Merge tag 'mlx5e-updates-2018-10-10' of git://git.kernel.org/pub/scm/linux/kernel/git/saeed/linux

Saeed Mahameed says:

====================
mlx5e-updates-2018-10-10

IPoIB netlink support and mlx5e pre-allocated netdevice initialization

IP link was broken due to the changes in IPoIB for the rdma_netdev
support after commit cd565b4b
("IB/IPoIB: Support acceleration options callbacks").

This patchset fixes IPoIB pkey creation and removal using rtnetlink by
adding support in both IPoIB ULP layer and mlx5 layer:

From Jason and Denis:
1) Introduces changes in the RDMA netdev code in order to
   allow allocation of the netdev to be done by the rtnl netdev code.
2) Reworks IPoIB initialization to use the two step rdma_netdev
   creation.

From Feras and Saeed, mlx5e netdev layer refactoring to allow accepting
pre-allocated netdevs:
3) Adds support to initialize/cleanup netdevs that are not created
   by mlx5 driver.
4) Change mlx5e netdevice layer to accept the pre-allocated netdevice
   queue number.
5) Initialize mlx5e generic structures in one place to be used for all
   netdevs types NIC/representors/IPoIB (both mlx5 allocated and
   pre-allocted).
====================
Signed-off-by: David S. Miller <davem@davemloft.net>

drivers/infiniband/core/verbs.c

@@ -2621,3 +2621,49 @@ void ib_drain_qp(struct ib_qp *qp)
 		ib_drain_rq(qp);
 }
 EXPORT_SYMBOL(ib_drain_qp);
+
+struct net_device *rdma_alloc_netdev(struct ib_device *device, u8 port_num,
+				     enum rdma_netdev_t type, const char *name,
+				     unsigned char name_assign_type,
+				     void (*setup)(struct net_device *))
+{
+	struct rdma_netdev_alloc_params params;
+	struct net_device *netdev;
+	int rc;
+
+	if (!device->rdma_netdev_get_params)
+		return ERR_PTR(-EOPNOTSUPP);
+
+	rc = device->rdma_netdev_get_params(device, port_num, type, &params);
+	if (rc)
+		return ERR_PTR(rc);
+
+	netdev = alloc_netdev_mqs(params.sizeof_priv, name, name_assign_type,
+				  setup, params.txqs, params.rxqs);
+	if (!netdev)
+		return ERR_PTR(-ENOMEM);
+
+	return netdev;
+}
+EXPORT_SYMBOL(rdma_alloc_netdev);
+
+int rdma_init_netdev(struct ib_device *device, u8 port_num,
+		     enum rdma_netdev_t type, const char *name,
+		     unsigned char name_assign_type,
+		     void (*setup)(struct net_device *),
+		     struct net_device *netdev)
+{
+	struct rdma_netdev_alloc_params params;
+	int rc;
+
+	if (!device->rdma_netdev_get_params)
+		return -EOPNOTSUPP;
+
+	rc = device->rdma_netdev_get_params(device, port_num, type, &params);
+	if (rc)
+		return rc;
+
+	return params.initialize_rdma_netdev(device, port_num,
+					     netdev, params.param);
+}
+EXPORT_SYMBOL(rdma_init_netdev);

drivers/infiniband/hw/mlx5/main.c

@@ -5163,22 +5163,14 @@ static int mlx5_ib_get_hw_stats(struct ib_device *ibdev,
 	return num_counters;
 }
 
-static struct net_device*
-mlx5_ib_alloc_rdma_netdev(struct ib_device *hca,
-			  u8 port_num,
-			  enum rdma_netdev_t type,
-			  const char *name,
-			  unsigned char name_assign_type,
-			  void (*setup)(struct net_device *))
+static int mlx5_ib_rn_get_params(struct ib_device *device, u8 port_num,
+				 enum rdma_netdev_t type,
+				 struct rdma_netdev_alloc_params *params)
 {
-	struct net_device *netdev;
-
 	if (type != RDMA_NETDEV_IPOIB)
-		return ERR_PTR(-EOPNOTSUPP);
+		return -EOPNOTSUPP;
 
-	netdev = mlx5_rdma_netdev_alloc(to_mdev(hca)->mdev, hca,
-					name, setup);
-	return netdev;
+	return mlx5_rdma_rn_get_params(to_mdev(device)->mdev, device, params);
 }
 
 static void delay_drop_debugfs_cleanup(struct mlx5_ib_dev *dev)
@@ -5824,8 +5816,9 @@ int mlx5_ib_stage_caps_init(struct mlx5_ib_dev *dev)
 	dev->ib_dev.check_mr_status	= mlx5_ib_check_mr_status;
 	dev->ib_dev.get_dev_fw_str      = get_dev_fw_str;
 	dev->ib_dev.get_vector_affinity	= mlx5_ib_get_vector_affinity;
-	if (MLX5_CAP_GEN(mdev, ipoib_enhanced_offloads))
-		dev->ib_dev.alloc_rdma_netdev	= mlx5_ib_alloc_rdma_netdev;
+	if (MLX5_CAP_GEN(mdev, ipoib_enhanced_offloads) &&
+	    IS_ENABLED(CONFIG_MLX5_CORE_IPOIB))
+		dev->ib_dev.rdma_netdev_get_params = mlx5_ib_rn_get_params;
 
 	if (mlx5_core_is_pf(mdev)) {
 		dev->ib_dev.get_vf_config	= mlx5_ib_get_vf_config;

