qed*: Support PVID configuration
authorYuval Mintz <Yuval.Mintz@qlogic.com>
Wed, 11 May 2016 13:36:20 +0000 (16:36 +0300)
committerDavid S. Miller <davem@davemloft.net>
Thu, 12 May 2016 04:04:07 +0000 (00:04 -0400)
This adds support for PF control over the VF vlan configuration.
I.e., `ip link ... vf <x> vlan <vid>' should now be supported.

 1. <vid> != 0 => VF receives [unknowingly] only traffic tagged by
    <vid> and tags all outgoing traffic sent by VF with <vid>.
 2. <vid> == 0 ==> Remove the pvid configuration, reverting to previous.

Signed-off-by: Yuval Mintz <Yuval.Mintz@qlogic.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
drivers/net/ethernet/qlogic/qed/qed_dev.c
drivers/net/ethernet/qlogic/qed/qed_l2.c
drivers/net/ethernet/qlogic/qed/qed_l2.h
drivers/net/ethernet/qlogic/qed/qed_sriov.c
drivers/net/ethernet/qlogic/qed/qed_sriov.h
drivers/net/ethernet/qlogic/qed/qed_vf.c
drivers/net/ethernet/qlogic/qed/qed_vf.h
drivers/net/ethernet/qlogic/qede/qede_main.c
include/linux/qed/qed_iov_if.h

index 9d01a16..e75e73a 100644 (file)
@@ -1104,9 +1104,16 @@ static void qed_hw_get_resc(struct qed_hwfn *p_hwfn)
        u8 num_funcs = p_hwfn->num_funcs_on_engine;
        u32 *resc_num = p_hwfn->hw_info.resc_num;
        struct qed_sb_cnt_info sb_cnt_info;
-       int i;
+       int i, max_vf_vlan_filters;
 
        memset(&sb_cnt_info, 0, sizeof(sb_cnt_info));
+
+#ifdef CONFIG_QED_SRIOV
+       max_vf_vlan_filters = QED_ETH_MAX_VF_NUM_VLAN_FILTERS;
+#else
+       max_vf_vlan_filters = 0;
+#endif
+
        qed_int_get_num_sbs(p_hwfn, &sb_cnt_info);
 
        resc_num[QED_SB] = min_t(u32,
index 80f0b85..7fb6b82 100644 (file)
@@ -114,7 +114,8 @@ int qed_sp_vport_start(struct qed_hwfn *p_hwfn,
                                             p_params->mtu,
                                             p_params->remove_inner_vlan,
                                             p_params->tpa_mode,
-                                            p_params->max_buffers_per_cqe);
+                                            p_params->max_buffers_per_cqe,
+                                            p_params->only_untagged);
        }
 
        return qed_sp_eth_vport_start(p_hwfn, p_params);
@@ -367,6 +368,16 @@ int qed_sp_vport_update(struct qed_hwfn *p_hwfn,
        p_cmn->inner_vlan_removal_en = p_params->inner_vlan_removal_flg;
        val = p_params->update_inner_vlan_removal_flg;
        p_cmn->update_inner_vlan_removal_en_flg = val;
+
+       p_cmn->default_vlan_en = p_params->default_vlan_enable_flg;
+       val = p_params->update_default_vlan_enable_flg;
+       p_cmn->update_default_vlan_en_flg = val;
+
+       p_cmn->default_vlan = cpu_to_le16(p_params->default_vlan);
+       p_cmn->update_default_vlan_flg = p_params->update_default_vlan_flg;
+
+       p_cmn->silent_vlan_removal_en = p_params->silent_vlan_removal_flg;
+
        p_ramrod->common.tx_switching_en = p_params->tx_switching_flg;
        p_cmn->update_tx_switching_en_flg = p_params->update_tx_switching_flg;
 
