRV_LOG(ERR, "Not enough resources for a new flow.\n");
-               return -ENOMEM;
-       }
-
-       ret = bnxt_validate_and_parse_flow(dev, pattern, actions, attr,
-                                          error, filter);
-       /* No need to hold on to this filter if we are just validating flow */
-       filter->fw_l2_filter_id = UINT64_MAX;
-       bnxt_free_filter(bp, filter);
-
-       return ret;
-}
-
-static int
-bnxt_match_filter(struct bnxt *bp, struct bnxt_filter_info *nf)
-{
-       struct bnxt_filter_info *mf;
-       struct rte_flow *flow;
-       int i;
-
-       for (i = bp->nr_vnics - 1; i >= 0; i--) {
-               struct bnxt_vnic_info *vnic = &bp->vnic_info[i];
-
-               STAILQ_FOREACH(flow, &vnic->flow_list, next) {
-                       mf = flow->filter;
-
-                       if (mf->filter_type == nf->filter_type &&
-                           mf->flags == nf->flags &&
-                           mf->src_port == nf->src_port &&
-                           mf->src_port_mask == nf->src_port_mask &&
-                           mf->dst_port == nf->dst_port &&
-                           mf->dst_port_mask == nf->dst_port_mask &&
-                           mf->ip_protocol == nf->ip_protocol &&
-                           mf->ip_addr_type == nf->ip_addr_type &&
-                           mf->ethertype == nf->ethertype &&
-                           mf->vni == nf->vni &&
-                           mf->tunnel_type == nf->tunnel_type &&
-                           mf->l2_ovlan == nf->l2_ovlan &&
-                           mf->l2_ovlan_mask == nf->l2_ovlan_mask &&
-                           mf->l2_ivlan == nf->l2_ivlan &&
-                           mf->l2_ivlan_mask == nf->l2_ivlan_mask &&
-                           !memcmp(mf->l2_addr, nf->l2_addr, ETHER_ADDR_LEN) &&
-                           !memcmp(mf->l2_addr_mask, nf->l2_addr_mask,
-                                   ETHER_ADDR_LEN) &&
-                           !memcmp(mf->src_macaddr, nf->src_macaddr,
-                                   ETHER_ADDR_LEN) &&
-                           !memcmp(mf->dst_macaddr, nf->dst_macaddr,
-                                   ETHER_ADDR_LEN) &&
-                           !memcmp(mf->src_ipaddr, nf->src_ipaddr,
-                                   sizeof(nf->src_ipaddr)) &&
-                           !memcmp(mf->src_ipaddr_mask, nf->src_ipaddr_mask,
-                                   sizeof(nf->src_ipaddr_mask)) &&
-                           !memcmp(mf->dst_ipaddr, nf->dst_ipaddr,
-                                   sizeof(nf->dst_ipaddr)) &&
-                           !memcmp(mf->dst_ipaddr_mask, nf->dst_ipaddr_mask,
-                                   sizeof(nf->dst_ipaddr_mask))) {
-                               if (mf->dst_id == nf->dst_id)
-                                       return -EEXIST;
-                               /* Same Flow, Different queue
-                                * Clear the old ntuple filter
-                                */
-                               if (nf->filter_type == HWRM_CFA_EM_FILTER)
-                                       bnxt_hwrm_clear_em_filter(bp, mf);
-                               if (nf->filter_type == HWRM_CFA_NTUPLE_FILTER)
-                                       bnxt_hwrm_clear_ntuple_filter(bp, mf);
-                               /* Free the old filter, update flow
-                                * with new filter
-                                */
-                               bnxt_free_filter(bp, mf);
-                               flow->filter = nf;
-                               return -EXDEV;
-                       }
-               }
-       }
-       return 0;
-}
-
-static struct rte_flow *
-bnxt_flow_create(struct rte_eth_dev *dev,
-                 const struct rte_flow_attr *attr,
-                 const struct rte_flow_item pattern[],
-                 const struct rte_flow_action actions[],
-                 struct rte_flow_error *error)
-{
-       struct bnxt *bp = (struct bnxt *)dev->data->dev_private;
-       struct bnxt_filter_info *filter;
-       struct bnxt_vnic_info *vnic = NULL;
-       bool update_flow = false;
-       struct rte_flow *flow;
-       unsigned int i;
-       int ret = 0;
-
-       flow = rte_zmalloc("bnxt_flow", sizeof(struct rte_flow), 0);
-       if (!flow) {
-               rte_flow_error_set(error, ENOMEM,
-                                  RTE_FLOW_ERROR_TYPE_HANDLE, NULL,
-                                  "Failed to allocate memory");
-               return flow;
-       }
-
-       ret = bnxt_flow_agrs_validate(attr, pattern, actions, error);
-       if (ret != 0) {
-               PMD_DRV_LOG(ERR, "Not a validate flow.\n");
-               goto free_flow;
-       }
-
-       filter = bnxt_get_unused_filter(bp);
-       if (filter == NULL) {
-               PMD_DRV_LOG(ERR, "Not enough resources for a new flow.\n");
-               goto free_flow;
-       }
-
-       ret = bnxt_validate_and_parse_flow(dev, pattern, actions, attr,
-                                          error, filter);
-       if (ret != 0)
-               goto free_filter;
-
-       ret = bnxt_match_filter(bp, filter);
-       if (ret == -EEXIST) {
-               PMD_DRV_LOG(DEBUG, "Flow already exists.\n");
-               /* Clear the filter that was created as part of
-                * validate_and_parse_flow() above
-                */
-               bnxt_hwrm_clear_l2_filter(bp, filter);
-               goto free_filter;
-       } else if (ret == -EXDEV) {
-               PMD_DRV_LOG(DEBUG, "Flow with same pattern exists");
-               PMD_DRV_LOG(DEBUG, "Updating with different destination\n");
-               update_flow = true;
-       }
-
-       if (filter->filter_type == HWRM_CFA_EM_FILTER) {
-               filter->enables |=
-                       HWRM_CFA_EM_FLOW_ALLOC_INPUT_ENABLES_L2_FILTER_ID;
-               ret = bnxt_hwrm_set_em_filter(bp, filter->dst_id, filter);
-       }
-       if (filter->filter_type == HWRM_CFA_NTUPLE_FILTER) {
-               filter->enables |=
-                       HWRM_CFA_NTUPLE_FILTER_ALLOC_INPUT_ENABLES_L2_FILTER_ID;
-               ret = bnxt_hwrm_set_ntuple_filter(bp, filter->dst_id, filter);
-       }
-
-       for (i = 0; i < bp->nr_vnics; i++) {
-               vnic = &bp->vnic_info[i];
-               if (filter->dst_id == vnic->fw_vnic_id)
-                       break;
-       }
-
-       if (!ret) {
-               flow->filter = filter;
-               flow->vnic = vnic;
-               if (update_flow) {
-                       ret = -EXDEV;
-                       goto free_flow;
-               }
-               PMD_DRV_LOG(ERR, "Successfully created flow.\n");
-               STAILQ_INSERT_TAIL(&vnic->flow_list, flow, next);
-               return flow;
-       }
-free_filter:
-       bnxt_free_filter(bp, filter);
-free_flow:
-       if (ret == -EEXIST)
-               rte_flow_error_set(error, ret,
-                                  RTE_FLOW_ERROR_TYPE_HANDLE, NULL,
-                                  "Matching Flow exists.");
-       else if (ret == -EXDEV)
-               rte_flow_error_set(error, ret,
-                                  RTE_FLOW_ERROR_TYPE_HANDLE, NULL,
-                                  "Flow with pattern exists, updating destination queue");
-       else
-               rte_flow_error_set(error, -ret,
-                                  RTE_FLOW_ERROR_TYPE_HANDLE, NULL,
-                                  "Failed to create flow.");
-       rte_free(flow);
-       flow = NULL;
-       return flow;
-}
-
-static int
-bnxt_flow_destroy(struct rte_eth_dev *dev,
-                 struct rte_flow *flow,
-                 struct rte_flow_error *error)
-{
-       struct bnxt *bp = (struct bnxt *)dev->data->dev_private;
-       struct bnxt_filter_info *filter = flow->filter;
-       struct bnxt_vnic_info *vnic = flow->vnic;
-       int ret = 0;
-
-       ret = bnxt_match_filter(bp, filter);