drivers/infiniband/ulp/ipoib/ipoib.h

@@ -499,8 +499,10 @@ void ipoib_reap_ah(struct work_struct *work);
 struct ipoib_path *__path_find(struct net_device *dev, void *gid);
 void ipoib_mark_paths_invalid(struct net_device *dev);
 void ipoib_flush_paths(struct net_device *dev);
-struct ipoib_dev_priv *ipoib_intf_alloc(struct ib_device *hca, u8 port,
-					const char *format);
+struct net_device *ipoib_intf_alloc(struct ib_device *hca, u8 port,
+				    const char *format);
+int ipoib_intf_init(struct ib_device *hca, u8 port, const char *format,
+		    struct net_device *dev);
 void ipoib_ib_tx_timer_func(struct timer_list *t);
 void ipoib_ib_dev_flush_light(struct work_struct *work);
 void ipoib_ib_dev_flush_normal(struct work_struct *work);
@@ -531,6 +533,8 @@ int ipoib_dma_map_tx(struct ib_device *ca, struct ipoib_tx_buf *tx_req);
 void ipoib_dma_unmap_tx(struct ipoib_dev_priv *priv,
 			struct ipoib_tx_buf *tx_req);
 
+struct rtnl_link_ops *ipoib_get_link_ops(void);
+
 static inline void ipoib_build_sge(struct ipoib_dev_priv *priv,
 				   struct ipoib_tx_buf *tx_req)
 {

drivers/infiniband/ulp/ipoib/ipoib_main.c

@@ -2115,82 +2115,58 @@ static const struct net_device_ops ipoib_netdev_default_pf = {
 	.ndo_stop		 = ipoib_ib_dev_stop_default,
 };
 
-static struct net_device
-*ipoib_create_netdev_default(struct ib_device *hca,
-			     const char *name,
-			     unsigned char name_assign_type,
-			     void (*setup)(struct net_device *))
+static struct net_device *ipoib_alloc_netdev(struct ib_device *hca, u8 port,
+					     const char *name)
 {
 	struct net_device *dev;
-	struct rdma_netdev *rn;
 
-	dev = alloc_netdev((int)sizeof(struct rdma_netdev),
-			   name,
-			   name_assign_type, setup);
-	if (!dev)
-		return NULL;
-
-	rn = netdev_priv(dev);
-
-	rn->send = ipoib_send;
-	rn->attach_mcast = ipoib_mcast_attach;
-	rn->detach_mcast = ipoib_mcast_detach;
-	rn->hca = hca;
-	dev->netdev_ops = &ipoib_netdev_default_pf;
-
-	return dev;
-}
-
-static struct net_device *ipoib_get_netdev(struct ib_device *hca, u8 port,
-					   const char *name)
-{
-	struct net_device *dev;
-
-	if (hca->alloc_rdma_netdev) {
-		dev = hca->alloc_rdma_netdev(hca, port,
-					     RDMA_NETDEV_IPOIB, name,
-					     NET_NAME_UNKNOWN,
-					     ipoib_setup_common);
-		if (IS_ERR_OR_NULL(dev) && PTR_ERR(dev) != -EOPNOTSUPP)
-			return NULL;
-	}
-
-	if (!hca->alloc_rdma_netdev || PTR_ERR(dev) == -EOPNOTSUPP)
-		dev = ipoib_create_netdev_default(hca, name, NET_NAME_UNKNOWN,
-						  ipoib_setup_common);
+	dev = rdma_alloc_netdev(hca, port, RDMA_NETDEV_IPOIB, name,
+				NET_NAME_UNKNOWN, ipoib_setup_common);
+	if (!IS_ERR(dev) || PTR_ERR(dev) != -EOPNOTSUPP)
+		return dev;
 
+	dev = alloc_netdev(sizeof(struct rdma_netdev), name, NET_NAME_UNKNOWN,
+			   ipoib_setup_common);
+	if (!dev)
+		return ERR_PTR(-ENOMEM);
 	return dev;
 }
 
-struct ipoib_dev_priv *ipoib_intf_alloc(struct ib_device *hca, u8 port,
-					const char *name)
+int ipoib_intf_init(struct ib_device *hca, u8 port, const char *name,
+		    struct net_device *dev)
 {
-	struct net_device *dev;
+	struct rdma_netdev *rn = netdev_priv(dev);
 	struct ipoib_dev_priv *priv;
-	struct rdma_netdev *rn;
+	int rc;
 
 	priv = kzalloc(sizeof(*priv), GFP_KERNEL);
 	if (!priv)
-		return NULL;
+		return -ENOMEM;
 
 	priv->ca = hca;
 	priv->port = port;
 
-	dev = ipoib_get_netdev(hca, port, name);
-	if (!dev)
-		goto free_priv;
+	rc = rdma_init_netdev(hca, port, RDMA_NETDEV_IPOIB, name,
+			      NET_NAME_UNKNOWN, ipoib_setup_common, dev);
+	if (rc) {
+		if (rc != -EOPNOTSUPP)
+			goto out;
+
+		dev->netdev_ops = &ipoib_netdev_default_pf;
+		rn->send = ipoib_send;
+		rn->attach_mcast = ipoib_mcast_attach;
+		rn->detach_mcast = ipoib_mcast_detach;
+		rn->hca = hca;
+	}
 
 	priv->rn_ops = dev->netdev_ops;
 
