net/qede: fix multicast filtering
authorHarish Patil <harish.patil@cavium.com>
Sun, 20 May 2018 00:15:43 +0000 (17:15 -0700)
committerFerruh Yigit <ferruh.yigit@intel.com>
Tue, 22 May 2018 22:35:00 +0000 (00:35 +0200)
This patch is to fix multicast filtering using set_mc_addr_list().

Fixes: 77fac1b54fc9 ("net/qede: fix filtering code")
Cc: stable@dpdk.org
Signed-off-by: Harish Patil <harish.patil@cavium.com>
Signed-off-by: Shahed Shaikh <shahed.shaikh@cavium.com>
Signed-off-by: Rasesh Mody <rasesh.mody@cavium.com>
drivers/net/qede/base/ecore_l2.c
drivers/net/qede/base/ecore_l2_api.h
drivers/net/qede/base/ecore_sriov.c
drivers/net/qede/base/ecore_vf.c
drivers/net/qede/base/ecore_vfpf_if.h
drivers/net/qede/qede_ethdev.c

index c897fa5..fdb928a 100644 (file)
@@ -690,7 +690,7 @@ ecore_sp_update_mcast_bin(struct vport_update_ramrod_data *p_ramrod,
 
        p_ramrod->common.update_approx_mcast_flg = 1;
        for (i = 0; i < ETH_MULTICAST_MAC_BINS_IN_REGS; i++) {
-               u32 *p_bins = (u32 *)p_params->bins;
+               u32 *p_bins = p_params->bins;
 
                p_ramrod->approx_mcast.bins[i] = OSAL_CPU_TO_LE32(p_bins[i]);
        }
@@ -1568,8 +1568,8 @@ ecore_sp_eth_filter_mcast(struct ecore_hwfn *p_hwfn,
                          enum spq_mode comp_mode,
                          struct ecore_spq_comp_cb *p_comp_data)
 {
-       unsigned long bins[ETH_MULTICAST_MAC_BINS_IN_REGS];
        struct vport_update_ramrod_data *p_ramrod = OSAL_NULL;
+       u32 bins[ETH_MULTICAST_MAC_BINS_IN_REGS];
        struct ecore_spq_entry *p_ent = OSAL_NULL;
        struct ecore_sp_init_data init_data;
        u8 abs_vport_id = 0;
@@ -1608,8 +1608,7 @@ ecore_sp_eth_filter_mcast(struct ecore_hwfn *p_hwfn,
        /* explicitly clear out the entire vector */
        OSAL_MEMSET(&p_ramrod->approx_mcast.bins,
                    0, sizeof(p_ramrod->approx_mcast.bins));
-       OSAL_MEMSET(bins, 0, sizeof(unsigned long) *
-                   ETH_MULTICAST_MAC_BINS_IN_REGS);
+       OSAL_MEMSET(bins, 0, sizeof(u32) * ETH_MULTICAST_MAC_BINS_IN_REGS);
        /* filter ADD op is explicit set op and it removes
        *  any existing filters for the vport.
        */
@@ -1618,16 +1617,15 @@ ecore_sp_eth_filter_mcast(struct ecore_hwfn *p_hwfn,
                        u32 bit;
 
                        bit = ecore_mcast_bin_from_mac(p_filter_cmd->mac[i]);
-                       OSAL_SET_BIT(bit, bins);
+                       bins[bit / 32] |= 1 << (bit % 32);
                }
 
                /* Convert to correct endianity */
                for (i = 0; i < ETH_MULTICAST_MAC_BINS_IN_REGS; i++) {
                        struct vport_update_ramrod_mcast *p_ramrod_bins;
-                       u32 *p_bins = (u32 *)bins;
 
                        p_ramrod_bins = &p_ramrod->approx_mcast;
-                       p_ramrod_bins->bins[i] = OSAL_CPU_TO_LE32(p_bins[i]);
+                       p_ramrod_bins->bins[i] = OSAL_CPU_TO_LE32(bins[i]);
                }
        }
 
index ed9837b..8cbe8dd 100644 (file)
@@ -332,7 +332,7 @@ struct ecore_sp_vport_update_params {
        u8                      anti_spoofing_en;
        u8                      update_accept_any_vlan_flg;
        u8                      accept_any_vlan;
-       unsigned long           bins[8];
+       u32                     bins[8];
        struct ecore_rss_params *rss_params;
        struct ecore_filter_accept_flags accept_flags;
        struct ecore_sge_tpa_params *sge_tpa_params;
index 0279709..93674a2 100644 (file)
@@ -2979,8 +2979,7 @@ ecore_iov_vp_update_mcast_bin_param(struct ecore_hwfn *p_hwfn,
 
