net/mlx5e: IPoIB, Move to a separate directory
authorSaeed Mahameed <saeedm@mellanox.com>
Sun, 18 Jun 2017 14:13:44 +0000 (17:13 +0300)
committerSaeed Mahameed <saeedm@mellanox.com>
Mon, 19 Jun 2017 15:40:20 +0000 (18:40 +0300)
IPoIB netdevice driver was only introduced in previous kernel release
and it is growing in terms of features and LOC, move it to a separate
directory.

Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
drivers/net/ethernet/mellanox/mlx5/core/Makefile
drivers/net/ethernet/mellanox/mlx5/core/en_rx.c
drivers/net/ethernet/mellanox/mlx5/core/en_tx.c
drivers/net/ethernet/mellanox/mlx5/core/ipoib.c [deleted file]
drivers/net/ethernet/mellanox/mlx5/core/ipoib.h [deleted file]
drivers/net/ethernet/mellanox/mlx5/core/ipoib/ipoib.c [new file with mode: 0644]
drivers/net/ethernet/mellanox/mlx5/core/ipoib/ipoib.h [new file with mode: 0644]

index 12556c03eec40811aa7cfad0fa2d36d1fca73bd4..59b87b30e438d01318ddac6084de08c1230b9683 100644 (file)
@@ -15,4 +15,4 @@ mlx5_core-$(CONFIG_MLX5_CORE_EN) += wq.o eswitch.o eswitch_offloads.o \
 
 mlx5_core-$(CONFIG_MLX5_CORE_EN_DCB) +=  en_dcbnl.o
 
-mlx5_core-$(CONFIG_MLX5_CORE_IPOIB) += ipoib.o
+mlx5_core-$(CONFIG_MLX5_CORE_IPOIB) += ipoib/ipoib.o
index 66b5fec1531349d14b5324f68a28746cecdf1918..806139d6213935fe0911b2ea17f739676ce3f44d 100644 (file)
@@ -40,7 +40,7 @@
 #include "en_tc.h"
 #include "eswitch.h"
 #include "en_rep.h"
-#include "ipoib.h"
+#include "ipoib/ipoib.h"
 
 static inline bool mlx5e_rx_hw_stamp(struct mlx5e_tstamp *tstamp)
 {
index ef3e1918d8cc80cb0ff2d3f8c209d8ab06074299..354f474322cef4b5fbfeb199d0c19ad12b47243d 100644 (file)
@@ -33,7 +33,7 @@
 #include <linux/tcp.h>
 #include <linux/if_vlan.h>
 #include "en.h"
-#include "ipoib.h"
+#include "ipoib/ipoib.h"
 
 #define MLX5E_SQ_NOPS_ROOM  MLX5_SEND_WQE_MAX_WQEBBS
 #define MLX5E_SQ_STOP_ROOM (MLX5_SEND_WQE_MAX_WQEBBS +\
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/ipoib.c b/drivers/net/ethernet/mellanox/mlx5/core/ipoib.c
deleted file mode 100644 (file)
index 22ca591..0000000
+++ /dev/null
@@ -1,513 +0,0 @@
-/*
- * Copyright (c) 2017, Mellanox Technologies. All rights reserved.
- *
- * This software is available to you under a choice of one of two
- * licenses.  You may choose to be licensed under the terms of the GNU
- * General Public License (GPL) Version 2, available from the file
- * COPYING in the main directory of this source tree, or the
- * OpenIB.org BSD license below:
- *
- *     Redistribution and use in source and binary forms, with or
- *     without modification, are permitted provided that the following
- *     conditions are met:
- *
- *      - Redistributions of source code must retain the above
- *        copyright notice, this list of conditions and the following
- *        disclaimer.
- *
- *      - Redistributions in binary form must reproduce the above
- *        copyright notice, this list of conditions and the following
- *        disclaimer in the documentation and/or other materials
- *        provided with the distribution.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
- * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
- * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
- * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
- * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
- * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
- * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
- * SOFTWARE.
- */
-
-#include <rdma/ib_verbs.h>
-#include <linux/mlx5/fs.h>
-#include "en.h"
-#include "ipoib.h"
-
-#define IB_DEFAULT_Q_KEY   0xb1b
-
-static int mlx5i_open(struct net_device *netdev);
-static int mlx5i_close(struct net_device *netdev);
-static int  mlx5i_dev_init(struct net_device *dev);
-static void mlx5i_dev_cleanup(struct net_device *dev);
-
-static const struct net_device_ops mlx5i_netdev_ops = {
-       .ndo_open                = mlx5i_open,
-       .ndo_stop                = mlx5i_close,
-       .ndo_init                = mlx5i_dev_init,
-       .ndo_uninit              = mlx5i_dev_cleanup,
-};
-
-/* IPoIB mlx5 netdev profile */
-
-/* Called directly after IPoIB netdevice was created to initialize SW structs */
-static void mlx5i_init(struct mlx5_core_dev *mdev,
-                      struct net_device *netdev,
-                      const struct mlx5e_profile *profile,
-                      void *ppriv)
-{
-       struct mlx5e_priv *priv  = mlx5i_epriv(netdev);
-
-       priv->mdev        = mdev;
-       priv->netdev      = netdev;