-	/* fixme : should be after the query_cap */
-	if (priv->hca_caps & IB_DEVICE_VIRTUAL_FUNCTION)
+	if (hca->attrs.device_cap_flags & IB_DEVICE_VIRTUAL_FUNCTION)
 		dev->netdev_ops	= &ipoib_netdev_ops_vf;
 	else
 		dev->netdev_ops	= &ipoib_netdev_ops_pf;
 
-	rn = netdev_priv(dev);
 	rn->clnt_priv = priv;
-
 	/*
 	 * Only the child register_netdev flows can handle priv_destructor
 	 * being set, so we force it to NULL here and handle manually until it
@@ -2201,10 +2177,35 @@ struct ipoib_dev_priv *ipoib_intf_alloc(struct ib_device *hca, u8 port,
 
 	ipoib_build_priv(dev);
 
-	return priv;
-free_priv:
+	return 0;
+
+out:
 	kfree(priv);
-	return NULL;
+	return rc;
+}
+
+struct net_device *ipoib_intf_alloc(struct ib_device *hca, u8 port,
+				    const char *name)
+{
+	struct net_device *dev;
+	int rc;
+
+	dev = ipoib_alloc_netdev(hca, port, name);
+	if (IS_ERR(dev))
+		return dev;
+
+	rc = ipoib_intf_init(hca, port, name, dev);
+	if (rc) {
+		free_netdev(dev);
+		return ERR_PTR(rc);
+	}
+
+	/*
+	 * Upon success the caller must ensure ipoib_intf_free is called or
+	 * register_netdevice succeed'd and priv_destructor is set to
+	 * ipoib_intf_free.
+	 */
+	return dev;
 }
 
 void ipoib_intf_free(struct net_device *dev)
@@ -2387,16 +2388,19 @@ int ipoib_add_pkey_attr(struct net_device *dev)
 static struct net_device *ipoib_add_port(const char *format,
 					 struct ib_device *hca, u8 port)
 {
+	struct rtnl_link_ops *ops = ipoib_get_link_ops();
+	struct rdma_netdev_alloc_params params;
 	struct ipoib_dev_priv *priv;
 	struct net_device *ndev;
 	int result;
 
-	priv = ipoib_intf_alloc(hca, port, format);
-	if (!priv) {
-		pr_warn("%s, %d: ipoib_intf_alloc failed\n", hca->name, port);
-		return ERR_PTR(-ENOMEM);
+	ndev = ipoib_intf_alloc(hca, port, format);
+	if (IS_ERR(ndev)) {
+		pr_warn("%s, %d: ipoib_intf_alloc failed %ld\n", hca->name, port,
+			PTR_ERR(ndev));
+		return ndev;
 	}
-	ndev = priv->dev;
+	priv = ipoib_priv(ndev);
 
 	INIT_IB_EVENT_HANDLER(&priv->event_handler,
 			      priv->ca, ipoib_event);
@@ -2417,6 +2421,14 @@ static struct net_device *ipoib_add_port(const char *format,
 		return ERR_PTR(result);
 	}
 
+	if (hca->rdma_netdev_get_params) {
+		int rc = hca->rdma_netdev_get_params(hca, port,
+						     RDMA_NETDEV_IPOIB,
+						     &params);
+
+		if (!rc && ops->priv_size < params.sizeof_priv)
+			ops->priv_size = params.sizeof_priv;
+	}
 	/*
 	 * We cannot set priv_destructor before register_netdev because we
 	 * need priv to be always valid during the error flow to execute

drivers/infiniband/ulp/ipoib/ipoib_netlink.c

@@ -122,12 +122,26 @@ static int ipoib_new_child_link(struct net *src_net, struct net_device *dev,
 	} else
 		child_pkey  = nla_get_u16(data[IFLA_IPOIB_PKEY]);
 
+	err = ipoib_intf_init(ppriv->ca, ppriv->port, dev->name, dev);
+	if (err) {
+		ipoib_warn(ppriv, "failed to initialize pkey device\n");
+		return err;
+	}
+
 	err = __ipoib_vlan_add(ppriv, ipoib_priv(dev),
 			       child_pkey, IPOIB_RTNL_CHILD);
+	if (err)
+		return err;
 
-	if (!err && data)
+	if (data) {
 		err = ipoib_changelink(dev, tb, data, extack);
-	return err;
+		if (err) {
+			unregister_netdevice(dev);
+			return err;
+		}
+	}
+
+	return 0;
 }
 
 static size_t ipoib_get_size(const struct net_device *dev)
@@ -149,6 +163,11 @@ static struct rtnl_link_ops ipoib_link_ops __read_mostly = {
 	.fill_info	= ipoib_fill_info,
 };
 
+struct rtnl_link_ops *ipoib_get_link_ops(void)
+{
+	return &ipoib_link_ops;
+}
+
 int __init ipoib_netlink_init(void)
 {
 	return rtnl_link_register(&ipoib_link_ops);

drivers/infiniband/ulp/ipoib/ipoib_vlan.c

@@ -85,7 +85,7 @@ static bool is_child_unique(struct ipoib_dev_priv *ppriv,
 
 /*
  * NOTE: If this function fails then the priv->dev will remain valid, however
- * priv can have been freed and must not be touched by caller in the error
+ * priv will have been freed and must not be touched by caller in the error
  * case.
  *
  * If (ndev->reg_state == NETREG_UNINITIALIZED) then it is up to the caller to
@@ -100,6 +100,12 @@ int __ipoib_vlan_add(struct ipoib_dev_priv *ppriv, struct ipoib_dev_priv *priv,
 