@@ -1702,6 +1713,7 @@ static int qed_start_vport(struct qed_dev *cdev,
                start.tpa_mode = params->gro_enable ? QED_TPA_MODE_GRO :
                                                        QED_TPA_MODE_NONE;
                start.remove_inner_vlan = params->remove_inner_vlan;
+               start.only_untagged = true;     /* untagged only */
                start.drop_ttl0 = params->drop_ttl0;
                start.opaque_fid = p_hwfn->hw_info.opaque_fid;
                start.concrete_fid = p_hwfn->hw_info.concrete_fid;
index f9e677a..fad30ae 100644 (file)
@@ -94,6 +94,7 @@ enum qed_tpa_mode {
 struct qed_sp_vport_start_params {
        enum qed_tpa_mode tpa_mode;
        bool remove_inner_vlan;
+       bool only_untagged;
        bool drop_ttl0;
        u8 max_buffers_per_cqe;
        u32 concrete_fid;
@@ -140,6 +141,11 @@ struct qed_sp_vport_update_params {
        u8                              vport_active_tx_flg;
        u8                              update_inner_vlan_removal_flg;
        u8                              inner_vlan_removal_flg;
+       u8                              silent_vlan_removal_flg;
+       u8                              update_default_vlan_enable_flg;
+       u8                              default_vlan_enable_flg;
+       u8                              update_default_vlan_flg;
+       u16                             default_vlan;
        u8                              update_tx_switching_flg;
        u8                              tx_switching_flg;
        u8                              update_approx_mcast_flg;
index 29a53dd..77d44ba 100644 (file)
@@ -1075,6 +1075,7 @@ static void qed_iov_vf_cleanup(struct qed_hwfn *p_hwfn,
        p_vf->vport_instance = 0;
        p_vf->num_mac_filters = 0;
        p_vf->num_vlan_filters = 0;
+       p_vf->configured_features = 0;
 
        /* If VF previously requested less resources, go back to default */
        p_vf->num_rxqs = p_vf->num_sbs;
@@ -1085,6 +1086,7 @@ static void qed_iov_vf_cleanup(struct qed_hwfn *p_hwfn,
        for (i = 0; i < QED_MAX_VF_CHAINS_PER_PF; i++)
                p_vf->vf_queues[i].rxq_active = 0;
 
+       memset(&p_vf->shadow_config, 0, sizeof(p_vf->shadow_config));
        qed_iov_clean_vf(p_hwfn, p_vf->relative_vf_id);
 }
 
@@ -1232,6 +1234,149 @@ out:
                             sizeof(struct pfvf_acquire_resp_tlv), vfpf_status);
 }
 
