mlxsw: spectrum: Add initial support for Spectrum ASIC
authorJiri Pirko <jiri@mellanox.com>
Fri, 16 Oct 2015 12:01:37 +0000 (14:01 +0200)
committerDavid S. Miller <davem@davemloft.net>
Fri, 16 Oct 2015 14:15:23 +0000 (07:15 -0700)
Add support for new generation Mellanox Spectrum ASIC, 10/25/40/50 and
100Gb/s Ethernet Switch.

The initial driver implements bridge forwarding offload including
bridge internal VLAN support, FDB static entries, FDB learning and
HW ageing including their setup.

Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: Elad Raz <eladr@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
drivers/net/ethernet/mellanox/mlxsw/Kconfig
drivers/net/ethernet/mellanox/mlxsw/Makefile
drivers/net/ethernet/mellanox/mlxsw/core.h
drivers/net/ethernet/mellanox/mlxsw/pci.c
drivers/net/ethernet/mellanox/mlxsw/pci.h
drivers/net/ethernet/mellanox/mlxsw/spectrum.c [new file with mode: 0644]
drivers/net/ethernet/mellanox/mlxsw/spectrum.h [new file with mode: 0644]
drivers/net/ethernet/mellanox/mlxsw/spectrum_buffers.c [new file with mode: 0644]
drivers/net/ethernet/mellanox/mlxsw/spectrum_switchdev.c [new file with mode: 0644]
drivers/net/ethernet/mellanox/mlxsw/txheader.h

index 2941d9c..e36e122 100644 (file)
@@ -30,3 +30,14 @@ config MLXSW_SWITCHX2
 
          To compile this driver as a module, choose M here: the
          module will be called mlxsw_switchx2.
+
+config MLXSW_SPECTRUM
+       tristate "Mellanox Technologies Spectrum support"
+       depends on MLXSW_CORE && NET_SWITCHDEV
+       default m
+       ---help---
+         This driver supports Mellanox Technologies Spectrum Ethernet
+         Switch ASICs.
+
+         To compile this driver as a module, choose M here: the
+         module will be called mlxsw_spectrum.
index 0a05f65..af01581 100644 (file)
@@ -4,3 +4,6 @@ obj-$(CONFIG_MLXSW_PCI)         += mlxsw_pci.o
 mlxsw_pci-objs                 := pci.o
 obj-$(CONFIG_MLXSW_SWITCHX2)   += mlxsw_switchx2.o
 mlxsw_switchx2-objs            := switchx2.o
+obj-$(CONFIG_MLXSW_SPECTRUM)   += mlxsw_spectrum.o
+mlxsw_spectrum-objs            := spectrum.o spectrum_buffers.o \
+                                  spectrum_switchdev.o
index e92ab27..8078273 100644 (file)
@@ -54,6 +54,7 @@
        MODULE_ALIAS(MLXSW_MODULE_ALIAS_PREFIX kind)
 
 #define MLXSW_DEVICE_KIND_SWITCHX2 "switchx2"
+#define MLXSW_DEVICE_KIND_SPECTRUM "spectrum"
 
 struct mlxsw_core;
 struct mlxsw_driver;
index 0fa44c8..879e000 100644 (file)
@@ -57,6 +57,7 @@ static const char mlxsw_pci_driver_name[] = "mlxsw_pci";
 
 static const struct pci_device_id mlxsw_pci_id_table[] = {
        {PCI_VDEVICE(MELLANOX, PCI_DEVICE_ID_MELLANOX_SWITCHX2), 0},
+       {PCI_VDEVICE(MELLANOX, PCI_DEVICE_ID_MELLANOX_SPECTRUM), 0},
        {0, }
 };
 