 	ASSERT_RTNL();
 
+	/*
+	 * We do not need to touch priv if register_netdevice fails, so just
+	 * always use this flow.
+	 */
+	ndev->priv_destructor = ipoib_intf_free;
+
 	/*
 	 * Racing with unregister of the parent must be prevented by the
 	 * caller.
@@ -120,9 +126,6 @@ int __ipoib_vlan_add(struct ipoib_dev_priv *ppriv, struct ipoib_dev_priv *priv,
 		goto out_early;
 	}
 
-	/* We do not need to touch priv if register_netdevice fails */
-	ndev->priv_destructor = ipoib_intf_free;
-
 	result = register_netdevice(ndev);
 	if (result) {
 		ipoib_warn(priv, "failed to initialize; error %i", result);
@@ -182,12 +185,12 @@ int ipoib_vlan_add(struct net_device *pdev, unsigned short pkey)
 	snprintf(intf_name, sizeof(intf_name), "%s.%04x",
 		 ppriv->dev->name, pkey);
 
-	priv = ipoib_intf_alloc(ppriv->ca, ppriv->port, intf_name);
-	if (!priv) {
-		result = -ENOMEM;
+	ndev = ipoib_intf_alloc(ppriv->ca, ppriv->port, intf_name);
+	if (IS_ERR(ndev)) {
+		result = PTR_ERR(ndev);
 		goto out;
 	}
-	ndev = priv->dev;
+	priv = ipoib_priv(ndev);
 
 	result = __ipoib_vlan_add(ppriv, priv, pkey, IPOIB_LEGACY_CHILD);
 

drivers/net/ethernet/mellanox/mlx5/core/en.h

@@ -173,6 +173,7 @@ static inline u16 mlx5_min_rx_wqes(int wq_type, u32 wq_size)
 	}
 }
 
+/* Use this function to get max num channels (rxqs/txqs) only to create netdev */
 static inline int mlx5e_get_max_num_channels(struct mlx5_core_dev *mdev)
 {
 	return is_kdump_kernel() ?
@@ -181,6 +182,13 @@ static inline int mlx5e_get_max_num_channels(struct mlx5_core_dev *mdev)
 		      MLX5E_MAX_NUM_CHANNELS);
 }
 
+/* Use this function to get max num channels after netdev was created */
+static inline int mlx5e_get_netdev_max_channels(struct net_device *netdev)
+{
+	return min_t(unsigned int, netdev->num_rx_queues,
+		     netdev->num_tx_queues);
+}
+
 struct mlx5e_tx_wqe {
 	struct mlx5_wqe_ctrl_seg ctrl;
 	struct mlx5_wqe_eth_seg  eth;
@@ -673,7 +681,7 @@ struct mlx5e_priv {
 	struct work_struct         update_carrier_work;
 	struct work_struct         set_rx_mode_work;
 	struct work_struct         tx_timeout_work;
-	struct delayed_work        update_stats_work;
+	struct work_struct         update_stats_work;
 
 	struct mlx5_core_dev      *mdev;
 	struct net_device         *netdev;
@@ -698,7 +706,7 @@ struct mlx5e_priv {
 };
 
 struct mlx5e_profile {
-	void	(*init)(struct mlx5_core_dev *mdev,
+	int	(*init)(struct mlx5_core_dev *mdev,
 			struct net_device *netdev,
 			const struct mlx5e_profile *profile, void *ppriv);
 	void	(*cleanup)(struct mlx5e_priv *priv);
@@ -710,7 +718,6 @@ struct mlx5e_profile {
 	void	(*disable)(struct mlx5e_priv *priv);
 	void	(*update_stats)(struct mlx5e_priv *priv);
 	void	(*update_carrier)(struct mlx5e_priv *priv);
-	int	(*max_nch)(struct mlx5_core_dev *mdev);
 	struct {
 		mlx5e_fp_handle_rx_cqe handle_rx_cqe;
 		mlx5e_fp_handle_rx_cqe handle_rx_cqe_mpwqe;
@@ -926,8 +933,8 @@ int mlx5e_create_tises(struct mlx5e_priv *priv);
 void mlx5e_cleanup_nic_tx(struct mlx5e_priv *priv);
 int mlx5e_close(struct net_device *netdev);
 int mlx5e_open(struct net_device *netdev);
-void mlx5e_update_stats_work(struct work_struct *work);
 
+void mlx5e_queue_update_stats(struct mlx5e_priv *priv);
 int mlx5e_bits_invert(unsigned long a, int size);
 
 typedef int (*change_hw_mtu_cb)(struct mlx5e_priv *priv);
@@ -962,9 +969,15 @@ int mlx5e_ethtool_flash_device(struct mlx5e_priv *priv,
 			       struct ethtool_flash *flash);
 