-       priv->profile     = profile;
-       priv->ppriv       = ppriv;
-
-       mlx5e_build_nic_params(mdev, &priv->channels.params, profile->max_nch(mdev));
-
-       /* Override RQ params as IPoIB supports only LINKED LIST RQ for now */
-       mlx5e_set_rq_type_params(mdev, &priv->channels.params, MLX5_WQ_TYPE_LINKED_LIST);
-       priv->channels.params.lro_en = false;
-
-       mutex_init(&priv->state_lock);
-
-       netdev->hw_features    |= NETIF_F_SG;
-       netdev->hw_features    |= NETIF_F_IP_CSUM;
-       netdev->hw_features    |= NETIF_F_IPV6_CSUM;
-       netdev->hw_features    |= NETIF_F_GRO;
-       netdev->hw_features    |= NETIF_F_TSO;
-       netdev->hw_features    |= NETIF_F_TSO6;
-       netdev->hw_features    |= NETIF_F_RXCSUM;
-       netdev->hw_features    |= NETIF_F_RXHASH;
-
-       netdev->netdev_ops = &mlx5i_netdev_ops;
-}
-
-/* Called directly before IPoIB netdevice is destroyed to cleanup SW structs */
-static void mlx5i_cleanup(struct mlx5e_priv *priv)
-{
-       /* Do nothing .. */
-}
-
-#define MLX5_QP_ENHANCED_ULP_STATELESS_MODE 2
-
-static int mlx5i_create_underlay_qp(struct mlx5_core_dev *mdev, struct mlx5_core_qp *qp)
-{
-       struct mlx5_qp_context *context = NULL;
-       u32 *in = NULL;
-       void *addr_path;
-       int ret = 0;
-       int inlen;
-       void *qpc;
-
-       inlen = MLX5_ST_SZ_BYTES(create_qp_in);
-       in = kvzalloc(inlen, GFP_KERNEL);
-       if (!in)
-               return -ENOMEM;
-
-       qpc = MLX5_ADDR_OF(create_qp_in, in, qpc);
-       MLX5_SET(qpc, qpc, st, MLX5_QP_ST_UD);
-       MLX5_SET(qpc, qpc, pm_state, MLX5_QP_PM_MIGRATED);
-       MLX5_SET(qpc, qpc, ulp_stateless_offload_mode,
-                MLX5_QP_ENHANCED_ULP_STATELESS_MODE);
-
-       addr_path = MLX5_ADDR_OF(qpc, qpc, primary_address_path);
-       MLX5_SET(ads, addr_path, port, 1);
-       MLX5_SET(ads, addr_path, grh, 1);
-
-       ret = mlx5_core_create_qp(mdev, qp, in, inlen);
-       if (ret) {
-               mlx5_core_err(mdev, "Failed creating IPoIB QP err : %d\n", ret);
-               goto out;
-       }
-
-       /* QP states */
-       context = kzalloc(sizeof(*context), GFP_KERNEL);
-       if (!context) {
-               ret = -ENOMEM;
-               goto out;
-       }
-
-       context->flags = cpu_to_be32(MLX5_QP_PM_MIGRATED << 11);
-       context->pri_path.port = 1;
-       context->qkey = cpu_to_be32(IB_DEFAULT_Q_KEY);
-
-       ret = mlx5_core_qp_modify(mdev, MLX5_CMD_OP_RST2INIT_QP, 0, context, qp);
-       if (ret) {
-               mlx5_core_err(mdev, "Failed to modify qp RST2INIT, err: %d\n", ret);
-               goto out;
-       }
-       memset(context, 0, sizeof(*context));
-
-       ret = mlx5_core_qp_modify(mdev, MLX5_CMD_OP_INIT2RTR_QP, 0, context, qp);
-       if (ret) {
-               mlx5_core_err(mdev, "Failed to modify qp INIT2RTR, err: %d\n", ret);
-               goto out;
-       }
-
-       ret = mlx5_core_qp_modify(mdev, MLX5_CMD_OP_RTR2RTS_QP, 0, context, qp);
-       if (ret) {
-               mlx5_core_err(mdev, "Failed to modify qp RTR2RTS, err: %d\n", ret);
-               goto out;
-       }
-
-out:
-       kfree(context);
-       kvfree(in);
-       return ret;
-}
-
-static void mlx5i_destroy_underlay_qp(struct mlx5_core_dev *mdev, struct mlx5_core_qp *qp)
-{
-       mlx5_fs_remove_rx_underlay_qpn(mdev, qp->qpn);
-
-       mlx5_core_destroy_qp(mdev, qp);
-}
-
-static int mlx5i_init_tx(struct mlx5e_priv *priv)
-{
-       struct mlx5i_priv *ipriv = priv->ppriv;
-       int err;
-
-       err = mlx5i_create_underlay_qp(priv->mdev, &ipriv->qp);
-       if (err) {
-               mlx5_core_warn(priv->mdev, "create underlay QP failed, %d\n", err);
-               return err;
-       }
-
-       mlx5_fs_add_rx_underlay_qpn(priv->mdev, ipriv->qp.qpn);
-
-       err = mlx5e_create_tis(priv->mdev, 0 /* tc */, ipriv->qp.qpn, &priv->tisn[0]);
-       if (err) {
-               mlx5_core_warn(priv->mdev, "create tis failed, %d\n", err);
-               return err;
-       }
-
-       return 0;
-}
-
-static void mlx5i_cleanup_tx(struct mlx5e_priv *priv)
-{
-       struct mlx5i_priv *ipriv = priv->ppriv;
-
-       mlx5e_destroy_tis(priv->mdev, priv->tisn[0]);
-       mlx5i_destroy_underlay_qp(priv->mdev, &ipriv->qp);
-}
-
-static int mlx5i_create_flow_steering(struct mlx5e_priv *priv)
-{
-       int err;
-
-       priv->fs.ns = mlx5_get_flow_namespace(priv->mdev,
-                                              MLX5_FLOW_NAMESPACE_KERNEL);
-