-       if (ret == 0)
-               PMD_DRV_LOG(ERR, "Could not find matching flow\n");
-       if (filter->filter_type == HWRM_CFA_EM_FILTER)
-               ret = bnxt_hwrm_clear_em_filter(bp, filter);
-       if (filter->filter_type == HWRM_CFA_NTUPLE_FILTER)
-               ret = bnxt_hwrm_clear_ntuple_filter(bp, filter);
-       else
-               ret = bnxt_hwrm_clear_l2_filter(bp, filter);
-       if (!ret) {
-               STAILQ_REMOVE(&vnic->flow_list, flow, rte_flow, next);
-               rte_free(flow);
-       } else {
-               rte_flow_error_set(error, -ret,
-                                  RTE_FLOW_ERROR_TYPE_HANDLE, NULL,
-                                  "Failed to destroy flow.");
-       }
-
-       return ret;
-}
-
-static int
-bnxt_flow_flush(struct rte_eth_dev *dev, struct rte_flow_error *error)
-{
-       struct bnxt *bp = (struct bnxt *)dev->data->dev_private;
-       struct bnxt_vnic_info *vnic;
-       struct rte_flow *flow;
-       unsigned int i;
-       int ret = 0;
-
-       for (i = 0; i < bp->nr_vnics; i++) {
-               vnic = &bp->vnic_info[i];
-               STAILQ_FOREACH(flow, &vnic->flow_list, next) {
-                       struct bnxt_filter_info *filter = flow->filter;
-
-                       if (filter->filter_type == HWRM_CFA_EM_FILTER)
-                               ret = bnxt_hwrm_clear_em_filter(bp, filter);
-                       if (filter->filter_type == HWRM_CFA_NTUPLE_FILTER)
-                               ret = bnxt_hwrm_clear_ntuple_filter(bp, filter);
-
-                       if (ret) {
-                               rte_flow_error_set(error, -ret,
-                                                  RTE_FLOW_ERROR_TYPE_HANDLE,
-                                                  NULL,
-                                                  "Failed to flush flow in HW.");
-                               return -rte_errno;
-                       }
-
-                       STAILQ_REMOVE(&vnic->flow_list, flow,
-                                     rte_flow, next);
-                       rte_free(flow);
-               }
-       }
-
-       return ret;
-}
-
-const struct rte_flow_ops bnxt_flow_ops = {
-       .validate = bnxt_flow_validate,
-       .create = bnxt_flow_create,
-       .destroy = bnxt_flow_destroy,
-       .flush = bnxt_flow_flush,
-};
diff --git a/drivers/net/bnxt/bnxt_flow.c b/drivers/net/bnxt/bnxt_flow.c
new file mode 100644 (file)
index 0000000..a491e9d
--- /dev/null
@@ -0,0 +1,1167 @@
+/* SPDX-License-Identifier: BSD-3-Clause
+ * Copyright(c) 2014-2018 Broadcom
+ * All rights reserved.
+ */
+
+#include <sys/queue.h>
+
+#include <rte_log.h>
+#include <rte_malloc.h>
+#include <rte_flow.h>
+#include <rte_flow_driver.h>
+#include <rte_tailq.h>
+
+#include "bnxt.h"
+#include "bnxt_filter.h"
+#include "bnxt_hwrm.h"
+#include "bnxt_vnic.h"
+#include "bnxt_util.h"
+#include "hsi_struct_def_dpdk.h"
+
+static int
+bnxt_flow_args_validate(const struct rte_flow_attr *attr,
+                       const struct rte_flow_item pattern[],
+                       const struct rte_flow_action actions[],
+                       struct rte_flow_error *error)
+{
+       if (!pattern) {
+               rte_flow_error_set(error,
+                                  EINVAL,
+                                  RTE_FLOW_ERROR_TYPE_ITEM_NUM,
+                                  NULL,
+                                  "NULL pattern.");
+               return -rte_errno;
+       }
+
+       if (!actions) {
+               rte_flow_error_set(error,
+                                  EINVAL,
+                                  RTE_FLOW_ERROR_TYPE_ACTION_NUM,
+                                  NULL,
+                                  "NULL action.");
+               return -rte_errno;
+       }
+
+       if (!attr) {
+               rte_flow_error_set(error,
+                                  EINVAL,
+                                  RTE_FLOW_ERROR_TYPE_ATTR,
+                                  NULL,
+                                  "NULL attribute.");
+               return -rte_errno;
+       }
+
+       return 0;
+}
+
+static const struct rte_flow_item *
+bnxt_flow_non_void_item(const struct rte_flow_item *cur)
+{
+       while (1) {
+               if (cur->type != RTE_FLOW_ITEM_TYPE_VOID)
+                       return cur;
+               cur++;
+       }
+}
+
+static const struct rte_flow_action *
+bnxt_flow_non_void_action(const struct rte_flow_action *cur)
+{
+       while (1) {
+               if (cur->type != RTE_FLOW_ACTION_TYPE_VOID)
+                       return cur;
+               cur++;
+       }
+}
+
+static int
+bnxt_filter_type_check(const struct rte_flow_item pattern[],
+                      struct rte_flow_error *error __rte_unused)
+{
+       const struct rte_flow_item *item =
+               bnxt_flow_non_void_item(pattern);
+       int use_ntuple = 1;
+
+       while (item->type != RTE_FLOW_ITEM_TYPE_END) {
+               switch (item->type) {
+               case RTE_FLOW_ITEM_TYPE_ETH:
+                       use_ntuple = 1;
+                       break;
+               case RTE_FLOW_ITEM_TYPE_VLAN:
+                       use_ntuple = 0;
+                       break;
+               case RTE_FLOW_ITEM_TYPE_IPV4:
+               case RTE_FLOW_ITEM_TYPE_IPV6:
+               case RTE_FLOW_ITEM_TYPE_TCP:
+               case RTE_FLOW_ITEM_TYPE_UDP:
+                       /* FALLTHROUGH */
+                       /* need ntuple match, reset exact match */
+                       if (!use_ntuple) {
+                               PMD_DRV_LOG(ERR,
+                                       "VLAN flow cannot use NTUPLE filter\n");
+                               rte_flow_error_set
+                                       (error,
+                                        EINVAL,
+                                        RTE_FLOW_ERROR_TYPE_ITEM,
+                                        item,
+                                        "Cannot use VLAN with NTUPLE");
+                               return -rte_errno;
+                       }
+                       use_ntuple |= 1;
+                       break;
+               default:
+                       PMD_DRV_LOG(ERR, "Unknown Flow type\n");
+                       use_ntuple |= 1;
+               }
+               item++;
+       }
+       return use_ntuple;
+}
+
+static int
+bnxt_validate_and_parse_flow_type(struct bnxt *bp,
+                                 const struct rte_flow_attr *attr,
+                                 const struct rte_flow_item pattern[],
+                                 struct rte_flow_error *error,
+                                 struct bnxt_filter_info *filter)
+{
+       const struct rte_flow_item *item = bnxt_flow_non_void_item(pattern);
+       const struct rte_flow_item_vlan *vlan_spec, *vlan_mask;
+       const struct rte_flow_item_ipv4 *ipv4_spec, *ipv4_mask;
+       const struct rte_flow_item_ipv6 *ipv6_spec, *ipv6_mask;
+       const struct rte_flow_item_tcp *tcp_spec, *tcp_mask;
+       const struct rte_flow_item_udp *udp_spec, *udp_mask;
+       const struct rte_flow_item_eth *eth_spec, *eth_mask;
+       const struct rte_flow_item_nvgre *nvgre_spec;
+       const struct rte_flow_item_nvgre *nvgre_mask;
+       const struct rte_flow_item_vxlan *vxlan_spec;
+       const struct rte_flow_item_vxlan *vxlan_mask;
+       uint8_t vni_mask[] = {0xFF, 0xFF, 0xFF};
+       uint8_t tni_mask[] = {0xFF, 0xFF, 0xFF};
+       const struct rte_flow_item_vf *vf_spec;
+       uint32_t tenant_id_be = 0;
+       bool vni_masked = 0;
+       bool tni_masked = 0;
+       uint32_t vf = 0;
+       int use_ntuple;
+       uint32_t en = 0;
+       uint32_t en_ethertype;
+       int dflt_vnic;
+
+       use_ntuple = bnxt_filter_type_check(pattern, error);
+       PMD_DRV_LOG(DEBUG, "Use NTUPLE %d\n", use_ntuple);
+       if (use_ntuple < 0)
+               return use_ntuple;
+
+       filter->filter_type = use_ntuple ?