 /* mlx5e generic netdev management API */
+int mlx5e_netdev_init(struct net_device *netdev,
+		      struct mlx5e_priv *priv,
+		      struct mlx5_core_dev *mdev,
+		      const struct mlx5e_profile *profile,
+		      void *ppriv);
+void mlx5e_netdev_cleanup(struct net_device *netdev, struct mlx5e_priv *priv);
 struct net_device*
 mlx5e_create_netdev(struct mlx5_core_dev *mdev, const struct mlx5e_profile *profile,
-		    void *ppriv);
+		    int nch, void *ppriv);
 int mlx5e_attach_netdev(struct mlx5e_priv *priv);
 void mlx5e_detach_netdev(struct mlx5e_priv *priv);
 void mlx5e_destroy_netdev(struct mlx5e_priv *priv);

drivers/net/ethernet/mellanox/mlx5/core/en_ethtool.c

@@ -319,7 +319,7 @@ static int mlx5e_set_ringparam(struct net_device *dev,
 void mlx5e_ethtool_get_channels(struct mlx5e_priv *priv,
 				struct ethtool_channels *ch)
 {
-	ch->max_combined   = priv->profile->max_nch(priv->mdev);
+	ch->max_combined   = mlx5e_get_netdev_max_channels(priv->netdev);
 	ch->combined_count = priv->channels.params.num_channels;
 }
 

drivers/net/ethernet/mellanox/mlx5/core/en_rep.c

@@ -992,8 +992,7 @@ mlx5e_rep_get_stats(struct net_device *dev, struct rtnl_link_stats64 *stats)
 	struct mlx5e_priv *priv = netdev_priv(dev);
 
 	/* update HW stats in background for next time */
-	queue_delayed_work(priv->wq, &priv->update_stats_work, 0);
-
+	mlx5e_queue_update_stats(priv);
 	memcpy(stats, &priv->stats.vf_vport, sizeof(*stats));
 }
 
@@ -1078,28 +1077,33 @@ static void mlx5e_build_rep_netdev(struct net_device *netdev)
 	netdev->max_mtu = MLX5E_HW2SW_MTU(&priv->channels.params, max_mtu);
 }
 
-static void mlx5e_init_rep(struct mlx5_core_dev *mdev,
-			   struct net_device *netdev,
-			   const struct mlx5e_profile *profile,
-			   void *ppriv)
+static int mlx5e_init_rep(struct mlx5_core_dev *mdev,
+			  struct net_device *netdev,
+			  const struct mlx5e_profile *profile,
+			  void *ppriv)
 {
 	struct mlx5e_priv *priv = netdev_priv(netdev);
+	int err;
 
-	priv->mdev                         = mdev;
-	priv->netdev                       = netdev;
-	priv->profile                      = profile;
-	priv->ppriv                        = ppriv;
-
-	mutex_init(&priv->state_lock);
+	err = mlx5e_netdev_init(netdev, priv, mdev, profile, ppriv);
+	if (err)
+		return err;
 
-	INIT_DELAYED_WORK(&priv->update_stats_work, mlx5e_update_stats_work);
 
-	priv->channels.params.num_channels = 1;
+	priv->channels.params.num_channels =
+				mlx5e_get_netdev_max_channels(netdev);
 
 	mlx5e_build_rep_params(mdev, &priv->channels.params, netdev->mtu);
 	mlx5e_build_rep_netdev(netdev);
 
 	mlx5e_timestamp_init(priv);
+
+	return 0;
+}
+
+static void mlx5e_cleanup_rep(struct mlx5e_priv *priv)
+{
+	mlx5e_netdev_cleanup(priv->netdev, priv);
 }
 
 static int mlx5e_create_rep_ttc_table(struct mlx5e_priv *priv)
@@ -1224,12 +1228,12 @@ static int mlx5e_init_rep_tx(struct mlx5e_priv *priv)
 
 static const struct mlx5e_profile mlx5e_rep_profile = {
 	.init			= mlx5e_init_rep,
+	.cleanup		= mlx5e_cleanup_rep,
 	.init_rx		= mlx5e_init_rep_rx,
 	.cleanup_rx		= mlx5e_cleanup_rep_rx,
 	.init_tx		= mlx5e_init_rep_tx,
 	.cleanup_tx		= mlx5e_cleanup_nic_tx,
 	.update_stats           = mlx5e_rep_update_hw_counters,
-	.max_nch		= mlx5e_get_max_num_channels,
 	.update_carrier		= NULL,
 	.rx_handlers.handle_rx_cqe       = mlx5e_handle_rx_cqe_rep,
 	.rx_handlers.handle_rx_cqe_mpwqe = mlx5e_handle_rx_cqe_mpwrq,
@@ -1292,13 +1296,14 @@ mlx5e_vport_rep_load(struct mlx5_core_dev *dev, struct mlx5_eswitch_rep *rep)
 	struct mlx5e_rep_priv *rpriv;
 	struct net_device *netdev;
 	struct mlx5e_priv *upriv;
-	int err;
+	int nch, err;
 
 	rpriv = kzalloc(sizeof(*rpriv), GFP_KERNEL);
 	if (!rpriv)
 		return -ENOMEM;
 
-	netdev = mlx5e_create_netdev(dev, &mlx5e_rep_profile, rpriv);
+	nch = mlx5e_get_max_num_channels(dev);
+	netdev = mlx5e_create_netdev(dev, &mlx5e_rep_profile, nch, rpriv);
 	if (!netdev) {
 		pr_warn("Failed to create representor netdev for vport %d\n",
 			rep->vport);

drivers/net/ethernet/mellanox/mlx5/core/en_stats.c

@@ -133,7 +133,7 @@ void mlx5e_grp_sw_update_stats(struct mlx5e_priv *priv)
 
 	memset(s, 0, sizeof(*s));
 