+static int qed_iov_reconfigure_unicast_vlan(struct qed_hwfn *p_hwfn,
+                                           struct qed_vf_info *p_vf)
+{
+       struct qed_filter_ucast filter;
+       int rc = 0;
+       int i;
+
+       memset(&filter, 0, sizeof(filter));
+       filter.is_rx_filter = 1;
+       filter.is_tx_filter = 1;
+       filter.vport_to_add_to = p_vf->vport_id;
+       filter.opcode = QED_FILTER_ADD;
+
+       /* Reconfigure vlans */
+       for (i = 0; i < QED_ETH_VF_NUM_VLAN_FILTERS + 1; i++) {
+               if (!p_vf->shadow_config.vlans[i].used)
+                       continue;
+
+               filter.type = QED_FILTER_VLAN;
+               filter.vlan = p_vf->shadow_config.vlans[i].vid;
+               DP_VERBOSE(p_hwfn,
+                          QED_MSG_IOV,
+                          "Reconfiguring VLAN [0x%04x] for VF [%04x]\n",
+                          filter.vlan, p_vf->relative_vf_id);
+               rc = qed_sp_eth_filter_ucast(p_hwfn,
+                                            p_vf->opaque_fid,
+                                            &filter,
+                                            QED_SPQ_MODE_CB, NULL);
+               if (rc) {
+                       DP_NOTICE(p_hwfn,
+                                 "Failed to configure VLAN [%04x] to VF [%04x]\n",
+                                 filter.vlan, p_vf->relative_vf_id);
+                       break;
+               }
+       }
+
+       return rc;
+}
+
+static int
+qed_iov_reconfigure_unicast_shadow(struct qed_hwfn *p_hwfn,
+                                  struct qed_vf_info *p_vf, u64 events)
+{
+       int rc = 0;
+
+       if ((events & (1 << VLAN_ADDR_FORCED)) &&
+           !(p_vf->configured_features & (1 << VLAN_ADDR_FORCED)))
+               rc = qed_iov_reconfigure_unicast_vlan(p_hwfn, p_vf);
+
+       return rc;
+}
+
+static int qed_iov_configure_vport_forced(struct qed_hwfn *p_hwfn,
+                                         struct qed_vf_info *p_vf, u64 events)
+{
+       int rc = 0;
+       struct qed_filter_ucast filter;
+
+       if (!p_vf->vport_instance)
+               return -EINVAL;
+
+       if (events & (1 << VLAN_ADDR_FORCED)) {
+               struct qed_sp_vport_update_params vport_update;
+               u8 removal;
+               int i;
+
+               memset(&filter, 0, sizeof(filter));
+               filter.type = QED_FILTER_VLAN;
+               filter.is_rx_filter = 1;
+               filter.is_tx_filter = 1;
+               filter.vport_to_add_to = p_vf->vport_id;
+               filter.vlan = p_vf->bulletin.p_virt->pvid;
+               filter.opcode = filter.vlan ? QED_FILTER_REPLACE :
+                                             QED_FILTER_FLUSH;
+
+               /* Send the ramrod */
+               rc = qed_sp_eth_filter_ucast(p_hwfn, p_vf->opaque_fid,
+                                            &filter, QED_SPQ_MODE_CB, NULL);
+               if (rc) {
+                       DP_NOTICE(p_hwfn,
+                                 "PF failed to configure VLAN for VF\n");
+                       return rc;
+               }
+
+               /* Update the default-vlan & silent vlan stripping */
+               memset(&vport_update, 0, sizeof(vport_update));
+               vport_update.opaque_fid = p_vf->opaque_fid;
+               vport_update.vport_id = p_vf->vport_id;
+               vport_update.update_default_vlan_enable_flg = 1;
+               vport_update.default_vlan_enable_flg = filter.vlan ? 1 : 0;
+               vport_update.update_default_vlan_flg = 1;
+               vport_update.default_vlan = filter.vlan;
+
+               vport_update.update_inner_vlan_removal_flg = 1;
+               removal = filter.vlan ? 1
+                                     : p_vf->shadow_config.inner_vlan_removal;
+               vport_update.inner_vlan_removal_flg = removal;
+               vport_update.silent_vlan_removal_flg = filter.vlan ? 1 : 0;
+               rc = qed_sp_vport_update(p_hwfn,
+                                        &vport_update,
+                                        QED_SPQ_MODE_EBLOCK, NULL);
+               if (rc) {
+                       DP_NOTICE(p_hwfn,
+                                 "PF failed to configure VF vport for vlan\n");
+                       return rc;
+               }
+
+               /* Update all the Rx queues */
+               for (i = 0; i < QED_MAX_VF_CHAINS_PER_PF; i++) {
+                       u16 qid;
+
+                       if (!p_vf->vf_queues[i].rxq_active)
+                               continue;
+
+                       qid = p_vf->vf_queues[i].fw_rx_qid;
+
+                       rc = qed_sp_eth_rx_queues_update(p_hwfn, qid,
+                                                        1, 0, 1,
+                                                        QED_SPQ_MODE_EBLOCK,
+                                                        NULL);
+                       if (rc) {
+                               DP_NOTICE(p_hwfn,
+                                         "Failed to send Rx update fo queue[0x%04x]\n",
+                                         qid);
+                               return rc;
+                       }
+               }
+
+               if (filter.vlan)
+                       p_vf->configured_features |= 1 << VLAN_ADDR_FORCED;
+               else
+                       p_vf->configured_features &= ~(1 << VLAN_ADDR_FORCED);
+       }
+
+       /* If forced features are terminated, we need to configure the shadow
+        * configuration back again.
+        */
+       if (events)
+               qed_iov_reconfigure_unicast_shadow(p_hwfn, p_vf, events);
+
+       return rc;
+}
+
 static void qed_iov_vf_mbx_start_vport(struct qed_hwfn *p_hwfn,
                                       struct qed_ptt *p_ptt,
                                       struct qed_vf_info *vf)
@@ -1241,6 +1386,7 @@ static void qed_iov_vf_mbx_start_vport(struct qed_hwfn *p_hwfn,
        struct vfpf_vport_start_tlv *start;
        u8 status = PFVF_STATUS_SUCCESS;
        struct qed_vf_info *vf_info;
+       u64 *p_bitmap;
        int sb_id;
        int rc;
 
@@ -1272,10 +1418,24 @@ static void qed_iov_vf_mbx_start_vport(struct qed_hwfn *p_hwfn,
        qed_iov_enable_vf_traffic(p_hwfn, p_ptt, vf);
 
        vf->mtu = start->mtu;
+       vf->shadow_config.inner_vlan_removal = start->inner_vlan_removal;
+
+       /* Take into consideration configuration forced by hypervisor;
+        * If none is configured, use the supplied VF values [for old
+        * vfs that would still be fine, since they passed '0' as padding].
+        */
+       p_bitmap = &vf_info->bulletin.p_virt->valid_bitmap;
+       if (!(*p_bitmap & (1 << VFPF_BULLETIN_UNTAGGED_DEFAULT_FORCED))) {
+               u8 vf_req = start->only_untagged;
+
+               vf_info->bulletin.p_virt->default_only_untagged = vf_req;
+               *p_bitmap |= 1 << VFPF_BULLETIN_UNTAGGED_DEFAULT;
+       }
 
        params.tpa_mode = start->tpa_mode;
        params.remove_inner_vlan = start->inner_vlan_removal;
 
