net/i40e: support AVF basic interface
authorJingjing Wu <jingjing.wu@intel.com>
Wed, 10 Jan 2018 13:02:03 +0000 (21:02 +0800)
committerFerruh Yigit <ferruh.yigit@intel.com>
Tue, 16 Jan 2018 17:47:49 +0000 (18:47 +0100)
Enable Virtchnl offload Caps negotiation and RSS_PF offload
to support AVF basic interface.

Signed-off-by: Jingjing Wu <jingjing.wu@intel.com>
drivers/net/i40e/i40e_ethdev.c
drivers/net/i40e/i40e_ethdev.h
drivers/net/i40e/i40e_pf.c
drivers/net/i40e/i40e_pf.h

index 0ac5909..27f766a 100644 (file)
@@ -3719,6 +3719,7 @@ i40e_get_rss_lut(struct i40e_vsi *vsi, uint8_t *lut, uint16_t lut_size)
 {
        struct i40e_pf *pf = I40E_VSI_TO_PF(vsi);
        struct i40e_hw *hw = I40E_VSI_TO_HW(vsi);
+       uint32_t reg;
        int ret;
 
        if (!lut)
@@ -3735,14 +3736,22 @@ i40e_get_rss_lut(struct i40e_vsi *vsi, uint8_t *lut, uint16_t lut_size)
                uint32_t *lut_dw = (uint32_t *)lut;
                uint16_t i, lut_size_dw = lut_size / 4;
 
-               for (i = 0; i < lut_size_dw; i++)
-                       lut_dw[i] = I40E_READ_REG(hw, I40E_PFQF_HLUT(i));
+               if (vsi->type == I40E_VSI_SRIOV) {
+                       for (i = 0; i <= lut_size_dw; i++) {
+                               reg = I40E_VFQF_HLUT1(i, vsi->user_param);
+                               lut_dw[i] = i40e_read_rx_ctl(hw, reg);
+                       }
+               } else {
+                       for (i = 0; i < lut_size_dw; i++)
+                               lut_dw[i] = I40E_READ_REG(hw,
+                                                         I40E_PFQF_HLUT(i));
+               }
        }
 
        return 0;
 }
 
-static int
+int
 i40e_set_rss_lut(struct i40e_vsi *vsi, uint8_t *lut, uint16_t lut_size)
 {
        struct i40e_pf *pf;
@@ -3766,8 +3775,17 @@ i40e_set_rss_lut(struct i40e_vsi *vsi, uint8_t *lut, uint16_t lut_size)
                uint32_t *lut_dw = (uint32_t *)lut;
                uint16_t i, lut_size_dw = lut_size / 4;
 
-               for (i = 0; i < lut_size_dw; i++)
-                       I40E_WRITE_REG(hw, I40E_PFQF_HLUT(i), lut_dw[i]);
+               if (vsi->type == I40E_VSI_SRIOV) {
+                       for (i = 0; i < lut_size_dw; i++)
+                               I40E_WRITE_REG(
+                                       hw,
+                                       I40E_VFQF_HLUT1(i, vsi->user_param),
+                                       lut_dw[i]);
+               } else {
+                       for (i = 0; i < lut_size_dw; i++)
+                               I40E_WRITE_REG(hw, I40E_PFQF_HLUT(i),
+                                              lut_dw[i]);
+               }
                I40E_WRITE_FLUSH(hw);
        }
 
@@ -6804,17 +6822,20 @@ i40e_pf_disable_rss(struct i40e_pf *pf)
        I40E_WRITE_FLUSH(hw);
 }
 