@@ -67,6 +68,8 @@ static const char *mlxsw_pci_device_kind_get(const struct pci_device_id *id)
        switch (id->device) {
        case PCI_DEVICE_ID_MELLANOX_SWITCHX2:
                return MLXSW_DEVICE_KIND_SWITCHX2;
+       case PCI_DEVICE_ID_MELLANOX_SPECTRUM:
+               return MLXSW_DEVICE_KIND_SPECTRUM;
        default:
                BUG();
        }
index 5b3453b..142f33d 100644 (file)
@@ -40,6 +40,7 @@
 #include "item.h"
 
 #define PCI_DEVICE_ID_MELLANOX_SWITCHX2        0xc738
+#define PCI_DEVICE_ID_MELLANOX_SPECTRUM        0xcb84
 #define MLXSW_PCI_BAR0_SIZE            (1024 * 1024) /* 1MB */
 #define MLXSW_PCI_PAGE_SIZE            4096
 
diff --git a/drivers/net/ethernet/mellanox/mlxsw/spectrum.c b/drivers/net/ethernet/mellanox/mlxsw/spectrum.c
new file mode 100644 (file)
index 0000000..6e9906d
--- /dev/null
@@ -0,0 +1,1948 @@
+/*
+ * drivers/net/ethernet/mellanox/mlxsw/spectrum.c
+ * Copyright (c) 2015 Mellanox Technologies. All rights reserved.
+ * Copyright (c) 2015 Jiri Pirko <jiri@mellanox.com>
+ * Copyright (c) 2015 Ido Schimmel <idosch@mellanox.com>
+ * Copyright (c) 2015 Elad Raz <eladr@mellanox.com>
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions are met:
+ *
+ * 1. Redistributions of source code must retain the above copyright
+ *    notice, this list of conditions and the following disclaimer.
+ * 2. 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.
+ * 3. Neither the names of the copyright holders nor the names of its
+ *    contributors may be used to endorse or promote products derived from
+ *    this software without specific prior written permission.
+ *
+ * Alternatively, this software may be distributed under the terms of the
+ * GNU General Public License ("GPL") version 2 as published by the Free
+ * Software Foundation.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+ * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
+ * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
+ * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
+ * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
+ * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
+ * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
+ * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
+ * POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#include <linux/kernel.h>
+#include <linux/module.h>
+#include <linux/types.h>
+#include <linux/netdevice.h>
+#include <linux/etherdevice.h>
+#include <linux/ethtool.h>
+#include <linux/slab.h>
+#include <linux/device.h>
+#include <linux/skbuff.h>
+#include <linux/if_vlan.h>
+#include <linux/if_bridge.h>
+#include <linux/workqueue.h>
+#include <linux/jiffies.h>
+#include <linux/bitops.h>
+#include <net/switchdev.h>
+#include <generated/utsrelease.h>
+
+#include "spectrum.h"
+#include "core.h"
+#include "reg.h"
+#include "port.h"
+#include "trap.h"
+#include "txheader.h"
+
+static const char mlxsw_sp_driver_name[] = "mlxsw_spectrum";