-       if (!priv->fs.ns)
-               return -EINVAL;
-
-       err = mlx5e_arfs_create_tables(priv);
-       if (err) {
-               netdev_err(priv->netdev, "Failed to create arfs tables, err=%d\n",
-                          err);
-               priv->netdev->hw_features &= ~NETIF_F_NTUPLE;
-       }
-
-       err = mlx5e_create_ttc_table(priv);
-       if (err) {
-               netdev_err(priv->netdev, "Failed to create ttc table, err=%d\n",
-                          err);
-               goto err_destroy_arfs_tables;
-       }
-
-       return 0;
-
-err_destroy_arfs_tables:
-       mlx5e_arfs_destroy_tables(priv);
-
-       return err;
-}
-
-static void mlx5i_destroy_flow_steering(struct mlx5e_priv *priv)
-{
-       mlx5e_destroy_ttc_table(priv);
-       mlx5e_arfs_destroy_tables(priv);
-}
-
-static int mlx5i_init_rx(struct mlx5e_priv *priv)
-{
-       int err;
-
-       err = mlx5e_create_indirect_rqt(priv);
-       if (err)
-               return err;
-
-       err = mlx5e_create_direct_rqts(priv);
-       if (err)
-               goto err_destroy_indirect_rqts;
-
-       err = mlx5e_create_indirect_tirs(priv);
-       if (err)
-               goto err_destroy_direct_rqts;
-
-       err = mlx5e_create_direct_tirs(priv);
-       if (err)
-               goto err_destroy_indirect_tirs;
-
-       err = mlx5i_create_flow_steering(priv);
-       if (err)
-               goto err_destroy_direct_tirs;
-
-       return 0;
-
-err_destroy_direct_tirs:
-       mlx5e_destroy_direct_tirs(priv);
-err_destroy_indirect_tirs:
-       mlx5e_destroy_indirect_tirs(priv);
-err_destroy_direct_rqts:
-       mlx5e_destroy_direct_rqts(priv);
-err_destroy_indirect_rqts:
-       mlx5e_destroy_rqt(priv, &priv->indir_rqt);
-       return err;
-}
-
-static void mlx5i_cleanup_rx(struct mlx5e_priv *priv)
-{
-       mlx5i_destroy_flow_steering(priv);
-       mlx5e_destroy_direct_tirs(priv);
-       mlx5e_destroy_indirect_tirs(priv);
-       mlx5e_destroy_direct_rqts(priv);
-       mlx5e_destroy_rqt(priv, &priv->indir_rqt);
-}
-
-static const struct mlx5e_profile mlx5i_nic_profile = {
-       .init              = mlx5i_init,
-       .cleanup           = mlx5i_cleanup,
-       .init_tx           = mlx5i_init_tx,
-       .cleanup_tx        = mlx5i_cleanup_tx,
-       .init_rx           = mlx5i_init_rx,
-       .cleanup_rx        = mlx5i_cleanup_rx,
-       .enable            = NULL, /* mlx5i_enable */
-       .disable           = NULL, /* mlx5i_disable */
-       .update_stats      = NULL, /* mlx5i_update_stats */
-       .max_nch           = mlx5e_get_max_num_channels,
-       .rx_handlers.handle_rx_cqe       = mlx5i_handle_rx_cqe,
-       .rx_handlers.handle_rx_cqe_mpwqe = NULL, /* Not supported */
-       .max_tc            = MLX5I_MAX_NUM_TC,
-};
-
-/* mlx5i netdev NDos */
-
-static int mlx5i_dev_init(struct net_device *dev)
-{
-       struct mlx5e_priv    *priv   = mlx5i_epriv(dev);
-       struct mlx5i_priv    *ipriv  = priv->ppriv;
-
-       /* Set dev address using underlay QP */
-       dev->dev_addr[1] = (ipriv->qp.qpn >> 16) & 0xff;
-       dev->dev_addr[2] = (ipriv->qp.qpn >>  8) & 0xff;
-       dev->dev_addr[3] = (ipriv->qp.qpn) & 0xff;
-
-       return 0;
-}
-
-static void mlx5i_dev_cleanup(struct net_device *dev)
-{
-       struct mlx5e_priv    *priv   = mlx5i_epriv(dev);
-       struct mlx5_core_dev *mdev   = priv->mdev;
-       struct mlx5i_priv    *ipriv  = priv->ppriv;
-       struct mlx5_qp_context context;
-
-       /* detach qp from flow-steering by reset it */
-       mlx5_core_qp_modify(mdev, MLX5_CMD_OP_2RST_QP, 0, &context, &ipriv->qp);
-}
-
-static int mlx5i_open(struct net_device *netdev)
-{
-       struct mlx5e_priv *priv = mlx5i_epriv(netdev);
-       int err;
-
-       mutex_lock(&priv->state_lock);
-
-       set_bit(MLX5E_STATE_OPENED, &priv->state);
-
-       err = mlx5e_open_channels(priv, &priv->channels);
-       if (err)
-               goto err_clear_state_opened_flag;
-
-       mlx5e_refresh_tirs(priv, false);
-       mlx5e_activate_priv_channels(priv);
-       mutex_unlock(&priv->state_lock);
-       return 0;
-
-err_clear_state_opened_flag:
-       clear_bit(MLX5E_STATE_OPENED, &priv->state);
-       mutex_unlock(&priv->state_lock);
-       return err;
-}
-
-static int mlx5i_close(struct net_device *netdev)
-{
-       struct mlx5e_priv *priv = mlx5i_epriv(netdev);
-
-       /* May already be CLOSED in case a previous configuration operation
-        * (e.g RX/TX queue size change) that involves close&open failed.