+       params.only_untagged = vf_info->bulletin.p_virt->default_only_untagged;
        params.drop_ttl0 = false;
        params.concrete_fid = vf->concrete_fid;
        params.opaque_fid = vf->opaque_fid;
@@ -1290,6 +1450,9 @@ static void qed_iov_vf_mbx_start_vport(struct qed_hwfn *p_hwfn,
                status = PFVF_STATUS_FAILURE;
        } else {
                vf->vport_instance++;
+
+               /* Force configuration if needed on the newly opened vport */
+               qed_iov_configure_vport_forced(p_hwfn, vf, *p_bitmap);
        }
        qed_iov_prepare_resp(p_hwfn, p_ptt, vf, CHANNEL_TLV_VPORT_START,
                             sizeof(struct pfvf_def_resp_tlv), status);
@@ -1311,6 +1474,10 @@ static void qed_iov_vf_mbx_stop_vport(struct qed_hwfn *p_hwfn,
                status = PFVF_STATUS_FAILURE;
        }
 
+       /* Forget the configuration on the vport */
+       vf->configured_features = 0;
+       memset(&vf->shadow_config, 0, sizeof(vf->shadow_config));
+
        qed_iov_prepare_resp(p_hwfn, p_ptt, vf, CHANNEL_TLV_VPORT_TEARDOWN,
                             sizeof(struct pfvf_def_resp_tlv), status);
 }
@@ -1634,8 +1801,13 @@ qed_iov_vp_update_vlan_param(struct qed_hwfn *p_hwfn,
        if (!p_vlan_tlv)
                return;
 
-       p_data->update_inner_vlan_removal_flg = 1;
-       p_data->inner_vlan_removal_flg = p_vlan_tlv->remove_vlan;
+       p_vf->shadow_config.inner_vlan_removal = p_vlan_tlv->remove_vlan;
+
+       /* Ignore the VF request if we're forcing a vlan */
+       if (!(p_vf->configured_features & (1 << VLAN_ADDR_FORCED))) {
+               p_data->update_inner_vlan_removal_flg = 1;
+               p_data->inner_vlan_removal_flg = p_vlan_tlv->remove_vlan;
+       }
 
        *tlvs_mask |= 1 << QED_IOV_VP_UPDATE_VLAN_STRIP;
 }
@@ -1886,6 +2058,67 @@ out:
        qed_iov_send_response(p_hwfn, p_ptt, vf, length, status);
 }
 
+static int qed_iov_vf_update_unicast_shadow(struct qed_hwfn *p_hwfn,
+                                           struct qed_vf_info *p_vf,
+                                           struct qed_filter_ucast *p_params)
+{
+       int i;
+
+       if (p_params->type == QED_FILTER_MAC)
+               return 0;
+
+       /* First remove entries and then add new ones */
+       if (p_params->opcode == QED_FILTER_REMOVE) {
+               for (i = 0; i < QED_ETH_VF_NUM_VLAN_FILTERS + 1; i++)
+                       if (p_vf->shadow_config.vlans[i].used &&
+                           p_vf->shadow_config.vlans[i].vid ==
+                           p_params->vlan) {
+                               p_vf->shadow_config.vlans[i].used = false;
+                               break;
+                       }
+               if (i == QED_ETH_VF_NUM_VLAN_FILTERS + 1) {
+                       DP_VERBOSE(p_hwfn,
+                                  QED_MSG_IOV,
+                                  "VF [%d] - Tries to remove a non-existing vlan\n",
+                                  p_vf->relative_vf_id);
+                       return -EINVAL;
+               }
+       } else if (p_params->opcode == QED_FILTER_REPLACE ||
+                  p_params->opcode == QED_FILTER_FLUSH) {
+               for (i = 0; i < QED_ETH_VF_NUM_VLAN_FILTERS + 1; i++)
+                       p_vf->shadow_config.vlans[i].used = false;
+       }
+
+       /* In forced mode, we're willing to remove entries - but we don't add
+        * new ones.
+        */
+       if (p_vf->bulletin.p_virt->valid_bitmap & (1 << VLAN_ADDR_FORCED))
+               return 0;
+
+       if (p_params->opcode == QED_FILTER_ADD ||
+           p_params->opcode == QED_FILTER_REPLACE) {
+               for (i = 0; i < QED_ETH_VF_NUM_VLAN_FILTERS + 1; i++) {
+                       if (p_vf->shadow_config.vlans[i].used)
+                               continue;
+
+                       p_vf->shadow_config.vlans[i].used = true;
+                       p_vf->shadow_config.vlans[i].vid = p_params->vlan;
+                       break;
+               }
+
+               if (i == QED_ETH_VF_NUM_VLAN_FILTERS + 1) {
+                       DP_VERBOSE(p_hwfn,
+                                  QED_MSG_IOV,
+                                  "VF [%d] - Tries to configure more than %d vlan filters\n",
+                                  p_vf->relative_vf_id,
+                                  QED_ETH_VF_NUM_VLAN_FILTERS + 1);
+                       return -EINVAL;
+               }
+       }
+
+       return 0;
+}
+
 int qed_iov_chk_ucast(struct qed_hwfn *hwfn,
                      int vfid, struct qed_filter_ucast *params)
 {
@@ -1907,6 +2140,7 @@ static void qed_iov_vf_mbx_ucast_filter(struct qed_hwfn *p_hwfn,
                                        struct qed_ptt *p_ptt,
                                        struct qed_vf_info *vf)
 {
+       struct qed_bulletin_content *p_bulletin = vf->bulletin.p_virt;
        struct qed_iov_vf_mbx *mbx = &vf->vf_mbx;
        struct vfpf_ucast_filter_tlv *req;
        u8 status = PFVF_STATUS_SUCCESS;
@@ -1946,6 +2180,25 @@ static void qed_iov_vf_mbx_ucast_filter(struct qed_hwfn *p_hwfn,
                goto out;
        }
 