-static int
+int
 i40e_set_rss_key(struct i40e_vsi *vsi, uint8_t *key, uint8_t key_len)
 {
        struct i40e_pf *pf = I40E_VSI_TO_PF(vsi);
        struct i40e_hw *hw = I40E_VSI_TO_HW(vsi);
+       uint16_t key_idx = (vsi->type == I40E_VSI_SRIOV) ?
+                          I40E_VFQF_HKEY_MAX_INDEX :
+                          I40E_PFQF_HKEY_MAX_INDEX;
        int ret = 0;
 
        if (!key || key_len == 0) {
                PMD_DRV_LOG(DEBUG, "No key to be configured");
                return 0;
-       } else if (key_len != (I40E_PFQF_HKEY_MAX_INDEX + 1) *
+       } else if (key_len != (key_idx + 1) *
                sizeof(uint32_t)) {
                PMD_DRV_LOG(ERR, "Invalid key length %u", key_len);
                return -EINVAL;
@@ -6831,8 +6852,18 @@ i40e_set_rss_key(struct i40e_vsi *vsi, uint8_t *key, uint8_t key_len)
                uint32_t *hash_key = (uint32_t *)key;
                uint16_t i;
 
-               for (i = 0; i <= I40E_PFQF_HKEY_MAX_INDEX; i++)
-                       i40e_write_rx_ctl(hw, I40E_PFQF_HKEY(i), hash_key[i]);
+               if (vsi->type == I40E_VSI_SRIOV) {
+                       for (i = 0; i <= I40E_VFQF_HKEY_MAX_INDEX; i++)
+                               I40E_WRITE_REG(
+                                       hw,
+                                       I40E_VFQF_HKEY1(i, vsi->user_param),
+                                       hash_key[i]);
+
+               } else {
+                       for (i = 0; i <= I40E_PFQF_HKEY_MAX_INDEX; i++)
+                               I40E_WRITE_REG(hw, I40E_PFQF_HKEY(i),
+                                              hash_key[i]);
+               }
                I40E_WRITE_FLUSH(hw);
        }
 
@@ -6844,6 +6875,7 @@ i40e_get_rss_key(struct i40e_vsi *vsi, uint8_t *key, uint8_t *key_len)
 {
        struct i40e_pf *pf = I40E_VSI_TO_PF(vsi);
        struct i40e_hw *hw = I40E_VSI_TO_HW(vsi);
+       uint32_t reg;
        int ret;
 
        if (!key || !key_len)
@@ -6860,11 +6892,22 @@ i40e_get_rss_key(struct i40e_vsi *vsi, uint8_t *key, uint8_t *key_len)
                uint32_t *key_dw = (uint32_t *)key;
                uint16_t i;
 
-               for (i = 0; i <= I40E_PFQF_HKEY_MAX_INDEX; i++)
-                       key_dw[i] = i40e_read_rx_ctl(hw, I40E_PFQF_HKEY(i));
+               if (vsi->type == I40E_VSI_SRIOV) {
+                       for (i = 0; i <= I40E_VFQF_HKEY_MAX_INDEX; i++) {
+                               reg = I40E_VFQF_HKEY1(i, vsi->user_param);
+                               key_dw[i] = i40e_read_rx_ctl(hw, reg);
+                       }
+                       *key_len = (I40E_VFQF_HKEY_MAX_INDEX + 1) *
+                                  sizeof(uint32_t);
+               } else {
+                       for (i = 0; i <= I40E_PFQF_HKEY_MAX_INDEX; i++) {
+                               reg = I40E_PFQF_HKEY(i);
+                               key_dw[i] = i40e_read_rx_ctl(hw, reg);
+                       }
+                       *key_len = (I40E_PFQF_HKEY_MAX_INDEX + 1) *
+                                  sizeof(uint32_t);
+               }
        }
-       *key_len = (I40E_PFQF_HKEY_MAX_INDEX + 1) * sizeof(uint32_t);
-
        return 0;
 }
 
index d598730..84a5133 100644 (file)
@@ -399,6 +399,9 @@ struct i40e_pf_vf {
        uint16_t lan_nb_qps; /* Actual queues allocated */
        uint16_t reset_cnt; /* Total vf reset times */
        struct ether_addr mac_addr;  /* Default MAC address */
+       /* version of the virtchnl from VF */
+       struct virtchnl_version_info version;
+       uint32_t request_caps; /* offload caps requested from VF */
 };
 