-        */
-       mutex_lock(&priv->state_lock);
-
-       if (!test_bit(MLX5E_STATE_OPENED, &priv->state))
-               goto unlock;
-
-       clear_bit(MLX5E_STATE_OPENED, &priv->state);
-
-       netif_carrier_off(priv->netdev);
-       mlx5e_deactivate_priv_channels(priv);
-       mlx5e_close_channels(&priv->channels);
-unlock:
-       mutex_unlock(&priv->state_lock);
-       return 0;
-}
-
-/* IPoIB RDMA netdev callbacks */
-static int mlx5i_attach_mcast(struct net_device *netdev, struct ib_device *hca,
-                             union ib_gid *gid, u16 lid, int set_qkey,
-                             u32 qkey)
-{
-       struct mlx5e_priv    *epriv = mlx5i_epriv(netdev);
-       struct mlx5_core_dev *mdev  = epriv->mdev;
-       struct mlx5i_priv    *ipriv = epriv->ppriv;
-       int err;
-
-       mlx5_core_dbg(mdev, "attaching QPN 0x%x, MGID %pI6\n", ipriv->qp.qpn, gid->raw);
-       err = mlx5_core_attach_mcg(mdev, gid, ipriv->qp.qpn);
-       if (err)
-               mlx5_core_warn(mdev, "failed attaching QPN 0x%x, MGID %pI6\n",
-                              ipriv->qp.qpn, gid->raw);
-
-       if (set_qkey) {
-               mlx5_core_dbg(mdev, "%s setting qkey 0x%x\n",
-                             netdev->name, qkey);
-               ipriv->qkey = qkey;
-       }
-
-       return err;
-}
-
-static int mlx5i_detach_mcast(struct net_device *netdev, struct ib_device *hca,
-                             union ib_gid *gid, u16 lid)
-{
-       struct mlx5e_priv    *epriv = mlx5i_epriv(netdev);
-       struct mlx5_core_dev *mdev  = epriv->mdev;
-       struct mlx5i_priv    *ipriv = epriv->ppriv;
-       int err;
-
-       mlx5_core_dbg(mdev, "detaching QPN 0x%x, MGID %pI6\n", ipriv->qp.qpn, gid->raw);
-
-       err = mlx5_core_detach_mcg(mdev, gid, ipriv->qp.qpn);
-       if (err)
-               mlx5_core_dbg(mdev, "failed dettaching QPN 0x%x, MGID %pI6\n",
-                             ipriv->qp.qpn, gid->raw);
-
-       return err;
-}
-
-static int mlx5i_xmit(struct net_device *dev, struct sk_buff *skb,
-                     struct ib_ah *address, u32 dqpn)
-{
-       struct mlx5e_priv *epriv = mlx5i_epriv(dev);
-       struct mlx5e_txqsq *sq   = epriv->txq2sq[skb_get_queue_mapping(skb)];
-       struct mlx5_ib_ah *mah   = to_mah(address);
-       struct mlx5i_priv *ipriv = epriv->ppriv;
-
-       return mlx5i_sq_xmit(sq, skb, &mah->av, dqpn, ipriv->qkey);
-}
-
-static int mlx5i_check_required_hca_cap(struct mlx5_core_dev *mdev)
-{
-       if (MLX5_CAP_GEN(mdev, port_type) != MLX5_CAP_PORT_TYPE_IB)
-               return -EOPNOTSUPP;
-
-       if (!MLX5_CAP_GEN(mdev, ipoib_enhanced_offloads)) {
-               mlx5_core_warn(mdev, "IPoIB enhanced offloads are not supported\n");
-               return -EOPNOTSUPP;
-       }
-
-       return 0;
-}
-
-struct net_device *mlx5_rdma_netdev_alloc(struct mlx5_core_dev *mdev,
-                                         struct ib_device *ibdev,
-                                         const char *name,
-                                         void (*setup)(struct net_device *))
-{
-       const struct mlx5e_profile *profile = &mlx5i_nic_profile;
-       int nch = profile->max_nch(mdev);
-       struct net_device *netdev;
-       struct mlx5i_priv *ipriv;
-       struct mlx5e_priv *epriv;
-       struct rdma_netdev *rn;
-       int err;
-
-       if (mlx5i_check_required_hca_cap(mdev)) {
-               mlx5_core_warn(mdev, "Accelerated mode is not supported\n");
-               return ERR_PTR(-EOPNOTSUPP);
-       }
-
-       /* This function should only be called once per mdev */
-       err = mlx5e_create_mdev_resources(mdev);
-       if (err)
-               return NULL;
-
-       netdev = alloc_netdev_mqs(sizeof(struct mlx5i_priv) + sizeof(struct mlx5e_priv),
-                                 name, NET_NAME_UNKNOWN,
-                                 setup,
-                                 nch * MLX5E_MAX_NUM_TC,
-                                 nch);
-       if (!netdev) {
-               mlx5_core_warn(mdev, "alloc_netdev_mqs failed\n");
-               goto free_mdev_resources;
-       }
-
-       ipriv = netdev_priv(netdev);
-       epriv = mlx5i_epriv(netdev);
-
-       epriv->wq = create_singlethread_workqueue("mlx5i");
-       if (!epriv->wq)
-               goto err_free_netdev;
-
-       profile->init(mdev, netdev, profile, ipriv);
-
-       mlx5e_attach_netdev(epriv);
-       netif_carrier_off(netdev);
-
-       /* set rdma_netdev func pointers */
-       rn = &ipriv->rn;
-       rn->hca  = ibdev;
-       rn->send = mlx5i_xmit;
-       rn->attach_mcast = mlx5i_attach_mcast;
-       rn->detach_mcast = mlx5i_detach_mcast;
-
-       return netdev;
-
-err_free_netdev:
-       free_netdev(netdev);
-free_mdev_resources:
-       mlx5e_destroy_mdev_resources(mdev);
-
-       return NULL;
-}
-EXPORT_SYMBOL(mlx5_rdma_netdev_alloc);
-
-void mlx5_rdma_netdev_free(struct net_device *netdev)
-{
-       struct mlx5e_priv          *priv    = mlx5i_epriv(netdev);
-       const struct mlx5e_profile *profile = priv->profile;
-
-       mlx5e_detach_netdev(priv);
-       profile->cleanup(priv);
-       destroy_workqueue(priv->wq);
-       free_netdev(netdev);
-
-       mlx5e_destroy_mdev_resources(priv->mdev);
-}
-EXPORT_SYMBOL(mlx5_rdma_netdev_free);
-
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/ipoib.h b/drivers/net/ethernet/mellanox/mlx5/core/ipoib.h
deleted file mode 100644 (file)
index 213191a..0000000
+++ /dev/null
@@ -1,56 +0,0 @@
-/*
- * Copyright (c) 2017, Mellanox Technologies. All rights reserved.
- *
- * This software is available to you under a choice of one of two
- * licenses.  You may choose to be licensed under the terms of the GNU
- * General Public License (GPL) Version 2, available from the file
- * COPYING in the main directory of this source tree, or the
- * OpenIB.org BSD license below:
- *
- *     Redistribution and use in source and binary forms, with or
- *     without modification, are permitted provided that the following
- *     conditions are met:
- *
- *      - Redistributions of source code must retain the above
- *        copyright notice, this list of conditions and the following
- *        disclaimer.