+       /* Update shadow copy of the VF configuration */
+       if (qed_iov_vf_update_unicast_shadow(p_hwfn, vf, &params)) {
+               status = PFVF_STATUS_FAILURE;
+               goto out;
+       }
+
+       /* Determine if the unicast filtering is acceptible by PF */
+       if ((p_bulletin->valid_bitmap & (1 << VLAN_ADDR_FORCED)) &&
+           (params.type == QED_FILTER_VLAN ||
+            params.type == QED_FILTER_MAC_VLAN)) {
+               /* Once VLAN is forced or PVID is set, do not allow
+                * to add/replace any further VLANs.
+                */
+               if (params.opcode == QED_FILTER_ADD ||
+                   params.opcode == QED_FILTER_REPLACE)
+                       status = PFVF_STATUS_FORCED;
+               goto out;
+       }
+
        rc = qed_iov_chk_ucast(p_hwfn, vf->relative_vf_id, &params);
        if (rc) {
                status = PFVF_STATUS_FAILURE;
@@ -2449,6 +2702,29 @@ static int qed_iov_copy_vf_msg(struct qed_hwfn *p_hwfn, struct qed_ptt *ptt,
        return 0;
 }
 
+void qed_iov_bulletin_set_forced_vlan(struct qed_hwfn *p_hwfn,
+                                     u16 pvid, int vfid)
+{
+       struct qed_vf_info *vf_info;
+       u64 feature;
+
+       vf_info = qed_iov_get_vf_info(p_hwfn, (u16) vfid, true);
+       if (!vf_info) {
+               DP_NOTICE(p_hwfn->cdev,
+                         "Can not set forced MAC, invalid vfid [%d]\n", vfid);
+               return;
+       }
+
+       feature = 1 << VLAN_ADDR_FORCED;
+       vf_info->bulletin.p_virt->pvid = pvid;
+       if (pvid)
+               vf_info->bulletin.p_virt->valid_bitmap |= feature;
+       else
+               vf_info->bulletin.p_virt->valid_bitmap &= ~feature;
+
+       qed_iov_configure_vport_forced(p_hwfn, vf_info, feature);
+}
+
 bool qed_iov_is_vf_stopped(struct qed_hwfn *p_hwfn, int vfid)
 {
        struct qed_vf_info *p_vf_info;
@@ -2460,6 +2736,20 @@ bool qed_iov_is_vf_stopped(struct qed_hwfn *p_hwfn, int vfid)
        return p_vf_info->state == VF_STOPPED;
 }
 