+               HWRM_CFA_NTUPLE_FILTER : HWRM_CFA_EM_FILTER;
+       en_ethertype = use_ntuple ?
+               NTUPLE_FLTR_ALLOC_INPUT_EN_ETHERTYPE :
+               EM_FLOW_ALLOC_INPUT_EN_ETHERTYPE;
+
+       while (item->type != RTE_FLOW_ITEM_TYPE_END) {
+               if (item->last) {
+                       /* last or range is NOT supported as match criteria */
+                       rte_flow_error_set(error, EINVAL,
+                                          RTE_FLOW_ERROR_TYPE_ITEM,
+                                          item,
+                                          "No support for range");
+                       return -rte_errno;
+               }
+
+               if (!item->spec || !item->mask) {
+                       rte_flow_error_set(error, EINVAL,
+                                          RTE_FLOW_ERROR_TYPE_ITEM,
+                                          item,
+                                          "spec/mask is NULL");
+                       return -rte_errno;
+               }
+
+               switch (item->type) {
+               case RTE_FLOW_ITEM_TYPE_ETH:
+                       eth_spec = item->spec;
+                       eth_mask = item->mask;
+
+                       /* Source MAC address mask cannot be partially set.
+                        * Should be All 0's or all 1's.
+                        * Destination MAC address mask must not be partially
+                        * set. Should be all 1's or all 0's.
+                        */
+                       if ((!is_zero_ether_addr(&eth_mask->src) &&
+                            !is_broadcast_ether_addr(&eth_mask->src)) ||
+                           (!is_zero_ether_addr(&eth_mask->dst) &&
+                            !is_broadcast_ether_addr(&eth_mask->dst))) {
+                               rte_flow_error_set(error,
+                                                  EINVAL,
+                                                  RTE_FLOW_ERROR_TYPE_ITEM,
+                                                  item,
+                                                  "MAC_addr mask not valid");
+                               return -rte_errno;
+                       }
+
+                       /* Mask is not allowed. Only exact matches are */
+                       if (eth_mask->type &&
+                           eth_mask->type != RTE_BE16(0xffff)) {
+                               rte_flow_error_set(error, EINVAL,
+                                                  RTE_FLOW_ERROR_TYPE_ITEM,
+                                                  item,
+                                                  "ethertype mask not valid");
+                               return -rte_errno;
+                       }
+
+                       if (is_broadcast_ether_addr(&eth_mask->dst)) {
+                               rte_memcpy(filter->dst_macaddr,
+                                          &eth_spec->dst, 6);
+                               en |= use_ntuple ?
+                                       NTUPLE_FLTR_ALLOC_INPUT_EN_DST_MACADDR :
+                                       EM_FLOW_ALLOC_INPUT_EN_DST_MACADDR;
+                       }
+
+                       if (is_broadcast_ether_addr(&eth_mask->src)) {
+                               rte_memcpy(filter->src_macaddr,
+                                          &eth_spec->src, 6);
+                               en |= use_ntuple ?
+                                       NTUPLE_FLTR_ALLOC_INPUT_EN_SRC_MACADDR :
+                                       EM_FLOW_ALLOC_INPUT_EN_SRC_MACADDR;
+                       } /*
+                          * else {
+                          *  PMD_DRV_LOG(ERR, "Handle this condition\n");
+                          * }
+                          */
+                       if (eth_mask->type) {
+                               filter->ethertype =
+                                       rte_be_to_cpu_16(eth_spec->type);
+                               en |= en_ethertype;
+                       }
+
+                       break;
+               case RTE_FLOW_ITEM_TYPE_VLAN:
+                       vlan_spec = item->spec;
+                       vlan_mask = item->mask;
+                       if (en & en_ethertype) {
+                               rte_flow_error_set(error, EINVAL,
+                                                  RTE_FLOW_ERROR_TYPE_ITEM,
+                                                  item,
+                                                  "VLAN TPID matching is not"
+                                                  " supported");
+                               return -rte_errno;
+                       }
+                       if (vlan_mask->tci &&
+                           vlan_mask->tci == RTE_BE16(0x0fff)) {
+                               /* Only the VLAN ID can be matched. */
+                               filter->l2_ovlan =
+                                       rte_be_to_cpu_16(vlan_spec->tci &
+                                                        RTE_BE16(0x0fff));
+                               en |= EM_FLOW_ALLOC_INPUT_EN_OVLAN_VID;
+                       } else {
+                               rte_flow_error_set(error,
+                                                  EINVAL,
+                                                  RTE_FLOW_ERROR_TYPE_ITEM,
+                                                  item,
+                                                  "VLAN mask is invalid");
+                               return -rte_errno;
+                       }
+                       if (vlan_mask->inner_type &&
+                           vlan_mask->inner_type != RTE_BE16(0xffff)) {
+                               rte_flow_error_set(error, EINVAL,
+                                                  RTE_FLOW_ERROR_TYPE_ITEM,
+                                                  item,
+                                                  "inner ethertype mask not"
+                                                  " valid");
+                               return -rte_errno;
+                       }
+                       if (vlan_mask->inner_type) {
+                               filter->ethertype =
+                                       rte_be_to_cpu_16(vlan_spec->inner_type);
+                               en |= en_ethertype;
+                       }
+
+                       break;
+               case RTE_FLOW_ITEM_TYPE_IPV4:
+                       /* If mask is not involved, we could use EM filters. */
+                       ipv4_spec = item->spec;
+                       ipv4_mask = item->mask;
+                       /* Only IP DST and SRC fields are maskable. */
+                       if (ipv4_mask->hdr.version_ihl ||
+                           ipv4_mask->hdr.type_of_service ||
+                           ipv4_mask->hdr.total_length ||
+                           ipv4_mask->hdr.packet_id ||
+                           ipv4_mask->hdr.fragment_offset ||
+                           ipv4_mask->hdr.time_to_live ||
+                           ipv4_mask->hdr.next_proto_id ||
+                           ipv4_mask->hdr.hdr_checksum) {
+                               rte_flow_error_set(error,
+                                                  EINVAL,
+                                                  RTE_FLOW_ERROR_TYPE_ITEM,
+                                                  item,
+                                                  "Invalid IPv4 mask.");
+                               return -rte_errno;
+                       }
+
+                       filter->dst_ipaddr[0] = ipv4_spec->hdr.dst_addr;
+                       filter->src_ipaddr[0] = ipv4_spec->hdr.src_addr;
+
+                       if (use_ntuple)
+                               en |= NTUPLE_FLTR_ALLOC_INPUT_EN_SRC_IPADDR |
+                                       NTUPLE_FLTR_ALLOC_INPUT_EN_DST_IPADDR;
+                       else
+                               en |= EM_FLOW_ALLOC_INPUT_EN_SRC_IPADDR |
+                                       EM_FLOW_ALLOC_INPUT_EN_DST_IPADDR;
+
+                       if (ipv4_mask->hdr.src_addr) {
+                               filter->src_ipaddr_mask[0] =
+                                       ipv4_mask->hdr.src_addr;
+                               en |= !use_ntuple ? 0 :
+                                    NTUPLE_FLTR_ALLOC_INPUT_EN_SRC_IPADDR_MASK;
+                       }
+
+                       if (ipv4_mask->hdr.dst_addr) {
+                               filter->dst_ipaddr_mask[0] =
+                                       ipv4_mask->hdr.dst_addr;
+                               en |= !use_ntuple ? 0 :
+                                    NTUPLE_FLTR_ALLOC_INPUT_EN_DST_IPADDR_MASK;
+                       }
+
+                       filter->ip_addr_type = use_ntuple ?