- *
- *      - Redistributions in binary form must reproduce the above
- *        copyright notice, this list of conditions and the following
- *        disclaimer in the documentation and/or other materials
- *        provided with the distribution.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
- * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
- * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
- * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
- * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
- * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
- * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
- * SOFTWARE.
- */
-
-#ifndef __MLX5E_IPOB_H__
-#define __MLX5E_IPOB_H__
-
-#include <linux/mlx5/fs.h>
-#include "en.h"
-
-#define MLX5I_MAX_NUM_TC 1
-
-/* ipoib rdma netdev's private data structure */
-struct mlx5i_priv {
-       struct rdma_netdev rn; /* keep this first */
-       struct mlx5_core_qp qp;
-       u32    qkey;
-       char  *mlx5e_priv[0];
-};
-
-/* Extract mlx5e_priv from IPoIB netdev */
-#define mlx5i_epriv(netdev) ((void *)(((struct mlx5i_priv *)netdev_priv(netdev))->mlx5e_priv))
-
-netdev_tx_t mlx5i_sq_xmit(struct mlx5e_txqsq *sq, struct sk_buff *skb,
-                         struct mlx5_av *av, u32 dqpn, u32 dqkey);
-void mlx5i_handle_rx_cqe(struct mlx5e_rq *rq, struct mlx5_cqe64 *cqe);
-
-#endif /* __MLX5E_IPOB_H__ */
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/ipoib/ipoib.c b/drivers/net/ethernet/mellanox/mlx5/core/ipoib/ipoib.c
new file mode 100644 (file)
index 0000000..22ca591
--- /dev/null
@@ -0,0 +1,513 @@
+/*
+ * Copyright (c) 2017, Mellanox Technologies. All rights reserved.
+ *
+ * This software is available to you under a choice of one of two
+ * licenses.  You may choose to be licensed under the terms of the GNU
+ * General Public License (GPL) Version 2, available from the file
+ * COPYING in the main directory of this source tree, or the
+ * OpenIB.org BSD license below:
+ *
+ *     Redistribution and use in source and binary forms, with or
+ *     without modification, are permitted provided that the following
+ *     conditions are met:
+ *
+ *      - Redistributions of source code must retain the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer.
+ *
+ *      - Redistributions in binary form must reproduce the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer in the documentation and/or other materials
+ *        provided with the distribution.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+ * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+
+#include <rdma/ib_verbs.h>
+#include <linux/mlx5/fs.h>
+#include "en.h"
+#include "ipoib.h"
+
+#define IB_DEFAULT_Q_KEY   0xb1b
+
+static int mlx5i_open(struct net_device *netdev);
+static int mlx5i_close(struct net_device *netdev);
+static int  mlx5i_dev_init(struct net_device *dev);
+static void mlx5i_dev_cleanup(struct net_device *dev);
+
+static const struct net_device_ops mlx5i_netdev_ops = {
+       .ndo_open                = mlx5i_open,
+       .ndo_stop                = mlx5i_close,
+       .ndo_init                = mlx5i_dev_init,
+       .ndo_uninit              = mlx5i_dev_cleanup,
+};
+
+/* IPoIB mlx5 netdev profile */
+
+/* Called directly after IPoIB netdevice was created to initialize SW structs */
+static void mlx5i_init(struct mlx5_core_dev *mdev,
+                      struct net_device *netdev,
+                      const struct mlx5e_profile *profile,
+                      void *ppriv)
+{
+       struct mlx5e_priv *priv  = mlx5i_epriv(netdev);
+
+       priv->mdev        = mdev;
+       priv->netdev      = netdev;
+       priv->profile     = profile;
+       priv->ppriv       = ppriv;
+
+       mlx5e_build_nic_params(mdev, &priv->channels.params, profile->max_nch(mdev));
+
+       /* Override RQ params as IPoIB supports only LINKED LIST RQ for now */
+       mlx5e_set_rq_type_params(mdev, &priv->channels.params, MLX5_WQ_TYPE_LINKED_LIST);
+       priv->channels.params.lro_en = false;
+
+       mutex_init(&priv->state_lock);
+
+       netdev->hw_features    |= NETIF_F_SG;
+       netdev->hw_features    |= NETIF_F_IP_CSUM;
+       netdev->hw_features    |= NETIF_F_IPV6_CSUM;
+       netdev->hw_features    |= NETIF_F_GRO;
+       netdev->hw_features    |= NETIF_F_TSO;
+       netdev->hw_features    |= NETIF_F_TSO6;
+       netdev->hw_features    |= NETIF_F_RXCSUM;
+       netdev->hw_features    |= NETIF_F_RXHASH;
+
+       netdev->netdev_ops = &mlx5i_netdev_ops;
+}
+
+/* Called directly before IPoIB netdevice is destroyed to cleanup SW structs */
+static void mlx5i_cleanup(struct mlx5e_priv *priv)
+{
+       /* Do nothing .. */
+}
+
+#define MLX5_QP_ENHANCED_ULP_STATELESS_MODE 2
+
+static int mlx5i_create_underlay_qp(struct mlx5_core_dev *mdev, struct mlx5_core_qp *qp)
+{
+       struct mlx5_qp_context *context = NULL;
+       u32 *in = NULL;
+       void *addr_path;
+       int ret = 0;
+       int inlen;
+       void *qpc;