+u16 qed_iov_bulletin_get_forced_vlan(struct qed_hwfn *p_hwfn, u16 rel_vf_id)
+{
+       struct qed_vf_info *p_vf;
+
+       p_vf = qed_iov_get_vf_info(p_hwfn, rel_vf_id, true);
+       if (!p_vf || !p_vf->bulletin.p_virt)
+               return 0;
+
+       if (!(p_vf->bulletin.p_virt->valid_bitmap & (1 << VLAN_ADDR_FORCED)))
+               return 0;
+
+       return p_vf->bulletin.p_virt->pvid;
+}
+
 /**
  * qed_schedule_iov - schedules IOV task for VF and PF
  * @hwfn: hardware function pointer
@@ -2609,6 +2899,38 @@ static int qed_sriov_configure(struct qed_dev *cdev, int num_vfs_param)
                return qed_sriov_disable(cdev, true);
 }
 
+static int qed_sriov_pf_set_vlan(struct qed_dev *cdev, u16 vid, int vfid)
+{
+       int i;
+
+       if (!IS_QED_SRIOV(cdev) || !IS_PF_SRIOV_ALLOC(&cdev->hwfns[0])) {
+               DP_VERBOSE(cdev, QED_MSG_IOV,
+                          "Cannot set a VF MAC; Sriov is not enabled\n");
+               return -EINVAL;
+       }
+
+       if (!qed_iov_is_valid_vfid(&cdev->hwfns[0], vfid, true)) {
+               DP_VERBOSE(cdev, QED_MSG_IOV,
+                          "Cannot set VF[%d] MAC (VF is not active)\n", vfid);
+               return -EINVAL;
+       }
+
+       for_each_hwfn(cdev, i) {
+               struct qed_hwfn *hwfn = &cdev->hwfns[i];
+               struct qed_public_vf_info *vf_info;
+
+               vf_info = qed_iov_get_public_vf_info(hwfn, vfid, true);
+               if (!vf_info)
+                       continue;
+
+               /* Set the forced vlan, and schedule the IOV task */
+               vf_info->forced_vlan = vid;
+               qed_schedule_iov(hwfn, QED_IOV_WQ_SET_UNICAST_FILTER_FLAG);
+       }
+
+       return 0;
+}
+
 void qed_inform_vf_link_state(struct qed_hwfn *hwfn)
 {
        struct qed_mcp_link_capabilities caps;
@@ -2671,6 +2993,38 @@ static void qed_handle_vf_msg(struct qed_hwfn *hwfn)
        qed_ptt_release(hwfn, ptt);
 }
 
+static void qed_handle_pf_set_vf_unicast(struct qed_hwfn *hwfn)
+{
+       int i;
+
+       qed_for_each_vf(hwfn, i) {
+               struct qed_public_vf_info *info;
+               bool update = false;
+
+               info = qed_iov_get_public_vf_info(hwfn, i, true);
+               if (!info)
+                       continue;
+
+               /* Update data on bulletin board */
+
+               if (qed_iov_bulletin_get_forced_vlan(hwfn, i) ^
+                   info->forced_vlan) {
+                       DP_VERBOSE(hwfn,
+                                  QED_MSG_IOV,
+                                  "Handling PF setting of pvid [0x%04x] to VF 0x%02x [Abs 0x%02x]\n",
+                                  info->forced_vlan,
+                                  i,
+                                  hwfn->cdev->p_iov_info->first_vf_in_pf + i);
+                       qed_iov_bulletin_set_forced_vlan(hwfn,
+                                                        info->forced_vlan, i);
+                       update = true;
+               }
+
+               if (update)
+                       qed_schedule_iov(hwfn, QED_IOV_WQ_BULLETIN_UPDATE_FLAG);
+       }
+}
+
 static void qed_handle_bulletin_post(struct qed_hwfn *hwfn)
 {
        struct qed_ptt *ptt;
@@ -2715,6 +3069,11 @@ void qed_iov_pf_task(struct work_struct *work)
 
        if (test_and_clear_bit(QED_IOV_WQ_MSG_FLAG, &hwfn->iov_task_flags))
                qed_handle_vf_msg(hwfn);
+
+       if (test_and_clear_bit(QED_IOV_WQ_SET_UNICAST_FILTER_FLAG,
+                              &hwfn->iov_task_flags))
+               qed_handle_pf_set_vf_unicast(hwfn);
+
        if (test_and_clear_bit(QED_IOV_WQ_BULLETIN_UPDATE_FLAG,
                               &hwfn->iov_task_flags))
                qed_handle_bulletin_post(hwfn);
@@ -2774,4 +3133,5 @@ int qed_iov_wq_start(struct qed_dev *cdev)
 
 const struct qed_iov_hv_ops qed_iov_ops_pass = {
        .configure = &qed_sriov_configure,
+       .set_vlan = &qed_sriov_pf_set_vlan,
 };
index 2c94b44..e65f403 100644 (file)
@@ -24,6 +24,9 @@
 #define QED_MAX_VF_CHAINS_PER_PF 16
 #define QED_ETH_VF_NUM_VLAN_FILTERS 2
 