        p_data->update_approx_mcast_flg = 1;
        OSAL_MEMCPY(p_data->bins, p_mcast_tlv->bins,
-                   sizeof(unsigned long) *
-                   ETH_MULTICAST_MAC_BINS_IN_REGS);
+                   sizeof(u32) * ETH_MULTICAST_MAC_BINS_IN_REGS);
        *tlvs_mask |= 1 << ECORE_IOV_VP_UPDATE_MCAST;
 }
 
index e0f2dd5..8a08911 100644 (file)
@@ -1275,8 +1275,7 @@ ecore_vf_pf_vport_update(struct ecore_hwfn *p_hwfn,
                resp_size += sizeof(struct pfvf_def_resp_tlv);
 
                OSAL_MEMCPY(p_mcast_tlv->bins, p_params->bins,
-                           sizeof(unsigned long) *
-                           ETH_MULTICAST_MAC_BINS_IN_REGS);
+                           sizeof(u32) * ETH_MULTICAST_MAC_BINS_IN_REGS);
        }
 
        update_rx = p_params->accept_flags.update_rx_mode_config;
@@ -1473,7 +1472,7 @@ void ecore_vf_pf_filter_mcast(struct ecore_hwfn *p_hwfn,
                        u32 bit;
 
                        bit = ecore_mcast_bin_from_mac(p_filter_cmd->mac[i]);
-                       OSAL_SET_BIT(bit, sp_params.bins);
+                       sp_params.bins[bit / 32] |= 1 << (bit % 32);
                }
        }
 
index ecb0064..c6af9ca 100644 (file)
@@ -396,7 +396,13 @@ struct vfpf_vport_update_mcast_bin_tlv {
        struct channel_tlv      tl;
        u8                      padding[4];
 
-       u64             bins[8];
+       /* This was a mistake; There are only 256 approx bins,
+        * and in HSI they're divided into 32-bit values.
+        * As old VFs used to set-bit to the values on its side,
+        * the upper half of the array is never expected to contain any data.
+        */
+       u64             bins[4];
+       u64             obsolete_bins[4];
 };
 
 struct vfpf_vport_update_accept_param_tlv {
index 50a63be..3e1a62c 100644 (file)
@@ -894,47 +894,69 @@ qede_ucast_filter(struct rte_eth_dev *eth_dev, struct ecore_filter_ucast *ucast,
 }
 
 static int
-qede_mcast_filter(struct rte_eth_dev *eth_dev, struct ecore_filter_ucast *mcast,
-                 bool add)
+qede_add_mcast_filters(struct rte_eth_dev *eth_dev, struct ether_addr *mc_addrs,
+                      uint32_t mc_addrs_num)
 {
        struct qede_dev *qdev = QEDE_INIT_QDEV(eth_dev);
        struct ecore_dev *edev = QEDE_INIT_EDEV(qdev);
-       struct ether_addr *mac_addr;
-       struct qede_mcast_entry *tmp = NULL;
-       struct qede_mcast_entry *m;
+       struct ecore_filter_mcast mcast;
+       struct qede_mcast_entry *m = NULL;
+       uint8_t i;
+       int rc;
 