+
+       inlen = MLX5_ST_SZ_BYTES(create_qp_in);
+       in = kvzalloc(inlen, GFP_KERNEL);
+       if (!in)
+               return -ENOMEM;
+
+       qpc = MLX5_ADDR_OF(create_qp_in, in, qpc);
+       MLX5_SET(qpc, qpc, st, MLX5_QP_ST_UD);
+       MLX5_SET(qpc, qpc, pm_state, MLX5_QP_PM_MIGRATED);
+       MLX5_SET(qpc, qpc, ulp_stateless_offload_mode,
+                MLX5_QP_ENHANCED_ULP_STATELESS_MODE);
+
+       addr_path = MLX5_ADDR_OF(qpc, qpc, primary_address_path);
+       MLX5_SET(ads, addr_path, port, 1);
+       MLX5_SET(ads, addr_path, grh, 1);
+
+       ret = mlx5_core_create_qp(mdev, qp, in, inlen);
+       if (ret) {
+               mlx5_core_err(mdev, "Failed creating IPoIB QP err : %d\n", ret);
+               goto out;
+       }
+
+       /* QP states */
+       context = kzalloc(sizeof(*context), GFP_KERNEL);
+       if (!context) {
+               ret = -ENOMEM;
+               goto out;
+       }
+
+       context->flags = cpu_to_be32(MLX5_QP_PM_MIGRATED << 11);
+       context->pri_path.port = 1;
+       context->qkey = cpu_to_be32(IB_DEFAULT_Q_KEY);
+
+       ret = mlx5_core_qp_modify(mdev, MLX5_CMD_OP_RST2INIT_QP, 0, context, qp);
+       if (ret) {
+               mlx5_core_err(mdev, "Failed to modify qp RST2INIT, err: %d\n", ret);
+               goto out;
+       }
+       memset(context, 0, sizeof(*context));
+
+       ret = mlx5_core_qp_modify(mdev, MLX5_CMD_OP_INIT2RTR_QP, 0, context, qp);
+       if (ret) {
+               mlx5_core_err(mdev, "Failed to modify qp INIT2RTR, err: %d\n", ret);
+               goto out;
+       }
+
+       ret = mlx5_core_qp_modify(mdev, MLX5_CMD_OP_RTR2RTS_QP, 0, context, qp);
+       if (ret) {
+               mlx5_core_err(mdev, "Failed to modify qp RTR2RTS, err: %d\n", ret);
+               goto out;
+       }
+
+out:
+       kfree(context);
+       kvfree(in);
+       return ret;
+}
+
+static void mlx5i_destroy_underlay_qp(struct mlx5_core_dev *mdev, struct mlx5_core_qp *qp)
+{
+       mlx5_fs_remove_rx_underlay_qpn(mdev, qp->qpn);
+
+       mlx5_core_destroy_qp(mdev, qp);
+}
+
+static int mlx5i_init_tx(struct mlx5e_priv *priv)
+{
+       struct mlx5i_priv *ipriv = priv->ppriv;
+       int err;
+
+       err = mlx5i_create_underlay_qp(priv->mdev, &ipriv->qp);
+       if (err) {
+               mlx5_core_warn(priv->mdev, "create underlay QP failed, %d\n", err);
+               return err;
+       }
+
+       mlx5_fs_add_rx_underlay_qpn(priv->mdev, ipriv->qp.qpn);
+
+       err = mlx5e_create_tis(priv->mdev, 0 /* tc */, ipriv->qp.qpn, &priv->tisn[0]);
+       if (err) {
+               mlx5_core_warn(priv->mdev, "create tis failed, %d\n", err);
+               return err;
+       }
+
+       return 0;
+}
+
+static void mlx5i_cleanup_tx(struct mlx5e_priv *priv)
+{
+       struct mlx5i_priv *ipriv = priv->ppriv;
+
+       mlx5e_destroy_tis(priv->mdev, priv->tisn[0]);
+       mlx5i_destroy_underlay_qp(priv->mdev, &ipriv->qp);
+}
+
+static int mlx5i_create_flow_steering(struct mlx5e_priv *priv)
+{
+       int err;
+
+       priv->fs.ns = mlx5_get_flow_namespace(priv->mdev,
+                                              MLX5_FLOW_NAMESPACE_KERNEL);
+
+       if (!priv->fs.ns)
+               return -EINVAL;
+
+       err = mlx5e_arfs_create_tables(priv);
+       if (err) {
+               netdev_err(priv->netdev, "Failed to create arfs tables, err=%d\n",
+                          err);
+               priv->netdev->hw_features &= ~NETIF_F_NTUPLE;
+       }
+
+       err = mlx5e_create_ttc_table(priv);
+       if (err) {
+               netdev_err(priv->netdev, "Failed to create ttc table, err=%d\n",
+                          err);
+               goto err_destroy_arfs_tables;
+       }
+
+       return 0;
+
+err_destroy_arfs_tables:
+       mlx5e_arfs_destroy_tables(priv);
+
+       return err;
+}
+
+static void mlx5i_destroy_flow_steering(struct mlx5e_priv *priv)
+{
+       mlx5e_destroy_ttc_table(priv);
+       mlx5e_arfs_destroy_tables(priv);
+}
+
+static int mlx5i_init_rx(struct mlx5e_priv *priv)
+{
+       int err;
+
+       err = mlx5e_create_indirect_rqt(priv);
+       if (err)
+               return err;
+
+       err = mlx5e_create_direct_rqts(priv);
+       if (err)
+               goto err_destroy_indirect_rqts;
+
+       err = mlx5e_create_indirect_tirs(priv);
+       if (err)
+               goto err_destroy_direct_rqts;
+
+       err = mlx5e_create_direct_tirs(priv);
+       if (err)
+               goto err_destroy_indirect_tirs;
+
+       err = mlx5i_create_flow_steering(priv);
+       if (err)
+               goto err_destroy_direct_tirs;
+
+       return 0;
+
+err_destroy_direct_tirs:
+       mlx5e_destroy_direct_tirs(priv);
+err_destroy_indirect_tirs:
+       mlx5e_destroy_indirect_tirs(priv);
+err_destroy_direct_rqts:
+       mlx5e_destroy_direct_rqts(priv);
+err_destroy_indirect_rqts:
+       mlx5e_destroy_rqt(priv, &priv->indir_rqt);
+       return err;
+}
+
+static void mlx5i_cleanup_rx(struct mlx5e_priv *priv)