 /*
@@ -1172,6 +1175,8 @@ int i40e_flush_queue_region_all_conf(struct rte_eth_dev *dev,
                struct i40e_hw *hw, struct i40e_pf *pf, uint16_t on);
 void i40e_init_queue_region_conf(struct rte_eth_dev *dev);
 void i40e_flex_payload_reg_set_default(struct i40e_hw *hw);
+int i40e_set_rss_key(struct i40e_vsi *vsi, uint8_t *key, uint8_t key_len);
+int i40e_set_rss_lut(struct i40e_vsi *vsi, uint8_t *lut, uint16_t lut_size);
 
 #define I40E_DEV_TO_PCI(eth_dev) \
        RTE_DEV_TO_PCI((eth_dev)->device)
index 93b484b..ad03903 100644 (file)
@@ -244,19 +244,23 @@ i40e_pf_host_send_msg_to_vf(struct i40e_pf_vf *vf,
 }
 
 static void
-i40e_pf_host_process_cmd_version(struct i40e_pf_vf *vf, bool b_op)
+i40e_pf_host_process_cmd_version(struct i40e_pf_vf *vf, uint8_t *msg,
+                                bool b_op)
 {
        struct virtchnl_version_info info;
 
-       /* Respond like a Linux PF host in order to support both DPDK VF and
-        * Linux VF driver. The expense is original DPDK host specific feature
+       /* VF and PF drivers need to follow the Virtchnl definition, No matter
+        * it's DPDK or other kernel drivers.
+        * The original DPDK host specific feature
         * like CFG_VLAN_PVID and CONFIG_VSI_QUEUES_EXT will not available.
-        *
-        * DPDK VF also can't identify host driver by version number returned.
-        * It always assume talking with Linux PF.
         */
+
        info.major = VIRTCHNL_VERSION_MAJOR;
-       info.minor = VIRTCHNL_VERSION_MINOR_NO_VF_CAPS;
+       vf->version = *(struct virtchnl_version_info *)msg;
+       if (VF_IS_V10(&vf->version))
+               info.minor = VIRTCHNL_VERSION_MINOR_NO_VF_CAPS;
+       else
+               info.minor = VIRTCHNL_VERSION_MINOR;
 
        if (b_op)
                i40e_pf_host_send_msg_to_vf(vf, VIRTCHNL_OP_VERSION,
@@ -280,11 +284,13 @@ i40e_pf_host_process_cmd_reset_vf(struct i40e_pf_vf *vf)
 }
 
 static int
-i40e_pf_host_process_cmd_get_vf_resource(struct i40e_pf_vf *vf, bool b_op)
+i40e_pf_host_process_cmd_get_vf_resource(struct i40e_pf_vf *vf, uint8_t *msg,
+                                        bool b_op)
 {
        struct virtchnl_vf_resource *vf_res = NULL;
        struct i40e_hw *hw = I40E_PF_TO_HW(vf->pf);
        uint32_t len = 0;
+       uint64_t default_hena = I40E_RSS_HENA_ALL;
        int ret = I40E_SUCCESS;
 
        if (!b_op) {
@@ -308,11 +314,35 @@ i40e_pf_host_process_cmd_get_vf_resource(struct i40e_pf_vf *vf, bool b_op)
                goto send_msg;
        }
 