+#define QED_ETH_MAX_VF_NUM_VLAN_FILTERS        \
+       (MAX_NUM_VFS * QED_ETH_VF_NUM_VLAN_FILTERS)
+
 enum qed_iov_vport_update_flag {
        QED_IOV_VP_UPDATE_ACTIVATE,
        QED_IOV_VP_UPDATE_VLAN_STRIP,
@@ -40,6 +43,7 @@ struct qed_public_vf_info {
        /* These copies will later be reflected in the bulletin board,
         * but this copy should be newer.
         */
+       u16 forced_vlan;
        u8 mac[ETH_ALEN];
 };
 
@@ -98,6 +102,18 @@ enum vf_state {
        VF_STOPPED              /* VF, Stopped */
 };
 
+struct qed_vf_vlan_shadow {
+       bool used;
+       u16 vid;
+};
+
+struct qed_vf_shadow_config {
+       /* Shadow copy of all guest vlans */
+       struct qed_vf_vlan_shadow vlans[QED_ETH_VF_NUM_VLAN_FILTERS + 1];
+
+       u8 inner_vlan_removal;
+};
+
 /* PFs maintain an array of this structure, per VF */
 struct qed_vf_info {
        struct qed_iov_vf_mbx vf_mbx;
@@ -131,6 +147,16 @@ struct qed_vf_info {
        u16 igu_sbs[QED_MAX_VF_CHAINS_PER_PF];
        u8 num_active_rxqs;
        struct qed_public_vf_info p_vf_info;
+
+       /* Stores the configuration requested by VF */
+       struct qed_vf_shadow_config shadow_config;
+
+       /* A bitfield using bulletin's valid-map bits, used to indicate
+        * which of the bulletin board features have been configured.
+        */
+       u64 configured_features;
+#define QED_IOV_CONFIGURED_FEATURES_MASK        ((1 << MAC_ADDR_FORCED) | \
+                                                (1 << VLAN_ADDR_FORCED))
 };
 
 /* This structure is part of qed_hwfn and used only for PFs that have sriov
index e788954..3c8911d 100644 (file)
@@ -474,7 +474,7 @@ int qed_vf_pf_vport_start(struct qed_hwfn *p_hwfn,
                          u16 mtu,
                          u8 inner_vlan_removal,
                          enum qed_tpa_mode tpa_mode,
-                         u8 max_buffers_per_cqe)
+                         u8 max_buffers_per_cqe, u8 only_untagged)
 {
        struct qed_vf_iov *p_iov = p_hwfn->vf_iov_info;
        struct vfpf_vport_start_tlv *req;
@@ -489,6 +489,7 @@ int qed_vf_pf_vport_start(struct qed_hwfn *p_hwfn,
        req->inner_vlan_removal = inner_vlan_removal;
        req->tpa_mode = tpa_mode;
        req->max_buffers_per_cqe = max_buffers_per_cqe;
+       req->only_untagged = only_untagged;
 
        /* status blocks */
        for (i = 0; i < p_hwfn->vf_iov_info->acquire_resp.resc.num_sbs; i++)
@@ -547,6 +548,8 @@ qed_vf_handle_vp_update_is_needed(struct qed_hwfn *p_hwfn,
                return !!p_data->update_tx_switching_flg;
        case CHANNEL_TLV_VPORT_UPDATE_VLAN_STRIP:
                return !!p_data->update_inner_vlan_removal_flg;
+       case CHANNEL_TLV_VPORT_UPDATE_ACCEPT_ANY_VLAN:
+               return !!p_data->update_accept_any_vlan_flg;
        case CHANNEL_TLV_VPORT_UPDATE_MCAST:
                return !!p_data->update_approx_mcast_flg;
        case CHANNEL_TLV_VPORT_UPDATE_ACCEPT_PARAM:
@@ -696,6 +699,19 @@ int qed_vf_pf_vport_update(struct qed_hwfn *p_hwfn,
                       sizeof(rss_params->rss_key));
        }
 