+static const char mlxsw_sp_driver_version[] = "1.0";
+
+/* tx_hdr_version
+ * Tx header version.
+ * Must be set to 1.
+ */
+MLXSW_ITEM32(tx, hdr, version, 0x00, 28, 4);
+
+/* tx_hdr_ctl
+ * Packet control type.
+ * 0 - Ethernet control (e.g. EMADs, LACP)
+ * 1 - Ethernet data
+ */
+MLXSW_ITEM32(tx, hdr, ctl, 0x00, 26, 2);
+
+/* tx_hdr_proto
+ * Packet protocol type. Must be set to 1 (Ethernet).
+ */
+MLXSW_ITEM32(tx, hdr, proto, 0x00, 21, 3);
+
+/* tx_hdr_rx_is_router
+ * Packet is sent from the router. Valid for data packets only.
+ */
+MLXSW_ITEM32(tx, hdr, rx_is_router, 0x00, 19, 1);
+
+/* tx_hdr_fid_valid
+ * Indicates if the 'fid' field is valid and should be used for
+ * forwarding lookup. Valid for data packets only.
+ */
+MLXSW_ITEM32(tx, hdr, fid_valid, 0x00, 16, 1);
+
+/* tx_hdr_swid
+ * Switch partition ID. Must be set to 0.
+ */
+MLXSW_ITEM32(tx, hdr, swid, 0x00, 12, 3);
+
+/* tx_hdr_control_tclass
+ * Indicates if the packet should use the control TClass and not one
+ * of the data TClasses.
+ */
+MLXSW_ITEM32(tx, hdr, control_tclass, 0x00, 6, 1);
+
+/* tx_hdr_etclass
+ * Egress TClass to be used on the egress device on the egress port.
+ */
+MLXSW_ITEM32(tx, hdr, etclass, 0x00, 0, 4);
+
+/* tx_hdr_port_mid
+ * Destination local port for unicast packets.
+ * Destination multicast ID for multicast packets.
+ *
+ * Control packets are directed to a specific egress port, while data
+ * packets are transmitted through the CPU port (0) into the switch partition,
+ * where forwarding rules are applied.
+ */
+MLXSW_ITEM32(tx, hdr, port_mid, 0x04, 16, 16);
+
+/* tx_hdr_fid
+ * Forwarding ID used for L2 forwarding lookup. Valid only if 'fid_valid' is
+ * set, otherwise calculated based on the packet's VID using VID to FID mapping.
+ * Valid for data packets only.
+ */
+MLXSW_ITEM32(tx, hdr, fid, 0x08, 0, 16);
+
+/* tx_hdr_type
+ * 0 - Data packets
+ * 6 - Control packets
+ */
+MLXSW_ITEM32(tx, hdr, type, 0x0C, 0, 4);
+
+static void mlxsw_sp_txhdr_construct(struct sk_buff *skb,
+                                    const struct mlxsw_tx_info *tx_info)
+{
+       char *txhdr = skb_push(skb, MLXSW_TXHDR_LEN);
+
+       memset(txhdr, 0, MLXSW_TXHDR_LEN);
+
+       mlxsw_tx_hdr_version_set(txhdr, MLXSW_TXHDR_VERSION_1);
+       mlxsw_tx_hdr_ctl_set(txhdr, MLXSW_TXHDR_ETH_CTL);
+       mlxsw_tx_hdr_proto_set(txhdr, MLXSW_TXHDR_PROTO_ETH);
+       mlxsw_tx_hdr_swid_set(txhdr, 0);
+       mlxsw_tx_hdr_control_tclass_set(txhdr, 1);
+       mlxsw_tx_hdr_port_mid_set(txhdr, tx_info->local_port);
+       mlxsw_tx_hdr_type_set(txhdr, MLXSW_TXHDR_TYPE_CONTROL);
+}
+
+static int mlxsw_sp_base_mac_get(struct mlxsw_sp *mlxsw_sp)
+{
+       char spad_pl[MLXSW_REG_SPAD_LEN];
+       int err;
+
+       err = mlxsw_reg_query(mlxsw_sp->core, MLXSW_REG(spad), spad_pl);
+       if (err)
+               return err;
+       mlxsw_reg_spad_base_mac_memcpy_from(spad_pl, mlxsw_sp->base_mac);
+       return 0;
+}
+