+                        HWRM_CFA_NTUPLE_FILTER_ALLOC_INPUT_IP_ADDR_TYPE_IPV4 :
+                        HWRM_CFA_EM_FLOW_ALLOC_INPUT_IP_ADDR_TYPE_IPV4;
+
+                       if (ipv4_spec->hdr.next_proto_id) {
+                               filter->ip_protocol =
+                                       ipv4_spec->hdr.next_proto_id;
+                               if (use_ntuple)
+                                       en |= NTUPLE_FLTR_ALLOC_IN_EN_IP_PROTO;
+                               else
+                                       en |= EM_FLOW_ALLOC_INPUT_EN_IP_PROTO;
+                       }
+                       break;
+               case RTE_FLOW_ITEM_TYPE_IPV6:
+                       ipv6_spec = item->spec;
+                       ipv6_mask = item->mask;
+
+                       /* Only IP DST and SRC fields are maskable. */
+                       if (ipv6_mask->hdr.vtc_flow ||
+                           ipv6_mask->hdr.payload_len ||
+                           ipv6_mask->hdr.proto ||
+                           ipv6_mask->hdr.hop_limits) {
+                               rte_flow_error_set(error,
+                                                  EINVAL,
+                                                  RTE_FLOW_ERROR_TYPE_ITEM,
+                                                  item,
+                                                  "Invalid IPv6 mask.");
+                               return -rte_errno;
+                       }
+
+                       if (use_ntuple)
+                               en |= NTUPLE_FLTR_ALLOC_INPUT_EN_SRC_IPADDR |
+                                       NTUPLE_FLTR_ALLOC_INPUT_EN_DST_IPADDR;
+                       else
+                               en |= EM_FLOW_ALLOC_INPUT_EN_SRC_IPADDR |
+                                       EM_FLOW_ALLOC_INPUT_EN_DST_IPADDR;
+
+                       rte_memcpy(filter->src_ipaddr,
+                                  ipv6_spec->hdr.src_addr, 16);
+                       rte_memcpy(filter->dst_ipaddr,
+                                  ipv6_spec->hdr.dst_addr, 16);
+
+                       if (!bnxt_check_zero_bytes(ipv6_mask->hdr.src_addr,
+                                                  16)) {
+                               rte_memcpy(filter->src_ipaddr_mask,
+                                          ipv6_mask->hdr.src_addr, 16);
+                               en |= !use_ntuple ? 0 :
+                                   NTUPLE_FLTR_ALLOC_INPUT_EN_SRC_IPADDR_MASK;
+                       }
+
+                       if (!bnxt_check_zero_bytes(ipv6_mask->hdr.dst_addr,
+                                                  16)) {
+                               rte_memcpy(filter->dst_ipaddr_mask,
+                                          ipv6_mask->hdr.dst_addr, 16);
+                               en |= !use_ntuple ? 0 :
+                                    NTUPLE_FLTR_ALLOC_INPUT_EN_DST_IPADDR_MASK;
+                       }
+
+                       filter->ip_addr_type = use_ntuple ?
+                               NTUPLE_FLTR_ALLOC_INPUT_IP_ADDR_TYPE_IPV6 :
+                               EM_FLOW_ALLOC_INPUT_IP_ADDR_TYPE_IPV6;
+                       break;
+               case RTE_FLOW_ITEM_TYPE_TCP:
+                       tcp_spec = item->spec;
+                       tcp_mask = item->mask;
+
+                       /* Check TCP mask. Only DST & SRC ports are maskable */
+                       if (tcp_mask->hdr.sent_seq ||
+                           tcp_mask->hdr.recv_ack ||
+                           tcp_mask->hdr.data_off ||
+                           tcp_mask->hdr.tcp_flags ||
+                           tcp_mask->hdr.rx_win ||
+                           tcp_mask->hdr.cksum ||
+                           tcp_mask->hdr.tcp_urp) {
+                               rte_flow_error_set(error,
+                                                  EINVAL,
+                                                  RTE_FLOW_ERROR_TYPE_ITEM,
+                                                  item,
+                                                  "Invalid TCP mask");
+                               return -rte_errno;
+                       }
+
+                       filter->src_port = tcp_spec->hdr.src_port;
+                       filter->dst_port = tcp_spec->hdr.dst_port;
+
+                       if (use_ntuple)
+                               en |= NTUPLE_FLTR_ALLOC_INPUT_EN_SRC_PORT |
+                                       NTUPLE_FLTR_ALLOC_INPUT_EN_DST_PORT;
+                       else
+                               en |= EM_FLOW_ALLOC_INPUT_EN_SRC_PORT |
+                                       EM_FLOW_ALLOC_INPUT_EN_DST_PORT;
+
+                       if (tcp_mask->hdr.dst_port) {
+                               filter->dst_port_mask = tcp_mask->hdr.dst_port;
+                               en |= !use_ntuple ? 0 :
+                                 NTUPLE_FLTR_ALLOC_INPUT_EN_DST_PORT_MASK;
+                       }
+
+                       if (tcp_mask->hdr.src_port) {
+                               filter->src_port_mask = tcp_mask->hdr.src_port;
+                               en |= !use_ntuple ? 0 :
+                                 NTUPLE_FLTR_ALLOC_INPUT_EN_SRC_PORT_MASK;
+                       }
+                       break;
+               case RTE_FLOW_ITEM_TYPE_UDP:
+                       udp_spec = item->spec;
+                       udp_mask = item->mask;
+
+                       if (udp_mask->hdr.dgram_len ||
+                           udp_mask->hdr.dgram_cksum) {
+                               rte_flow_error_set(error,
+                                                  EINVAL,
+                                                  RTE_FLOW_ERROR_TYPE_ITEM,
+                                                  item,
+                                                  "Invalid UDP mask");
+                               return -rte_errno;
+                       }
+
+                       filter->src_port = udp_spec->hdr.src_port;
+                       filter->dst_port = udp_spec->hdr.dst_port;
+
+                       if (use_ntuple)
+                               en |= NTUPLE_FLTR_ALLOC_INPUT_EN_SRC_PORT |
+                                       NTUPLE_FLTR_ALLOC_INPUT_EN_DST_PORT;
+                       else
+                               en |= EM_FLOW_ALLOC_INPUT_EN_SRC_PORT |
+                                       EM_FLOW_ALLOC_INPUT_EN_DST_PORT;
+
+                       if (udp_mask->hdr.dst_port) {
+                               filter->dst_port_mask = udp_mask->hdr.dst_port;
+                               en |= !use_ntuple ? 0 :
+                                 NTUPLE_FLTR_ALLOC_INPUT_EN_DST_PORT_MASK;
+                       }
+
+                       if (udp_mask->hdr.src_port) {
+                               filter->src_port_mask = udp_mask->hdr.src_port;
+                               en |= !use_ntuple ? 0 :
+                                 NTUPLE_FLTR_ALLOC_INPUT_EN_SRC_PORT_MASK;
+                       }
+                       break;
+               case RTE_FLOW_ITEM_TYPE_VXLAN:
+                       vxlan_spec = item->spec;
+                       vxlan_mask = item->mask;
+                       /* Check if VXLAN item is used to describe protocol.