-       mac_addr  = (struct ether_addr *)mcast->mac;
-       if (add) {
-               SLIST_FOREACH(tmp, &qdev->mc_list_head, list) {
-                       if (memcmp(mac_addr, &tmp->mac, ETHER_ADDR_LEN) == 0) {
-                               DP_ERR(edev,
-                                       "Multicast MAC is already added\n");
-                               return -EEXIST;
-                       }
-               }
+       for (i = 0; i < mc_addrs_num; i++) {
                m = rte_malloc(NULL, sizeof(struct qede_mcast_entry),
-                       RTE_CACHE_LINE_SIZE);
+                              RTE_CACHE_LINE_SIZE);
                if (!m) {
-                       DP_ERR(edev,
-                               "Did not allocate memory for mcast\n");
+                       DP_ERR(edev, "Did not allocate memory for mcast\n");
                        return -ENOMEM;
                }
-               ether_addr_copy(mac_addr, &m->mac);
+               ether_addr_copy(&mc_addrs[i], &m->mac);
                SLIST_INSERT_HEAD(&qdev->mc_list_head, m, list);
-               qdev->num_mc_addr++;
-       } else {
-               SLIST_FOREACH(tmp, &qdev->mc_list_head, list) {
-                       if (memcmp(mac_addr, &tmp->mac, ETHER_ADDR_LEN) == 0)
-                               break;
-               }
-               if (tmp == NULL) {
-                       DP_INFO(edev, "Multicast mac is not found\n");
-                       return -EINVAL;
-               }
-               SLIST_REMOVE(&qdev->mc_list_head, tmp,
-                            qede_mcast_entry, list);
-               qdev->num_mc_addr--;
        }
+       memset(&mcast, 0, sizeof(mcast));
+       mcast.num_mc_addrs = mc_addrs_num;
+       mcast.opcode = ECORE_FILTER_ADD;
+       for (i = 0; i < mc_addrs_num; i++)
+               ether_addr_copy(&mc_addrs[i], (struct ether_addr *)
+                                                       &mcast.mac[i]);
+       rc = ecore_filter_mcast_cmd(edev, &mcast, ECORE_SPQ_MODE_CB, NULL);
+       if (rc != ECORE_SUCCESS) {
+               DP_ERR(edev, "Failed to add multicast filter (rc = %d\n)", rc);
+               return -1;
+       }
+
+       return 0;
+}
+
+static int qede_del_mcast_filters(struct rte_eth_dev *eth_dev)
+{
+       struct qede_dev *qdev = QEDE_INIT_QDEV(eth_dev);
+       struct ecore_dev *edev = QEDE_INIT_EDEV(qdev);
+       struct qede_mcast_entry *tmp = NULL;
+       struct ecore_filter_mcast mcast;
+       int j;
+       int rc;
+
+       memset(&mcast, 0, sizeof(mcast));
+       mcast.num_mc_addrs = qdev->num_mc_addr;
+       mcast.opcode = ECORE_FILTER_REMOVE;
+       j = 0;
+       SLIST_FOREACH(tmp, &qdev->mc_list_head, list) {
+               ether_addr_copy(&tmp->mac, (struct ether_addr *)&mcast.mac[j]);
+               j++;
+       }
+       rc = ecore_filter_mcast_cmd(edev, &mcast, ECORE_SPQ_MODE_CB, NULL);
+       if (rc != ECORE_SUCCESS) {
+               DP_ERR(edev, "Failed to delete multicast filter\n");
+               return -1;
+       }
+       /* Init the list */
+       while (!SLIST_EMPTY(&qdev->mc_list_head)) {
+               tmp = SLIST_FIRST(&qdev->mc_list_head);
+               SLIST_REMOVE_HEAD(&qdev->mc_list_head, list);
+       }
+       SLIST_INIT(&qdev->mc_list_head);
 
        return 0;
 }
@@ -945,59 +967,22 @@ qede_mac_int_ops(struct rte_eth_dev *eth_dev, struct ecore_filter_ucast *ucast,
 {
        struct qede_dev *qdev = QEDE_INIT_QDEV(eth_dev);
        struct ecore_dev *edev = QEDE_INIT_EDEV(qdev);
-       enum _ecore_status_t rc;
-       struct ecore_filter_mcast mcast;
-       struct qede_mcast_entry *tmp;
-       uint16_t j = 0;
+       enum _ecore_status_t rc = ECORE_INVAL;
 
-       /* Multicast */
-       if (is_multicast_ether_addr((struct ether_addr *)ucast->mac)) {
-               if (add) {
-                       if (qdev->num_mc_addr >= ECORE_MAX_MC_ADDRS) {
-                               DP_ERR(edev,
-                                      "Mcast filter table limit exceeded, "
-                                      "Please enable mcast promisc mode\n");
-                               return -ECORE_INVAL;
-                       }
-               }
-               rc = qede_mcast_filter(eth_dev, ucast, add);
-               if (rc == 0) {
-                       DP_INFO(edev, "num_mc_addrs = %u\n", qdev->num_mc_addr);
-                       memset(&mcast, 0, sizeof(mcast));
-                       mcast.num_mc_addrs = qdev->num_mc_addr;
-                       mcast.opcode = ECORE_FILTER_ADD;
-                       SLIST_FOREACH(tmp, &qdev->mc_list_head, list) {
-                               ether_addr_copy(&tmp->mac,
-                                       (struct ether_addr *)&mcast.mac[j]);
-                               j++;
-                       }
-                       rc = ecore_filter_mcast_cmd(edev, &mcast,
-                                                   ECORE_SPQ_MODE_CB, NULL);
-               }
-               if (rc != ECORE_SUCCESS) {
-                       DP_ERR(edev, "Failed to add multicast filter"
-                              " rc = %d, op = %d\n", rc, add);
-               }
-       } else { /* Unicast */
-               if (add) {
-                       if (qdev->num_uc_addr >=
-                           qdev->dev_info.num_mac_filters) {
-                               DP_ERR(edev,
-                                      "Ucast filter table limit exceeded,"
-                                      " Please enable promisc mode\n");
-                               return -ECORE_INVAL;
-                       }
-               }
-               rc = qede_ucast_filter(eth_dev, ucast, add);
-               if (rc == 0)
-                       rc = ecore_filter_ucast_cmd(edev, ucast,
-                                                   ECORE_SPQ_MODE_CB, NULL);
-               if (rc != ECORE_SUCCESS) {
-                       DP_ERR(edev, "MAC filter failed, rc = %d, op = %d\n",
-                              rc, add);
-               }
+       if (add && (qdev->num_uc_addr >= qdev->dev_info.num_mac_filters)) {
+               DP_ERR(edev, "Ucast filter table limit exceeded,"
+                             " Please enable promisc mode\n");
+                       return ECORE_INVAL;
        }
 