-	for (i = 0; i < priv->profile->max_nch(priv->mdev); i++) {
+	for (i = 0; i < mlx5e_get_netdev_max_channels(priv->netdev); i++) {
 		struct mlx5e_channel_stats *channel_stats =
 			&priv->channel_stats[i];
 		struct mlx5e_xdpsq_stats *xdpsq_red_stats = &channel_stats->xdpsq;
@@ -1217,7 +1217,7 @@ static const struct counter_desc ch_stats_desc[] = {
 
 static int mlx5e_grp_channels_get_num_stats(struct mlx5e_priv *priv)
 {
-	int max_nch = priv->profile->max_nch(priv->mdev);
+	int max_nch = mlx5e_get_netdev_max_channels(priv->netdev);
 
 	return (NUM_RQ_STATS * max_nch) +
 	       (NUM_CH_STATS * max_nch) +
@@ -1229,7 +1229,7 @@ static int mlx5e_grp_channels_get_num_stats(struct mlx5e_priv *priv)
 static int mlx5e_grp_channels_fill_strings(struct mlx5e_priv *priv, u8 *data,
 					   int idx)
 {
-	int max_nch = priv->profile->max_nch(priv->mdev);
+	int max_nch = mlx5e_get_netdev_max_channels(priv->netdev);
 	int i, j, tc;
 
 	for (i = 0; i < max_nch; i++)
@@ -1264,7 +1264,7 @@ static int mlx5e_grp_channels_fill_strings(struct mlx5e_priv *priv, u8 *data,
 static int mlx5e_grp_channels_fill_stats(struct mlx5e_priv *priv, u64 *data,
 					 int idx)
 {
-	int max_nch = priv->profile->max_nch(priv->mdev);
+	int max_nch = mlx5e_get_netdev_max_channels(priv->netdev);
 	int i, j, tc;
 
 	for (i = 0; i < max_nch; i++)

drivers/net/ethernet/mellanox/mlx5/core/ipoib/ipoib.c

@@ -71,27 +71,25 @@ static void mlx5i_build_nic_params(struct mlx5_core_dev *mdev,
 }
 
 /* Called directly after IPoIB netdevice was created to initialize SW structs */
-void mlx5i_init(struct mlx5_core_dev *mdev,
-		struct net_device *netdev,
-		const struct mlx5e_profile *profile,
-		void *ppriv)
+int 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);
 	u16 max_mtu;
+	int err;
 
-	/* priv init */
-	priv->mdev        = mdev;
-	priv->netdev      = netdev;
-	priv->profile     = profile;
-	priv->ppriv       = ppriv;
-	priv->max_opened_tc = 1;
-	mutex_init(&priv->state_lock);
+	err = mlx5e_netdev_init(netdev, priv, mdev, profile, ppriv);
+	if (err)
+		return err;
 
 	mlx5_query_port_max_mtu(mdev, &max_mtu, 1);
 	netdev->mtu = max_mtu;
 
 	mlx5e_build_nic_params(mdev, &priv->channels.params,
-			       profile->max_nch(mdev), netdev->mtu);
+			       mlx5e_get_netdev_max_channels(netdev),
+			       netdev->mtu);
 	mlx5i_build_nic_params(mdev, &priv->channels.params);
 
 	mlx5e_timestamp_init(priv);
@@ -108,20 +106,23 @@ void mlx5i_init(struct mlx5_core_dev *mdev,
 
 	netdev->netdev_ops = &mlx5i_netdev_ops;
 	netdev->ethtool_ops = &mlx5i_ethtool_ops;
+
+	return 0;
 }
 
 /* Called directly before IPoIB netdevice is destroyed to cleanup SW structs */
-static void mlx5i_cleanup(struct mlx5e_priv *priv)
+void mlx5i_cleanup(struct mlx5e_priv *priv)
 {
-	/* Do nothing .. */
+	mlx5e_netdev_cleanup(priv->netdev, priv);
 }
 
 static void mlx5i_grp_sw_update_stats(struct mlx5e_priv *priv)
 {
+	int max_nch = mlx5e_get_netdev_max_channels(priv->netdev);
 	struct mlx5e_sw_stats s = { 0 };
 	int i, j;
 
-	for (i = 0; i < priv->profile->max_nch(priv->mdev); i++) {
+	for (i = 0; i < max_nch; i++) {
 		struct mlx5e_channel_stats *channel_stats;
 		struct mlx5e_rq_stats *rq_stats;
 
@@ -418,7 +419,6 @@ static const struct mlx5e_profile mlx5i_nic_profile = {
 	.enable		   = NULL, /* mlx5i_enable */
 	.disable	   = NULL, /* mlx5i_disable */
 	.update_stats	   = NULL, /* mlx5i_update_stats */
-	.max_nch	   = mlx5e_get_max_num_channels,
 	.update_carrier    = NULL, /* no HW update in IB link */
 	.rx_handlers.handle_rx_cqe       = mlx5i_handle_rx_cqe,
 	.rx_handlers.handle_rx_cqe_mpwqe = NULL, /* Not supported */
@@ -650,7 +650,6 @@ static void mlx5_rdma_netdev_free(struct net_device *netdev)
 
 	mlx5e_detach_netdev(priv);
 	profile->cleanup(priv);
-	destroy_workqueue(priv->wq);
 
 	if (!ipriv->sub_interface) {
 		mlx5i_pkey_qpn_ht_cleanup(netdev);
@@ -658,58 +657,37 @@ static void mlx5_rdma_netdev_free(struct net_device *netdev)
 	}
 }
 