+                        * If yes, both spec and mask should be NULL.
+                        * If no, both spec and mask shouldn't be NULL.
+                        */
+                       if ((!vxlan_spec && vxlan_mask) ||
+                           (vxlan_spec && !vxlan_mask)) {
+                               rte_flow_error_set(error,
+                                                  EINVAL,
+                                                  RTE_FLOW_ERROR_TYPE_ITEM,
+                                                  item,
+                                                  "Invalid VXLAN item");
+                               return -rte_errno;
+                       }
+
+                       if (vxlan_spec->rsvd1 || vxlan_spec->rsvd0[0] ||
+                           vxlan_spec->rsvd0[1] || vxlan_spec->rsvd0[2] ||
+                           vxlan_spec->flags != 0x8) {
+                               rte_flow_error_set(error,
+                                                  EINVAL,
+                                                  RTE_FLOW_ERROR_TYPE_ITEM,
+                                                  item,
+                                                  "Invalid VXLAN item");
+                               return -rte_errno;
+                       }
+
+                       /* Check if VNI is masked. */
+                       if (vxlan_spec && vxlan_mask) {
+                               vni_masked =
+                                       !!memcmp(vxlan_mask->vni, vni_mask,
+                                                RTE_DIM(vni_mask));
+                               if (vni_masked) {
+                                       rte_flow_error_set
+                                               (error,
+                                                EINVAL,
+                                                RTE_FLOW_ERROR_TYPE_ITEM,
+                                                item,
+                                                "Invalid VNI mask");
+                                       return -rte_errno;
+                               }
+
+                               rte_memcpy(((uint8_t *)&tenant_id_be + 1),
+                                          vxlan_spec->vni, 3);
+                               filter->vni =
+                                       rte_be_to_cpu_32(tenant_id_be);
+                               filter->tunnel_type =
+                                CFA_NTUPLE_FILTER_ALLOC_REQ_TUNNEL_TYPE_VXLAN;
+                       }
+                       break;
+               case RTE_FLOW_ITEM_TYPE_NVGRE:
+                       nvgre_spec = item->spec;
+                       nvgre_mask = item->mask;
+                       /* Check if NVGRE item is used to describe protocol.
+                        * If yes, both spec and mask should be NULL.
+                        * If no, both spec and mask shouldn't be NULL.
+                        */
+                       if ((!nvgre_spec && nvgre_mask) ||
+                           (nvgre_spec && !nvgre_mask)) {
+                               rte_flow_error_set(error,
+                                                  EINVAL,
+                                                  RTE_FLOW_ERROR_TYPE_ITEM,
+                                                  item,
+                                                  "Invalid NVGRE item");
+                               return -rte_errno;
+                       }
+
+                       if (nvgre_spec->c_k_s_rsvd0_ver != 0x2000 ||
+                           nvgre_spec->protocol != 0x6558) {
+                               rte_flow_error_set(error,
+                                                  EINVAL,
+                                                  RTE_FLOW_ERROR_TYPE_ITEM,
+                                                  item,
+                                                  "Invalid NVGRE item");
+                               return -rte_errno;
+                       }
+
+                       if (nvgre_spec && nvgre_mask) {
+                               tni_masked =
+                                       !!memcmp(nvgre_mask->tni, tni_mask,
+                                                RTE_DIM(tni_mask));
+                               if (tni_masked) {
+                                       rte_flow_error_set
+                                               (error,
+                                                EINVAL,
+                                                RTE_FLOW_ERROR_TYPE_ITEM,
+                                                item,
+                                                "Invalid TNI mask");
+                                       return -rte_errno;
+                               }
+                               rte_memcpy(((uint8_t *)&tenant_id_be + 1),
+                                          nvgre_spec->tni, 3);
+                               filter->vni =
+                                       rte_be_to_cpu_32(tenant_id_be);
+                               filter->tunnel_type =
+                                CFA_NTUPLE_FILTER_ALLOC_REQ_TUNNEL_TYPE_NVGRE;
+                       }
+                       break;
+               case RTE_FLOW_ITEM_TYPE_VF:
+                       vf_spec = item->spec;
+                       vf = vf_spec->id;
+
+                       if (!BNXT_PF(bp)) {
+                               rte_flow_error_set(error,
+                                                  EINVAL,
+                                                  RTE_FLOW_ERROR_TYPE_ITEM,
+                                                  item,
+                                                  "Configuring on a VF!");
+                               return -rte_errno;
+                       }
+
+                       if (vf >= bp->pdev->max_vfs) {
+                               rte_flow_error_set(error,
+                                                  EINVAL,
+                                                  RTE_FLOW_ERROR_TYPE_ITEM,
+                                                  item,
+                                                  "Incorrect VF id!");
+                               return -rte_errno;
+                       }
+
+                       if (!attr->transfer) {
+                               rte_flow_error_set(error,
+                                                  ENOTSUP,
+                                                  RTE_FLOW_ERROR_TYPE_ITEM,
+                                                  item,
+                                                  "Matching VF traffic without"
+                                                  " affecting it (transfer attribute)"
+                                                  " is unsupported");
+                               return -rte_errno;
+                       }
+
+                       filter->mirror_vnic_id =
+                       dflt_vnic = bnxt_hwrm_func_qcfg_vf_dflt_vnic_id(bp, vf);
+                       if (dflt_vnic < 0) {
+                               /* This simply indicates there's no driver
+                                * loaded. This is not an error.