+{
+       mlx5i_destroy_flow_steering(priv);
+       mlx5e_destroy_direct_tirs(priv);
+       mlx5e_destroy_indirect_tirs(priv);
+       mlx5e_destroy_direct_rqts(priv);
+       mlx5e_destroy_rqt(priv, &priv->indir_rqt);
+}
+
+static const struct mlx5e_profile mlx5i_nic_profile = {
+       .init              = mlx5i_init,
+       .cleanup           = mlx5i_cleanup,
+       .init_tx           = mlx5i_init_tx,
+       .cleanup_tx        = mlx5i_cleanup_tx,
+       .init_rx           = mlx5i_init_rx,
+       .cleanup_rx        = mlx5i_cleanup_rx,
+       .enable            = NULL, /* mlx5i_enable */
+       .disable           = NULL, /* mlx5i_disable */
+       .update_stats      = NULL, /* mlx5i_update_stats */
+       .max_nch           = mlx5e_get_max_num_channels,
+       .rx_handlers.handle_rx_cqe       = mlx5i_handle_rx_cqe,
+       .rx_handlers.handle_rx_cqe_mpwqe = NULL, /* Not supported */
+       .max_tc            = MLX5I_MAX_NUM_TC,
+};
+
+/* mlx5i netdev NDos */
+
+static int mlx5i_dev_init(struct net_device *dev)
+{
+       struct mlx5e_priv    *priv   = mlx5i_epriv(dev);
+       struct mlx5i_priv    *ipriv  = priv->ppriv;
+
+       /* Set dev address using underlay QP */
+       dev->dev_addr[1] = (ipriv->qp.qpn >> 16) & 0xff;
+       dev->dev_addr[2] = (ipriv->qp.qpn >>  8) & 0xff;
+       dev->dev_addr[3] = (ipriv->qp.qpn) & 0xff;
+
+       return 0;
+}
+
+static void mlx5i_dev_cleanup(struct net_device *dev)
+{
+       struct mlx5e_priv    *priv   = mlx5i_epriv(dev);
+       struct mlx5_core_dev *mdev   = priv->mdev;
+       struct mlx5i_priv    *ipriv  = priv->ppriv;
+       struct mlx5_qp_context context;
+
+       /* detach qp from flow-steering by reset it */
+       mlx5_core_qp_modify(mdev, MLX5_CMD_OP_2RST_QP, 0, &context, &ipriv->qp);
+}
+
+static int mlx5i_open(struct net_device *netdev)
+{
+       struct mlx5e_priv *priv = mlx5i_epriv(netdev);
+       int err;
+
+       mutex_lock(&priv->state_lock);
+
+       set_bit(MLX5E_STATE_OPENED, &priv->state);
+
+       err = mlx5e_open_channels(priv, &priv->channels);
+       if (err)
+               goto err_clear_state_opened_flag;
+
+       mlx5e_refresh_tirs(priv, false);
+       mlx5e_activate_priv_channels(priv);
+       mutex_unlock(&priv->state_lock);
+       return 0;
+
+err_clear_state_opened_flag:
+       clear_bit(MLX5E_STATE_OPENED, &priv->state);
+       mutex_unlock(&priv->state_lock);
+       return err;
+}
+
+static int mlx5i_close(struct net_device *netdev)
+{
+       struct mlx5e_priv *priv = mlx5i_epriv(netdev);
+
+       /* May already be CLOSED in case a previous configuration operation
+        * (e.g RX/TX queue size change) that involves close&open failed.
+        */
+       mutex_lock(&priv->state_lock);
+
+       if (!test_bit(MLX5E_STATE_OPENED, &priv->state))
+               goto unlock;
+
+       clear_bit(MLX5E_STATE_OPENED, &priv->state);
+
+       netif_carrier_off(priv->netdev);
+       mlx5e_deactivate_priv_channels(priv);
+       mlx5e_close_channels(&priv->channels);
+unlock:
+       mutex_unlock(&priv->state_lock);
+       return 0;
+}
+
+/* IPoIB RDMA netdev callbacks */
+static int mlx5i_attach_mcast(struct net_device *netdev, struct ib_device *hca,
+                             union ib_gid *gid, u16 lid, int set_qkey,
+                             u32 qkey)
+{
+       struct mlx5e_priv    *epriv = mlx5i_epriv(netdev);
+       struct mlx5_core_dev *mdev  = epriv->mdev;
+       struct mlx5i_priv    *ipriv = epriv->ppriv;
+       int err;
+
+       mlx5_core_dbg(mdev, "attaching QPN 0x%x, MGID %pI6\n", ipriv->qp.qpn, gid->raw);
+       err = mlx5_core_attach_mcg(mdev, gid, ipriv->qp.qpn);
+       if (err)
+               mlx5_core_warn(mdev, "failed attaching QPN 0x%x, MGID %pI6\n",
+                              ipriv->qp.qpn, gid->raw);
+
+       if (set_qkey) {
+               mlx5_core_dbg(mdev, "%s setting qkey 0x%x\n",
+                             netdev->name, qkey);
+               ipriv->qkey = qkey;
+       }
+
+       return err;
+}
+
+static int mlx5i_detach_mcast(struct net_device *netdev, struct ib_device *hca,
+                             union ib_gid *gid, u16 lid)
+{
+       struct mlx5e_priv    *epriv = mlx5i_epriv(netdev);
+       struct mlx5_core_dev *mdev  = epriv->mdev;
+       struct mlx5i_priv    *ipriv = epriv->ppriv;
+       int err;
+
+       mlx5_core_dbg(mdev, "detaching QPN 0x%x, MGID %pI6\n", ipriv->qp.qpn, gid->raw);
+
+       err = mlx5_core_detach_mcg(mdev, gid, ipriv->qp.qpn);
+       if (err)
+               mlx5_core_dbg(mdev, "failed dettaching QPN 0x%x, MGID %pI6\n",
+                             ipriv->qp.qpn, gid->raw);
+
+       return err;
+}
+
+static int mlx5i_xmit(struct net_device *dev, struct sk_buff *skb,
+                     struct ib_ah *address, u32 dqpn)
+{
+       struct mlx5e_priv *epriv = mlx5i_epriv(dev);
+       struct mlx5e_txqsq *sq   = epriv->txq2sq[skb_get_queue_mapping(skb)];