-       vf_res->vf_cap_flags = VIRTCHNL_VF_OFFLOAD_L2 |
-                              VIRTCHNL_VF_OFFLOAD_VLAN;
+       if (VF_IS_V10(&vf->version)) /* doesn't support offload negotiate */
+               vf->request_caps = VIRTCHNL_VF_OFFLOAD_L2 |
+                                  VIRTCHNL_VF_OFFLOAD_VLAN;
+       else
+               vf->request_caps = *(uint32_t *)msg;
+
+       /* enable all RSS by default,
+        * doesn't support hena setting by virtchnnl yet.
+        */
+       if (vf->request_caps & VIRTCHNL_VF_OFFLOAD_RSS_PF) {
+               I40E_WRITE_REG(hw, I40E_VFQF_HENA1(0, vf->vf_idx),
+                              (uint32_t)default_hena);
+               I40E_WRITE_REG(hw, I40E_VFQF_HENA1(1, vf->vf_idx),
+                              (uint32_t)(default_hena >> 32));
+               I40E_WRITE_FLUSH(hw);
+       }
+
+       vf_res->vf_cap_flags = vf->request_caps &
+                                  I40E_VIRTCHNL_OFFLOAD_CAPS;
+       /* For X722, it supports write back on ITR
+        * without binding queue to interrupt vector.
+        */
+       if (hw->mac.type == I40E_MAC_X722)
+               vf_res->vf_cap_flags |= VIRTCHNL_VF_OFFLOAD_WB_ON_ITR;
        vf_res->max_vectors = hw->func_caps.num_msix_vectors_vf;
        vf_res->num_queue_pairs = vf->vsi->nb_qps;
        vf_res->num_vsis = I40E_DEFAULT_VF_VSI_NUM;
+       vf_res->rss_key_size = (I40E_PFQF_HKEY_MAX_INDEX + 1) * 4;
+       vf_res->rss_lut_size = (I40E_VFQF_HLUT1_MAX_INDEX + 1) * 4;
 
        /* Change below setting if PF host can support more VSIs for VF */
        vf_res->vsi_res[0].vsi_type = VIRTCHNL_VSI_SRIOV;
@@ -1061,6 +1091,84 @@ i40e_pf_host_process_cmd_disable_vlan_strip(struct i40e_pf_vf *vf, bool b_op)
        return ret;
 }
 