+                                */
+                               rte_flow_error_set
+                                       (error,
+                                        EINVAL,
+                                        RTE_FLOW_ERROR_TYPE_ITEM,
+                                        item,
+                                        "Unable to get default VNIC for VF");
+                               return -rte_errno;
+                       }
+
+                       filter->mirror_vnic_id = dflt_vnic;
+                       en |= NTUPLE_FLTR_ALLOC_INPUT_EN_MIRROR_VNIC_ID;
+                       break;
+               default:
+                       break;
+               }
+               item++;
+       }
+       filter->enables = en;
+
+       return 0;
+}
+
+/* Parse attributes */
+static int
+bnxt_flow_parse_attr(const struct rte_flow_attr *attr,
+                    struct rte_flow_error *error)
+{
+       /* Must be input direction */
+       if (!attr->ingress) {
+               rte_flow_error_set(error,
+                                  EINVAL,
+                                  RTE_FLOW_ERROR_TYPE_ATTR_INGRESS,
+                                  attr,
+                                  "Only support ingress.");
+               return -rte_errno;
+       }
+
+       /* Not supported */
+       if (attr->egress) {
+               rte_flow_error_set(error,
+                                  EINVAL,
+                                  RTE_FLOW_ERROR_TYPE_ATTR_EGRESS,
+                                  attr,
+                                  "No support for egress.");
+               return -rte_errno;
+       }
+
+       /* Not supported */
+       if (attr->priority) {
+               rte_flow_error_set(error,
+                                  EINVAL,
+                                  RTE_FLOW_ERROR_TYPE_ATTR_PRIORITY,
+                                  attr,
+                                  "No support for priority.");
+               return -rte_errno;
+       }
+
+       /* Not supported */
+       if (attr->group) {
+               rte_flow_error_set(error,
+                                  EINVAL,
+                                  RTE_FLOW_ERROR_TYPE_ATTR_GROUP,
+                                  attr,
+                                  "No support for group.");
+               return -rte_errno;
+       }
+
+       return 0;
+}
+
+struct bnxt_filter_info *
+bnxt_get_l2_filter(struct bnxt *bp, struct bnxt_filter_info *nf,
+                  struct bnxt_vnic_info *vnic)
+{
+       struct bnxt_filter_info *filter1, *f0;
+       struct bnxt_vnic_info *vnic0;
+       int rc;
+
+       vnic0 = STAILQ_FIRST(&bp->ff_pool[0]);
+       f0 = STAILQ_FIRST(&vnic0->filter);
+
+       /* This flow has same DST MAC as the port/l2 filter. */
+       if (memcmp(f0->l2_addr, nf->dst_macaddr, ETHER_ADDR_LEN) == 0)
+               return f0;
+
+       /* This flow needs DST MAC which is not same as port/l2 */
+       PMD_DRV_LOG(DEBUG, "Create L2 filter for DST MAC\n");
+       filter1 = bnxt_get_unused_filter(bp);
+       if (filter1 == NULL)
+               return NULL;
+
+       filter1->flags = HWRM_CFA_L2_FILTER_ALLOC_INPUT_FLAGS_PATH_RX;
+       filter1->enables = HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_L2_ADDR |
+                       L2_FILTER_ALLOC_INPUT_EN_L2_ADDR_MASK;
+       memcpy(filter1->l2_addr, nf->dst_macaddr, ETHER_ADDR_LEN);
+       memset(filter1->l2_addr_mask, 0xff, ETHER_ADDR_LEN);
+       rc = bnxt_hwrm_set_l2_filter(bp, vnic->fw_vnic_id,
+                                    filter1);
+       if (rc) {
+               bnxt_free_filter(bp, filter1);
+               return NULL;
+       }
+       return filter1;
+}
+
+static int
+bnxt_validate_and_parse_flow(struct rte_eth_dev *dev,
+                            const struct rte_flow_item pattern[],
+                            const struct rte_flow_action actions[],
+                            const struct rte_flow_attr *attr,
+                            struct rte_flow_error *error,
+                            struct bnxt_filter_info *filter)
+{
+       const struct rte_flow_action *act =
+               bnxt_flow_non_void_action(actions);
+       struct bnxt *bp = (struct bnxt *)dev->data->dev_private;
+       const struct rte_flow_action_queue *act_q;
+       const struct rte_flow_action_vf *act_vf;
+       struct bnxt_vnic_info *vnic, *vnic0;
+       struct bnxt_filter_info *filter1;
+       uint32_t vf = 0;
+       int dflt_vnic;
+       int rc;
+
+       if (bp->eth_dev->data->dev_conf.rxmode.mq_mode & ETH_MQ_RX_RSS) {
+               PMD_DRV_LOG(ERR, "Cannot create flow on RSS queues\n");
+               rte_flow_error_set(error,
+                                  EINVAL,
+                                  RTE_FLOW_ERROR_TYPE_UNSPECIFIED,
+                                  NULL,
+                                  "Cannot create flow on RSS queues");
+               rc = -rte_errno;
+               goto ret;
+       }
+
+       rc =
+       bnxt_validate_and_parse_flow_type(bp, attr, pattern, error, filter);
+       if (rc != 0)
+               goto ret;
+
+       rc = bnxt_flow_parse_attr(attr, error);
+       if (rc != 0)
+               goto ret;
+
+       /* Since we support ingress attribute only - right now. */
+       if (filter->filter_type == HWRM_CFA_EM_FILTER)
+               filter->flags = HWRM_CFA_EM_FLOW_ALLOC_INPUT_FLAGS_PATH_RX;
+
+       switch (act->type) {
+       case RTE_FLOW_ACTION_TYPE_QUEUE:
+               /* Allow this flow. Redirect to a VNIC. */
+               act_q = (const struct rte_flow_action_queue *)act->conf;
+               if (act_q->index >= bp->rx_nr_rings) {
+                       rte_flow_error_set(error,
+                                          EINVAL,
+                                          RTE_FLOW_ERROR_TYPE_ACTION,
+                                          act,
+                                          "Invalid queue ID.");
+                       rc = -rte_errno;
+                       goto ret;
+               }
+               PMD_DRV_LOG(DEBUG, "Queue index %d\n", act_q->index);
+
+               vnic0 = STAILQ_FIRST(&bp->ff_pool[0]);
+               vnic = STAILQ_FIRST(&bp->ff_pool[act_q->index]);
+               if (vnic == NULL) {
+                       rte_flow_error_set(error,
+                                          EINVAL,
+                                          RTE_FLOW_ERROR_TYPE_ACTION,
+                                          act,
+                                          "No matching VNIC for queue ID.");
+                       rc = -rte_errno;
+                       goto ret;
+               }
+
+               filter->dst_id = vnic->fw_vnic_id;
+               filter1 = bnxt_get_l2_filter(bp, filter, vnic);
+               if (filter1 == NULL) {
+                       rc = -ENOSPC;
+                       goto ret;
+               }
+
+               filter->fw_l2_filter_id = filter1->fw_l2_filter_id;
+               PMD_DRV_LOG(DEBUG, "VNIC found\n");
+               break;
+       case RTE_FLOW_ACTION_TYPE_DROP:
+               vnic0 = STAILQ_FIRST(&bp->ff_pool[0]);
+               filter1 = bnxt_get_l2_filter(bp, filter, vnic0);
+               if (filter1 == NULL) {
+                       rc = -ENOSPC;
+                       goto ret;
+               }
+
+               filter->fw_l2_filter_id = filter1->fw_l2_filter_id;
+               if (filter->filter_type == HWRM_CFA_EM_FILTER)
+                       filter->flags =
+                               HWRM_CFA_EM_FLOW_ALLOC_INPUT_FLAGS_DROP;
+               else
+                       filter->flags =
+                               HWRM_CFA_NTUPLE_FILTER_ALLOC_INPUT_FLAGS_DROP;
+               break;
+       case RTE_FLOW_ACTION_TYPE_COUNT:
+               vnic0 = STAILQ_FIRST(&bp->ff_pool[0]);
+               filter1 = bnxt_get_l2_filter(bp, filter, vnic0);
+               if (filter1 == NULL) {
+                       rc = -ENOSPC;
+                       goto ret;
+               }
+
+               filter->fw_l2_filter_id = filter1->fw_l2_filter_id;
+               filter->flags = HWRM_CFA_NTUPLE_FILTER_ALLOC_INPUT_FLAGS_METER;
+               break;
+       case RTE_FLOW_ACTION_TYPE_VF:
+               act_vf = (const struct rte_flow_action_vf *)act->conf;
+               vf = act_vf->id;
+
+               if (!BNXT_PF(bp)) {
+                       rte_flow_error_set(error,
+                                          EINVAL,
+                                          RTE_FLOW_ERROR_TYPE_ACTION,
+                                          act,
+                                          "Configuring on a VF!");
+                       rc = -rte_errno;
+                       goto ret;
+               }
+
+               if (vf >= bp->pdev->max_vfs) {
+                       rte_flow_error_set(error,
+                                          EINVAL,
+                                          RTE_FLOW_ERROR_TYPE_ACTION,
+                                          act,
+                                          "Incorrect VF id!");
+                       rc = -rte_errno;
+                       goto ret;
+               }
+
+               filter->mirror_vnic_id =
+               dflt_vnic = bnxt_hwrm_func_qcfg_vf_dflt_vnic_id(bp, vf);
+               if (dflt_vnic < 0) {
+                       /* This simply indicates there's no driver loaded.