+       if (p_params->update_accept_any_vlan_flg) {
+               struct vfpf_vport_update_accept_any_vlan_tlv *p_any_vlan_tlv;
+
+               size = sizeof(struct vfpf_vport_update_accept_any_vlan_tlv);
+               tlv = CHANNEL_TLV_VPORT_UPDATE_ACCEPT_ANY_VLAN;
+               p_any_vlan_tlv = qed_add_tlv(p_hwfn, &p_iov->offset, tlv, size);
+
+               resp_size += sizeof(struct pfvf_def_resp_tlv);
+               p_any_vlan_tlv->accept_any_vlan = p_params->accept_any_vlan;
+               p_any_vlan_tlv->update_accept_any_vlan_flg =
+                   p_params->update_accept_any_vlan_flg;
+       }
+
        /* add list termination tlv */
        qed_add_tlv(p_hwfn, &p_iov->offset,
                    CHANNEL_TLV_LIST_END, sizeof(struct channel_list_end_tlv));
index d9a8aa6..35eced3 100644 (file)
@@ -417,6 +417,16 @@ union pfvf_tlvs {
        struct pfvf_start_queue_resp_tlv queue_start;
 };
 
+enum qed_bulletin_bit {
+       /* Alert the VF that a forced VLAN was set by the PF */
+       VLAN_ADDR_FORCED = 2,
+
+       /* Indicate that `default_only_untagged' contains actual data */
+       VFPF_BULLETIN_UNTAGGED_DEFAULT = 3,
+       VFPF_BULLETIN_UNTAGGED_DEFAULT_FORCED = 4,
+
+};
+
 struct qed_bulletin_content {
        /* crc of structure to ensure is not in mid-update */
        u32 crc;
@@ -465,6 +475,10 @@ struct qed_bulletin_content {
        u32 partner_adv_speed;
 
        u32 capability_speed;
+
+       /* Forced vlan */
+       u16 pvid;
+       u16 padding5;
 };
 
 struct qed_bulletin {
@@ -737,7 +751,7 @@ int qed_vf_pf_vport_start(struct qed_hwfn *p_hwfn,
                          u16 mtu,
                          u8 inner_vlan_removal,
                          enum qed_tpa_mode tpa_mode,
-                         u8 max_buffers_per_cqe);
+                         u8 max_buffers_per_cqe, u8 only_untagged);
 
 /**
  * @brief qed_vf_pf_vport_stop - stop the VF's vport
@@ -898,7 +912,8 @@ static inline int qed_vf_pf_vport_start(struct qed_hwfn *p_hwfn,
                                        u16 mtu,
                                        u8 inner_vlan_removal,
                                        enum qed_tpa_mode tpa_mode,
-                                       u8 max_buffers_per_cqe)
+                                       u8 max_buffers_per_cqe,
+                                       u8 only_untagged)
 {
        return -EINVAL;
 }
index bf54cfc..4d59d7e 100644 (file)
@@ -103,6 +103,21 @@ static int qede_alloc_rx_buffer(struct qede_dev *edev,
 static void qede_link_update(void *dev, struct qed_link_output *link);
 
 #ifdef CONFIG_QED_SRIOV
+static int qede_set_vf_vlan(struct net_device *ndev, int vf, u16 vlan, u8 qos)
+{
+       struct qede_dev *edev = netdev_priv(ndev);
+
+       if (vlan > 4095) {
+               DP_NOTICE(edev, "Illegal vlan value %d\n", vlan);
+               return -EINVAL;
+       }
+
+       DP_VERBOSE(edev, QED_MSG_IOV, "Setting Vlan 0x%04x to VF [%d]\n",
+                  vlan, vf);
+
+       return edev->ops->iov->set_vlan(edev->cdev, vlan, vf);
+}
+
 static int qede_sriov_configure(struct pci_dev *pdev, int num_vfs_param)
 {
        struct qede_dev *edev = netdev_priv(pci_get_drvdata(pdev));
@@ -2071,6 +2086,9 @@ static const struct net_device_ops qede_netdev_ops = {
        .ndo_set_mac_address = qede_set_mac_addr,
        .ndo_validate_addr = eth_validate_addr,
        .ndo_change_mtu = qede_change_mtu,
+#ifdef CONFIG_QED_SRIOV
+       .ndo_set_vf_vlan = qede_set_vf_vlan,
+#endif
        .ndo_vlan_rx_add_vid = qede_vlan_rx_add_vid,
        .ndo_vlan_rx_kill_vid = qede_vlan_rx_kill_vid,
        .ndo_get_stats64 = qede_get_stats64,
index c53bfa6..825c007 100644 (file)
@@ -15,6 +15,7 @@
 struct qed_iov_hv_ops {
        int (*configure)(struct qed_dev *cdev, int num_vfs_param);
 
+       int (*set_vlan) (struct qed_dev *cdev, u16 vid, int vfid);
 };
 
 #endif