-struct net_device *mlx5_rdma_netdev_alloc(struct mlx5_core_dev *mdev,
-					  struct ib_device *ibdev,
-					  const char *name,
-					  void (*setup)(struct net_device *))
+static bool mlx5_is_sub_interface(struct mlx5_core_dev *mdev)
+{
+	return mdev->mlx5e_res.pdn != 0;
+}
+
+static const struct mlx5e_profile *mlx5_get_profile(struct mlx5_core_dev *mdev)
 {
-	const struct mlx5e_profile *profile;
-	struct net_device *netdev;
+	if (mlx5_is_sub_interface(mdev))
+		return mlx5i_pkey_get_profile();
+	return &mlx5i_nic_profile;
+}
+
+static int mlx5_rdma_setup_rn(struct ib_device *ibdev, u8 port_num,
+			      struct net_device *netdev, void *param)
+{
+	struct mlx5_core_dev *mdev = (struct mlx5_core_dev *)param;
+	const struct mlx5e_profile *prof = mlx5_get_profile(mdev);
 	struct mlx5i_priv *ipriv;
 	struct mlx5e_priv *epriv;
 	struct rdma_netdev *rn;
-	bool sub_interface;
-	int nch;
 	int err;
 
-	if (mlx5i_check_required_hca_cap(mdev)) {
-		mlx5_core_warn(mdev, "Accelerated mode is not supported\n");
-		return ERR_PTR(-EOPNOTSUPP);
-	}
-
-	/* TODO: Need to find a better way to check if child device*/
-	sub_interface = (mdev->mlx5e_res.pdn != 0);
-
-	if (sub_interface)
-		profile = mlx5i_pkey_get_profile();
-	else
-		profile = &mlx5i_nic_profile;
-
-	nch = profile->max_nch(mdev);
-
-	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");
-		return NULL;
-	}
-
 	ipriv = netdev_priv(netdev);
 	epriv = mlx5i_epriv(netdev);
 
-	epriv->wq = create_singlethread_workqueue("mlx5i");
-	if (!epriv->wq)
-		goto err_free_netdev;
-
-	ipriv->sub_interface = sub_interface;
+	ipriv->sub_interface = mlx5_is_sub_interface(mdev);
 	if (!ipriv->sub_interface) {
 		err = mlx5i_pkey_qpn_ht_init(netdev);
 		if (err) {
 			mlx5_core_warn(mdev, "allocate qpn_to_netdev ht failed\n");
-			goto destroy_wq;
+			return err;
 		}
 
 		/* This should only be called once per mdev */
@@ -718,7 +696,7 @@ struct net_device *mlx5_rdma_netdev_alloc(struct mlx5_core_dev *mdev,
 			goto destroy_ht;
 	}
 
-	profile->init(mdev, netdev, profile, ipriv);
+	prof->init(mdev, netdev, prof, ipriv);
 
 	mlx5e_attach_netdev(epriv);
 	netif_carrier_off(netdev);
@@ -734,15 +712,35 @@ struct net_device *mlx5_rdma_netdev_alloc(struct mlx5_core_dev *mdev,
 	netdev->priv_destructor = mlx5_rdma_netdev_free;
 	netdev->needs_free_netdev = 1;
 
-	return netdev;
+	return 0;
 
 destroy_ht:
 	mlx5i_pkey_qpn_ht_cleanup(netdev);
-destroy_wq:
-	destroy_workqueue(epriv->wq);
-err_free_netdev:
-	free_netdev(netdev);
+	return err;
+}
 
-	return NULL;
+int mlx5_rdma_rn_get_params(struct mlx5_core_dev *mdev,
+			    struct ib_device *device,
+			    struct rdma_netdev_alloc_params *params)
+{
+	int nch;
+	int rc;
+
+	rc = mlx5i_check_required_hca_cap(mdev);
+	if (rc)
+		return rc;
+
+	nch = mlx5e_get_max_num_channels(mdev);
+
+	*params = (struct rdma_netdev_alloc_params){
+		.sizeof_priv = sizeof(struct mlx5i_priv) +
+			       sizeof(struct mlx5e_priv),
+		.txqs = nch * MLX5E_MAX_NUM_TC,
+		.rxqs = nch,
+		.param = mdev,
+		.initialize_rdma_netdev = mlx5_rdma_setup_rn,
+	};
+
+	return 0;
 }
-EXPORT_SYMBOL(mlx5_rdma_netdev_alloc);
+EXPORT_SYMBOL(mlx5_rdma_rn_get_params);

drivers/net/ethernet/mellanox/mlx5/core/ipoib/ipoib.h

@@ -84,10 +84,11 @@ void mlx5i_dev_cleanup(struct net_device *dev);
 int mlx5i_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd);
 
 /* Parent profile functions */
-void mlx5i_init(struct mlx5_core_dev *mdev,
-		struct net_device *netdev,
-		const struct mlx5e_profile *profile,
-		void *ppriv);
+int mlx5i_init(struct mlx5_core_dev *mdev,
+	       struct net_device *netdev,
+	       const struct mlx5e_profile *profile,
+	       void *ppriv);
+void mlx5i_cleanup(struct mlx5e_priv *priv);
 
 /* Get child interface nic profile */
 const struct mlx5e_profile *mlx5i_pkey_get_profile(void);

drivers/net/ethernet/mellanox/mlx5/core/ipoib/ipoib_vlan.c

@@ -275,14 +275,17 @@ static int mlx5i_pkey_change_mtu(struct net_device *netdev, int new_mtu)
 }
 