+       rc = qede_ucast_filter(eth_dev, ucast, add);
+       if (rc == 0)
+               rc = ecore_filter_ucast_cmd(edev, ucast,
+                                           ECORE_SPQ_MODE_CB, NULL);
+       if (rc != ECORE_SUCCESS)
+               DP_ERR(edev, "MAC filter failed, rc = %d, op = %d\n",
+                      rc, add);
+
        return rc;
 }
 
@@ -2042,6 +2027,35 @@ static void qede_allmulticast_disable(struct rte_eth_dev *eth_dev)
                                QED_FILTER_RX_MODE_TYPE_REGULAR);
 }
 
+static int
+qede_set_mc_addr_list(struct rte_eth_dev *eth_dev, struct ether_addr *mc_addrs,
+                     uint32_t mc_addrs_num)
+{
+       struct qede_dev *qdev = QEDE_INIT_QDEV(eth_dev);
+       struct ecore_dev *edev = QEDE_INIT_EDEV(qdev);
+       uint8_t i;
+
+       if (mc_addrs_num > ECORE_MAX_MC_ADDRS) {
+               DP_ERR(edev, "Reached max multicast filters limit,"
+                            "Please enable multicast promisc mode\n");
+               return -ENOSPC;
+       }
+
+       for (i = 0; i < mc_addrs_num; i++) {
+               if (!is_multicast_ether_addr(&mc_addrs[i])) {
+                       DP_ERR(edev, "Not a valid multicast MAC\n");
+                       return -EINVAL;
+               }
+       }
+
+       /* Flush all existing entries */
+       if (qede_del_mcast_filters(eth_dev))
+               return -1;
+
+       /* Set new mcast list */
+       return qede_add_mcast_filters(eth_dev, mc_addrs, mc_addrs_num);
+}
+
 static int qede_flow_ctrl_set(struct rte_eth_dev *eth_dev,
                              struct rte_eth_fc_conf *fc_conf)
 {
@@ -2926,6 +2940,7 @@ static const struct eth_dev_ops qede_eth_dev_ops = {
        .promiscuous_disable = qede_promiscuous_disable,
        .allmulticast_enable = qede_allmulticast_enable,
        .allmulticast_disable = qede_allmulticast_disable,
+       .set_mc_addr_list = qede_set_mc_addr_list,
        .dev_stop = qede_dev_stop,
        .dev_close = qede_dev_close,
        .stats_get = qede_get_stats,
@@ -2966,6 +2981,7 @@ static const struct eth_dev_ops qede_eth_vf_dev_ops = {
        .promiscuous_disable = qede_promiscuous_disable,
        .allmulticast_enable = qede_allmulticast_enable,
        .allmulticast_disable = qede_allmulticast_disable,
+       .set_mc_addr_list = qede_set_mc_addr_list,
        .dev_stop = qede_dev_stop,
        .dev_close = qede_dev_close,
        .stats_get = qede_get_stats,
@@ -3174,6 +3190,7 @@ static int qede_common_dev_init(struct rte_eth_dev *eth_dev, bool is_vf)
        SLIST_INIT(&adapter->fdir_info.fdir_list_head);
        SLIST_INIT(&adapter->vlan_list_head);
        SLIST_INIT(&adapter->uc_list_head);
+       SLIST_INIT(&adapter->mc_list_head);
        adapter->mtu = ETHER_MTU;
        adapter->vport_started = false;