+                        * This is not an error.
+                        */
+                       rte_flow_error_set(error,
+                                          EINVAL,
+                                          RTE_FLOW_ERROR_TYPE_ACTION,
+                                          act,
+                                          "Unable to get default VNIC for VF");
+                       rc = -rte_errno;
+                       goto ret;
+               }
+
+               filter->mirror_vnic_id = dflt_vnic;
+               filter->enables |= NTUPLE_FLTR_ALLOC_INPUT_EN_MIRROR_VNIC_ID;
+
+               vnic0 = STAILQ_FIRST(&bp->ff_pool[0]);
+               filter1 = bnxt_get_l2_filter(bp, filter, vnic0);
+               if (filter1 == NULL) {
+                       rc = -ENOSPC;
+                       goto ret;
+               }
+
+               filter->fw_l2_filter_id = filter1->fw_l2_filter_id;
+               break;
+
+       default:
+               rte_flow_error_set(error,
+                                  EINVAL,
+                                  RTE_FLOW_ERROR_TYPE_ACTION,
+                                  act,
+                                  "Invalid action.");
+               rc = -rte_errno;
+               goto ret;
+       }
+
+       if (filter1) {
+               bnxt_free_filter(bp, filter1);
+               filter1->fw_l2_filter_id = -1;
+       }
+
+       act = bnxt_flow_non_void_action(++act);
+       if (act->type != RTE_FLOW_ACTION_TYPE_END) {
+               rte_flow_error_set(error,
+                                  EINVAL,
+                                  RTE_FLOW_ERROR_TYPE_ACTION,
+                                  act,
+                                  "Invalid action.");
+               rc = -rte_errno;
+               goto ret;
+       }
+ret:
+       return rc;
+}
+
+static int
+bnxt_flow_validate(struct rte_eth_dev *dev,
+                  const struct rte_flow_attr *attr,
+                  const struct rte_flow_item pattern[],
+                  const struct rte_flow_action actions[],
+                  struct rte_flow_error *error)
+{
+       struct bnxt *bp = (struct bnxt *)dev->data->dev_private;
+       struct bnxt_filter_info *filter;
+       int ret = 0;
+
+       ret = bnxt_flow_args_validate(attr, pattern, actions, error);
+       if (ret != 0)
+               return ret;
+
+       filter = bnxt_get_unused_filter(bp);
+       if (filter == NULL) {
+               PMD_DRV_LOG(ERR, "Not enough resources for a new flow.\n");
+               return -ENOMEM;
+       }
+
+       ret = bnxt_validate_and_parse_flow(dev, pattern, actions, attr,
+                                          error, filter);
+       /* No need to hold on to this filter if we are just validating flow */
+       filter->fw_l2_filter_id = UINT64_MAX;
+       bnxt_free_filter(bp, filter);
+
+       return ret;
+}
+
+static int
+bnxt_match_filter(struct bnxt *bp, struct bnxt_filter_info *nf)
+{
+       struct bnxt_filter_info *mf;
+       struct rte_flow *flow;
+       int i;
+
+       for (i = bp->nr_vnics - 1; i >= 0; i--) {
+               struct bnxt_vnic_info *vnic = &bp->vnic_info[i];
+
+               STAILQ_FOREACH(flow, &vnic->flow_list, next) {
+                       mf = flow->filter;
+
+                       if (mf->filter_type == nf->filter_type &&
+                           mf->flags == nf->flags &&
+                           mf->src_port == nf->src_port &&
+                           mf->src_port_mask == nf->src_port_mask &&
+                           mf->dst_port == nf->dst_port &&
+                           mf->dst_port_mask == nf->dst_port_mask &&
+                           mf->ip_protocol == nf->ip_protocol &&
+                           mf->ip_addr_type == nf->ip_addr_type &&
+                           mf->ethertype == nf->ethertype &&
+                           mf->vni == nf->vni &&
+                           mf->tunnel_type == nf->tunnel_type &&
+                           mf->l2_ovlan == nf->l2_ovlan &&
+                           mf->l2_ovlan_mask == nf->l2_ovlan_mask &&
+                           mf->l2_ivlan == nf->l2_ivlan &&
+                           mf->l2_ivlan_mask == nf->l2_ivlan_mask &&
+                           !memcmp(mf->l2_addr, nf->l2_addr, ETHER_ADDR_LEN) &&
+                           !memcmp(mf->l2_addr_mask, nf->l2_addr_mask,
+                                   ETHER_ADDR_LEN) &&
+                           !memcmp(mf->src_macaddr, nf->src_macaddr,
+                                   ETHER_ADDR_LEN) &&
+                           !memcmp(mf->dst_macaddr, nf->dst_macaddr,
+                                   ETHER_ADDR_LEN) &&
+                           !memcmp(mf->src_ipaddr, nf->src_ipaddr,
+                                   sizeof(nf->src_ipaddr)) &&
+                           !memcmp(mf->src_ipaddr_mask, nf->src_ipaddr_mask,
+                                   sizeof(nf->src_ipaddr_mask)) &&
+                           !memcmp(mf->dst_ipaddr, nf->dst_ipaddr,
+                                   sizeof(nf->dst_ipaddr)) &&
+                           !memcmp(mf->dst_ipaddr_mask, nf->dst_ipaddr_mask,
+                                   sizeof(nf->dst_ipaddr_mask))) {
+                               if (mf->dst_id == nf->dst_id)
+                                       return -EEXIST;
+                               /* Same Flow, Different queue
+                                * Clear the old ntuple filter
+                                */
+                               if (nf->filter_type == HWRM_CFA_EM_FILTER)
+                                       bnxt_hwrm_clear_em_filter(bp, mf);
+                               if (nf->filter_type == HWRM_CFA_NTUPLE_FILTER)
+                                       bnxt_hwrm_clear_ntuple_filter(bp, mf);
+                               /* Free the old filter, update flow
+                                * with new filter
+                                */
+                               bnxt_free_filter(bp, mf);
+                               flow->filter = nf;
+                               return -EXDEV;
+                       }
+               }
+       }
+       return 0;
+}
+
+static struct rte_flow *
+bnxt_flow_create(struct rte_eth_dev *dev,
+                const struct rte_flow_attr *attr,
+                const struct rte_flow_item pattern[],
+                const struct rte_flow_action actions[],
+                struct rte_flow_error *error)
+{
+       struct bnxt *bp = (struct bnxt *)dev->data->dev_private;