+       struct mlx5_ib_ah *mah   = to_mah(address);
+       struct mlx5i_priv *ipriv = epriv->ppriv;
+
+       return mlx5i_sq_xmit(sq, skb, &mah->av, dqpn, ipriv->qkey);
+}
+
+static int mlx5i_check_required_hca_cap(struct mlx5_core_dev *mdev)
+{
+       if (MLX5_CAP_GEN(mdev, port_type) != MLX5_CAP_PORT_TYPE_IB)
+               return -EOPNOTSUPP;
+
+       if (!MLX5_CAP_GEN(mdev, ipoib_enhanced_offloads)) {
+               mlx5_core_warn(mdev, "IPoIB enhanced offloads are not supported\n");
+               return -EOPNOTSUPP;
+       }
+
+       return 0;
+}
+
+struct net_device *mlx5_rdma_netdev_alloc(struct mlx5_core_dev *mdev,
+                                         struct ib_device *ibdev,
+                                         const char *name,
+                                         void (*setup)(struct net_device *))
+{
+       const struct mlx5e_profile *profile = &mlx5i_nic_profile;
+       int nch = profile->max_nch(mdev);
+       struct net_device *netdev;
+       struct mlx5i_priv *ipriv;
+       struct mlx5e_priv *epriv;
+       struct rdma_netdev *rn;
+       int err;
+
+       if (mlx5i_check_required_hca_cap(mdev)) {
+               mlx5_core_warn(mdev, "Accelerated mode is not supported\n");
+               return ERR_PTR(-EOPNOTSUPP);
+       }
+
+       /* This function should only be called once per mdev */
+       err = mlx5e_create_mdev_resources(mdev);
+       if (err)
+               return NULL;
+
+       netdev = alloc_netdev_mqs(sizeof(struct mlx5i_priv) + sizeof(struct mlx5e_priv),
+                                 name, NET_NAME_UNKNOWN,
+                                 setup,
+                                 nch * MLX5E_MAX_NUM_TC,
+                                 nch);
+       if (!netdev) {
+               mlx5_core_warn(mdev, "alloc_netdev_mqs failed\n");
+               goto free_mdev_resources;
+       }
+
+       ipriv = netdev_priv(netdev);
+       epriv = mlx5i_epriv(netdev);
+
+       epriv->wq = create_singlethread_workqueue("mlx5i");
+       if (!epriv->wq)
+               goto err_free_netdev;
+
+       profile->init(mdev, netdev, profile, ipriv);
+
+       mlx5e_attach_netdev(epriv);
+       netif_carrier_off(netdev);
+
+       /* set rdma_netdev func pointers */
+       rn = &ipriv->rn;
+       rn->hca  = ibdev;
+       rn->send = mlx5i_xmit;
+       rn->attach_mcast = mlx5i_attach_mcast;
+       rn->detach_mcast = mlx5i_detach_mcast;
+
+       return netdev;
+
+err_free_netdev:
+       free_netdev(netdev);
+free_mdev_resources:
+       mlx5e_destroy_mdev_resources(mdev);
+
+       return NULL;
+}
+EXPORT_SYMBOL(mlx5_rdma_netdev_alloc);
+
+void mlx5_rdma_netdev_free(struct net_device *netdev)
+{
+       struct mlx5e_priv          *priv    = mlx5i_epriv(netdev);
+       const struct mlx5e_profile *profile = priv->profile;
+
+       mlx5e_detach_netdev(priv);
+       profile->cleanup(priv);
+       destroy_workqueue(priv->wq);
+       free_netdev(netdev);
+
+       mlx5e_destroy_mdev_resources(priv->mdev);
+}
+EXPORT_SYMBOL(mlx5_rdma_netdev_free);
+
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/ipoib/ipoib.h b/drivers/net/ethernet/mellanox/mlx5/core/ipoib/ipoib.h
new file mode 100644 (file)
index 0000000..213191a
--- /dev/null
@@ -0,0 +1,56 @@
+/*
+ * Copyright (c) 2017, Mellanox Technologies. All rights reserved.
+ *
+ * This software is available to you under a choice of one of two
+ * licenses.  You may choose to be licensed under the terms of the GNU
+ * General Public License (GPL) Version 2, available from the file
+ * COPYING in the main directory of this source tree, or the
+ * OpenIB.org BSD license below:
+ *
+ *     Redistribution and use in source and binary forms, with or
+ *     without modification, are permitted provided that the following
+ *     conditions are met:
+ *
+ *      - Redistributions of source code must retain the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer.
+ *
+ *      - Redistributions in binary form must reproduce the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer in the documentation and/or other materials
+ *        provided with the distribution.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+ * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+
+#ifndef __MLX5E_IPOB_H__
+#define __MLX5E_IPOB_H__
+
+#include <linux/mlx5/fs.h>
+#include "en.h"
+
+#define MLX5I_MAX_NUM_TC 1
+
+/* ipoib rdma netdev's private data structure */
+struct mlx5i_priv {
+       struct rdma_netdev rn; /* keep this first */
+       struct mlx5_core_qp qp;
+       u32    qkey;
+       char  *mlx5e_priv[0];
+};
+
+/* Extract mlx5e_priv from IPoIB netdev */
+#define mlx5i_epriv(netdev) ((void *)(((struct mlx5i_priv *)netdev_priv(netdev))->mlx5e_priv))
+
+netdev_tx_t mlx5i_sq_xmit(struct mlx5e_txqsq *sq, struct sk_buff *skb,
+                         struct mlx5_av *av, u32 dqpn, u32 dqkey);
+void mlx5i_handle_rx_cqe(struct mlx5e_rq *rq, struct mlx5_cqe64 *cqe);
+
+#endif /* __MLX5E_IPOB_H__ */