 /* Called directly after IPoIB netdevice was created to initialize SW structs */
-static void mlx5i_pkey_init(struct mlx5_core_dev *mdev,
-			     struct net_device *netdev,
-			     const struct mlx5e_profile *profile,
-			     void *ppriv)
+static int mlx5i_pkey_init(struct mlx5_core_dev *mdev,
+			   struct net_device *netdev,
+			   const struct mlx5e_profile *profile,
+			   void *ppriv)
 {
 	struct mlx5e_priv *priv  = mlx5i_epriv(netdev);
+	int err;
 
-	mlx5i_init(mdev, netdev, profile, ppriv);
+	err = mlx5i_init(mdev, netdev, profile, ppriv);
+	if (err)
+		return err;
 
 	/* Override parent ndo */
 	netdev->netdev_ops = &mlx5i_pkey_netdev_ops;
@@ -292,12 +295,14 @@ static void mlx5i_pkey_init(struct mlx5_core_dev *mdev,
 
 	/* Use dummy rqs */
 	priv->channels.params.log_rq_mtu_frames = MLX5E_PARAMS_MINIMUM_LOG_RQ_SIZE;
+
+	return 0;
 }
 
 /* Called directly before IPoIB netdevice is destroyed to cleanup SW structs */
 static void mlx5i_pkey_cleanup(struct mlx5e_priv *priv)
 {
-	/* Do nothing .. */
+	mlx5i_cleanup(priv);
 }
 
 static int mlx5i_pkey_init_tx(struct mlx5e_priv *priv)
@@ -346,7 +351,6 @@ static const struct mlx5e_profile mlx5i_pkey_nic_profile = {
 	.enable		   = NULL,
 	.disable	   = NULL,
 	.update_stats	   = NULL,
-	.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,

include/linux/mlx5/driver.h

@@ -1228,21 +1228,15 @@ int mlx5_lag_query_cong_counters(struct mlx5_core_dev *dev,
 struct mlx5_uars_page *mlx5_get_uars_page(struct mlx5_core_dev *mdev);
 void mlx5_put_uars_page(struct mlx5_core_dev *mdev, struct mlx5_uars_page *up);
 
-#ifndef CONFIG_MLX5_CORE_IPOIB
-static inline
-struct net_device *mlx5_rdma_netdev_alloc(struct mlx5_core_dev *mdev,
-					  struct ib_device *ibdev,
-					  const char *name,
-					  void (*setup)(struct net_device *))
-{
-	return ERR_PTR(-EOPNOTSUPP);
-}
-#else
+#ifdef CONFIG_MLX5_CORE_IPOIB
 struct net_device *mlx5_rdma_netdev_alloc(struct mlx5_core_dev *mdev,
 					  struct ib_device *ibdev,
 					  const char *name,
 					  void (*setup)(struct net_device *));
 #endif /* CONFIG_MLX5_CORE_IPOIB */
+int mlx5_rdma_rn_get_params(struct mlx5_core_dev *mdev,
+			    struct ib_device *device,
+			    struct rdma_netdev_alloc_params *params);
 
 struct mlx5_profile {
 	u64	mask;

include/rdma/ib_verbs.h

@@ -2223,6 +2223,16 @@ struct rdma_netdev {
 			    union ib_gid *gid, u16 mlid);
 };
 
+struct rdma_netdev_alloc_params {
+	size_t sizeof_priv;
+	unsigned int txqs;
+	unsigned int rxqs;
+	void *param;
+
+	int (*initialize_rdma_netdev)(struct ib_device *device, u8 port_num,
+				      struct net_device *netdev, void *param);
+};
+
 struct ib_port_pkey_list {
 	/* Lock to hold while modifying the list. */
 	spinlock_t                    list_lock;
@@ -2523,8 +2533,8 @@ struct ib_device {
 	/**
 	 * rdma netdev operation
 	 *
-	 * Driver implementing alloc_rdma_netdev must return -EOPNOTSUPP if it
-	 * doesn't support the specified rdma netdev type.
+	 * Driver implementing alloc_rdma_netdev or rdma_netdev_get_params
+	 * must return -EOPNOTSUPP if it doesn't support the specified type.
 	 */
 	struct net_device *(*alloc_rdma_netdev)(
 					struct ib_device *device,
@@ -2534,6 +2544,10 @@ struct ib_device {
 					unsigned char name_assign_type,
 					void (*setup)(struct net_device *));
 
+	int (*rdma_netdev_get_params)(struct ib_device *device, u8 port_num,
+				      enum rdma_netdev_t type,
+				      struct rdma_netdev_alloc_params *params);
+
 	struct module               *owner;
 	struct device                dev;
 	struct kobject               *ports_parent;
@@ -4179,4 +4193,16 @@ struct ib_ucontext *ib_uverbs_get_ucontext(struct ib_uverbs_file *ufile);
 
 int uverbs_destroy_def_handler(struct ib_uverbs_file *file,
 			       struct uverbs_attr_bundle *attrs);
+
+struct net_device *rdma_alloc_netdev(struct ib_device *device, u8 port_num,
+				     enum rdma_netdev_t type, const char *name,
+				     unsigned char name_assign_type,
+				     void (*setup)(struct net_device *));
+
+int rdma_init_netdev(struct ib_device *device, u8 port_num,
+		     enum rdma_netdev_t type, const char *name,
+		     unsigned char name_assign_type,
+		     void (*setup)(struct net_device *),
+		     struct net_device *netdev);
+
 #endif /* IB_VERBS_H */