+       struct bnxt_filter_info *filter;
+       struct bnxt_vnic_info *vnic = NULL;
+       bool update_flow = false;
+       struct rte_flow *flow;
+       unsigned int i;
+       int ret = 0;
+
+       flow = rte_zmalloc("bnxt_flow", sizeof(struct rte_flow), 0);
+       if (!flow) {
+               rte_flow_error_set(error, ENOMEM,
+                                  RTE_FLOW_ERROR_TYPE_HANDLE, NULL,
+                                  "Failed to allocate memory");
+               return flow;
+       }
+
+       ret = bnxt_flow_args_validate(attr, pattern, actions, error);
+       if (ret != 0) {
+               PMD_DRV_LOG(ERR, "Not a validate flow.\n");
+               goto free_flow;
+       }
+
+       filter = bnxt_get_unused_filter(bp);
+       if (filter == NULL) {
+               PMD_DRV_LOG(ERR, "Not enough resources for a new flow.\n");
+               goto free_flow;
+       }
+
+       ret = bnxt_validate_and_parse_flow(dev, pattern, actions, attr,
+                                          error, filter);
+       if (ret != 0)
+               goto free_filter;
+
+       ret = bnxt_match_filter(bp, filter);
+       if (ret == -EEXIST) {
+               PMD_DRV_LOG(DEBUG, "Flow already exists.\n");
+               /* Clear the filter that was created as part of
+                * validate_and_parse_flow() above
+                */
+               bnxt_hwrm_clear_l2_filter(bp, filter);
+               goto free_filter;
+       } else if (ret == -EXDEV) {
+               PMD_DRV_LOG(DEBUG, "Flow with same pattern exists\n");
+               PMD_DRV_LOG(DEBUG, "Updating with different destination\n");
+               update_flow = true;
+       }
+
+       if (filter->filter_type == HWRM_CFA_EM_FILTER) {
+               filter->enables |=
+                       HWRM_CFA_EM_FLOW_ALLOC_INPUT_ENABLES_L2_FILTER_ID;
+               ret = bnxt_hwrm_set_em_filter(bp, filter->dst_id, filter);
+       }
+
+       if (filter->filter_type == HWRM_CFA_NTUPLE_FILTER) {
+               filter->enables |=
+                       HWRM_CFA_NTUPLE_FILTER_ALLOC_INPUT_ENABLES_L2_FILTER_ID;
+               ret = bnxt_hwrm_set_ntuple_filter(bp, filter->dst_id, filter);
+       }
+
+       for (i = 0; i < bp->nr_vnics; i++) {
+               vnic = &bp->vnic_info[i];
+               if (filter->dst_id == vnic->fw_vnic_id)
+                       break;
+       }
+
+       if (!ret) {
+               flow->filter = filter;
+               flow->vnic = vnic;
+               if (update_flow) {
+                       ret = -EXDEV;
+                       goto free_flow;
+               }
+               PMD_DRV_LOG(ERR, "Successfully created flow.\n");
+               STAILQ_INSERT_TAIL(&vnic->flow_list, flow, next);
+               return flow;
+       }
+free_filter:
+       bnxt_free_filter(bp, filter);
+free_flow:
+       if (ret == -EEXIST)
+               rte_flow_error_set(error, ret,
+                                  RTE_FLOW_ERROR_TYPE_HANDLE, NULL,
+                                  "Matching Flow exists.");
+       else if (ret == -EXDEV)
+               rte_flow_error_set(error, ret,
+                                  RTE_FLOW_ERROR_TYPE_HANDLE, NULL,
+                                  "Flow with pattern exists, updating destination queue");
+       else
+               rte_flow_error_set(error, -ret,
+                                  RTE_FLOW_ERROR_TYPE_HANDLE, NULL,
+                                  "Failed to create flow.");
+       rte_free(flow);
+       flow = NULL;
+       return flow;
+}
+
+static int
+bnxt_flow_destroy(struct rte_eth_dev *dev,
+                 struct rte_flow *flow,
+                 struct rte_flow_error *error)
+{
+       struct bnxt *bp = (struct bnxt *)dev->data->dev_private;
+       struct bnxt_filter_info *filter = flow->filter;
+       struct bnxt_vnic_info *vnic = flow->vnic;
+       int ret = 0;
+
+       ret = bnxt_match_filter(bp, filter);
+       if (ret == 0)
+               PMD_DRV_LOG(ERR, "Could not find matching flow\n");
+       if (filter->filter_type == HWRM_CFA_EM_FILTER)
+               ret = bnxt_hwrm_clear_em_filter(bp, filter);
+       if (filter->filter_type == HWRM_CFA_NTUPLE_FILTER)
+               ret = bnxt_hwrm_clear_ntuple_filter(bp, filter);
+       else
+               ret = bnxt_hwrm_clear_l2_filter(bp, filter);
+       if (!ret) {
+               STAILQ_REMOVE(&vnic->flow_list, flow, rte_flow, next);
+               rte_free(flow);
+       } else {
+               rte_flow_error_set(error, -ret,
+                                  RTE_FLOW_ERROR_TYPE_HANDLE, NULL,
+                                  "Failed to destroy flow.");
+       }
+
+       return ret;
+}
+
+static int
+bnxt_flow_flush(struct rte_eth_dev *dev, struct rte_flow_error *error)
+{
+       struct bnxt *bp = (struct bnxt *)dev->data->dev_private;
+       struct bnxt_vnic_info *vnic;
+       struct rte_flow *flow;
+       unsigned int i;
+       int ret = 0;
+
+       for (i = 0; i < bp->nr_vnics; i++) {
+               vnic = &bp->vnic_info[i];
+               STAILQ_FOREACH(flow, &vnic->flow_list, next) {
+                       struct bnxt_filter_info *filter = flow->filter;
+
+                       if (filter->filter_type == HWRM_CFA_EM_FILTER)
+                               ret = bnxt_hwrm_clear_em_filter(bp, filter);
+                       if (filter->filter_type == HWRM_CFA_NTUPLE_FILTER)
+                               ret = bnxt_hwrm_clear_ntuple_filter(bp, filter);
+
+                       if (ret) {
+                               rte_flow_error_set
+                                       (error,
+                                        -ret,
+                                        RTE_FLOW_ERROR_TYPE_HANDLE,
+                                        NULL,
+                                        "Failed to flush flow in HW.");
+                               return -rte_errno;
+                       }
+
+                       STAILQ_REMOVE(&vnic->flow_list, flow,
+                                     rte_flow, next);
+                       rte_free(flow);
+               }
+       }
+
+       return ret;
+}
+
+const struct rte_flow_ops bnxt_flow_ops = {
+       .validate = bnxt_flow_validate,
+       .create = bnxt_flow_create,
+       .destroy = bnxt_flow_destroy,
+       .flush = bnxt_flow_flush,
+};
index bfb448f..e130f27 100644 (file)
@@ -6,6 +6,7 @@ version = 2
 sources = files('bnxt_cpr.c',
        'bnxt_ethdev.c',
        'bnxt_filter.c',
+       'bnxt_flow.c',
        'bnxt_hwrm.c',
        'bnxt_irq.c',
        'bnxt_ring.c',