+static int
+i40e_pf_host_process_cmd_set_rss_lut(struct i40e_pf_vf *vf,
+                                    uint8_t *msg,
+                                    uint16_t msglen,
+                                    bool b_op)
+{
+       struct virtchnl_rss_lut *rss_lut = (struct virtchnl_rss_lut *)msg;
+       uint16_t valid_len;
+       int ret = I40E_SUCCESS;
+
+       if (!b_op) {
+               i40e_pf_host_send_msg_to_vf(
+                       vf,
+                       VIRTCHNL_OP_CONFIG_RSS_LUT,
+                       I40E_NOT_SUPPORTED, NULL, 0);
+               return ret;
+       }
+
+       if (!msg || msglen <= sizeof(struct virtchnl_rss_lut)) {
+               PMD_DRV_LOG(ERR, "set_rss_lut argument too short");
+               ret = I40E_ERR_PARAM;
+               goto send_msg;
+       }
+       valid_len = sizeof(struct virtchnl_rss_lut) + rss_lut->lut_entries - 1;
+       if (msglen < valid_len) {
+               PMD_DRV_LOG(ERR, "set_rss_lut length mismatch");
+               ret = I40E_ERR_PARAM;
+               goto send_msg;
+       }
+
+       ret = i40e_set_rss_lut(vf->vsi, rss_lut->lut, rss_lut->lut_entries);
+
+send_msg:
+       i40e_pf_host_send_msg_to_vf(vf, VIRTCHNL_OP_CONFIG_RSS_LUT,
+                                   ret, NULL, 0);
+
+       return ret;
+}
+
+static int
+i40e_pf_host_process_cmd_set_rss_key(struct i40e_pf_vf *vf,
+                                    uint8_t *msg,
+                                    uint16_t msglen,
+                                    bool b_op)
+{
+       struct virtchnl_rss_key *rss_key = (struct virtchnl_rss_key *)msg;
+       uint16_t valid_len;
+       int ret = I40E_SUCCESS;
+
+       if (!b_op) {
+               i40e_pf_host_send_msg_to_vf(
+                       vf,
+                       VIRTCHNL_OP_DEL_VLAN,
+                       VIRTCHNL_OP_CONFIG_RSS_KEY, NULL, 0);
+               return ret;
+       }
+
+       if (!msg || msglen <= sizeof(struct virtchnl_rss_key)) {
+               PMD_DRV_LOG(ERR, "set_rss_key argument too short");
+               ret = I40E_ERR_PARAM;
+               goto send_msg;
+       }
+       valid_len = sizeof(struct virtchnl_rss_key) + rss_key->key_len - 1;
+       if (msglen < valid_len) {
+               PMD_DRV_LOG(ERR, "set_rss_key length mismatch");
+               ret = I40E_ERR_PARAM;
+               goto send_msg;
+       }
+
+       ret = i40e_set_rss_key(vf->vsi, rss_key->key, rss_key->key_len);
+
+send_msg:
+       i40e_pf_host_send_msg_to_vf(vf, VIRTCHNL_OP_CONFIG_RSS_KEY,
+                                   ret, NULL, 0);
+
+       return ret;
+}
+
 void
 i40e_notify_vf_link_status(struct rte_eth_dev *dev, struct i40e_pf_vf *vf)
 {
@@ -1167,7 +1275,7 @@ i40e_pf_host_handle_vf_msg(struct rte_eth_dev *dev,
        switch (opcode) {
        case VIRTCHNL_OP_VERSION:
                PMD_DRV_LOG(INFO, "OP_VERSION received");
-               i40e_pf_host_process_cmd_version(vf, b_op);
+               i40e_pf_host_process_cmd_version(vf, msg, b_op);
                break;
        case VIRTCHNL_OP_RESET_VF:
                PMD_DRV_LOG(INFO, "OP_RESET_VF received");
@@ -1175,7 +1283,7 @@ i40e_pf_host_handle_vf_msg(struct rte_eth_dev *dev,
                break;
        case VIRTCHNL_OP_GET_VF_RESOURCES:
                PMD_DRV_LOG(INFO, "OP_GET_VF_RESOURCES received");
-               i40e_pf_host_process_cmd_get_vf_resource(vf, b_op);
+               i40e_pf_host_process_cmd_get_vf_resource(vf, msg, b_op);
                break;
        case VIRTCHNL_OP_CONFIG_VSI_QUEUES:
                PMD_DRV_LOG(INFO, "OP_CONFIG_VSI_QUEUES received");
@@ -1236,6 +1344,14 @@ i40e_pf_host_handle_vf_msg(struct rte_eth_dev *dev,
                PMD_DRV_LOG(INFO, "OP_DISABLE_VLAN_STRIPPING received");
                i40e_pf_host_process_cmd_disable_vlan_strip(vf, b_op);
                break;
+       case VIRTCHNL_OP_CONFIG_RSS_LUT:
+               PMD_DRV_LOG(INFO, "OP_CONFIG_RSS_LUT received");
+               i40e_pf_host_process_cmd_set_rss_lut(vf, msg, msglen, b_op);
+               break;
+       case VIRTCHNL_OP_CONFIG_RSS_KEY:
+               PMD_DRV_LOG(INFO, "OP_CONFIG_RSS_KEY received");
+               i40e_pf_host_process_cmd_set_rss_key(vf, msg, msglen, b_op);
+               break;
        /* Don't add command supported below, which will
         * return an error code.
         */
index 429f347..1809ba4 100644 (file)
@@ -8,6 +8,12 @@
 /* Default setting on number of VSIs that VF can contain */
 #define I40E_DEFAULT_VF_VSI_NUM 1
 
+#define I40E_VIRTCHNL_OFFLOAD_CAPS ( \
+       VIRTCHNL_VF_OFFLOAD_L2 | \
+       VIRTCHNL_VF_OFFLOAD_VLAN | \
+       VIRTCHNL_VF_OFFLOAD_RSS_PF | \
+       VIRTCHNL_VF_OFFLOAD_RX_POLLING)
+
 struct virtchnl_vlan_offload_info {
        uint16_t vsi_id;
        uint8_t enable_vlan_strip;