+static int mlxsw_sp_port_admin_status_set(struct mlxsw_sp_port *mlxsw_sp_port,
+                                         bool is_up)
+{
+       struct mlxsw_sp *mlxsw_sp = mlxsw_sp_port->mlxsw_sp;
+       char paos_pl[MLXSW_REG_PAOS_LEN];
+
+       mlxsw_reg_paos_pack(paos_pl, mlxsw_sp_port->local_port,
+                           is_up ? MLXSW_PORT_ADMIN_STATUS_UP :
+                           MLXSW_PORT_ADMIN_STATUS_DOWN);
+       return mlxsw_reg_write(mlxsw_sp->core, MLXSW_REG(paos), paos_pl);
+}
+
+static int mlxsw_sp_port_oper_status_get(struct mlxsw_sp_port *mlxsw_sp_port,
+                                        bool *p_is_up)
+{
+       struct mlxsw_sp *mlxsw_sp = mlxsw_sp_port->mlxsw_sp;
+       char paos_pl[MLXSW_REG_PAOS_LEN];
+       u8 oper_status;
+       int err;
+
+       mlxsw_reg_paos_pack(paos_pl, mlxsw_sp_port->local_port, 0);
+       err = mlxsw_reg_query(mlxsw_sp->core, MLXSW_REG(paos), paos_pl);
+       if (err)
+               return err;
+       oper_status = mlxsw_reg_paos_oper_status_get(paos_pl);
+       *p_is_up = oper_status == MLXSW_PORT_ADMIN_STATUS_UP ? true : false;
+       return 0;
+}
+
+static int mlxsw_sp_vfid_create(struct mlxsw_sp *mlxsw_sp, u16 vfid)
+{
+       char sfmr_pl[MLXSW_REG_SFMR_LEN];
+       int err;
+
+       mlxsw_reg_sfmr_pack(sfmr_pl, MLXSW_REG_SFMR_OP_CREATE_FID,
+                           MLXSW_SP_VFID_BASE + vfid, 0);
+       err = mlxsw_reg_write(mlxsw_sp->core, MLXSW_REG(sfmr), sfmr_pl);
+
+       if (err)
+               return err;
+
+       set_bit(vfid, mlxsw_sp->active_vfids);
+       return 0;
+}
+
+static void mlxsw_sp_vfid_destroy(struct mlxsw_sp *mlxsw_sp, u16 vfid)
+{
+       char sfmr_pl[MLXSW_REG_SFMR_LEN];
+
+       clear_bit(vfid, mlxsw_sp->active_vfids);
+
+       mlxsw_reg_sfmr_pack(sfmr_pl, MLXSW_REG_SFMR_OP_DESTROY_FID,
+                           MLXSW_SP_VFID_BASE + vfid, 0);
+       mlxsw_reg_write(mlxsw_sp->core, MLXSW_REG(sfmr), sfmr_pl);
+}
+
+static int mlxsw_sp_port_dev_addr_set(struct mlxsw_sp_port *mlxsw_sp_port,
+                                     unsigned char *addr)
+{
+       struct mlxsw_sp *mlxsw_sp = mlxsw_sp_port->mlxsw_sp;
+       char ppad_pl[MLXSW_REG_PPAD_LEN];
+
+       mlxsw_reg_ppad_pack(ppad_pl, true, mlxsw_sp_port->local_port);
+       mlxsw_reg_ppad_mac_memcpy_to(ppad_pl, addr);
+       return mlxsw_reg_write(mlxsw_sp->core, MLXSW_REG(ppad), ppad_pl);
+}
+
+static int mlxsw_sp_port_dev_addr_init(struct mlxsw_sp_port *mlxsw_sp_port)
+{
+       struct mlxsw_sp *mlxsw_sp = mlxsw_sp_port->mlxsw_sp;
+       unsigned char *addr = mlxsw_sp_port->dev->dev_addr;
+
+       ether_addr_copy(addr, mlxsw_sp->base_mac);
+       addr[ETH_ALEN - 1] += mlxsw_sp_port->local_port;
+       return mlxsw_sp_port_dev_addr_set(mlxsw_sp_port, addr);
+}
+
+static int mlxsw_sp_port_stp_state_set(struct mlxsw_sp_port *mlxsw_sp_port,
+                                      u16 vid, enum mlxsw_reg_spms_state state)
+{
+       struct mlxsw_sp *mlxsw_sp = mlxsw_sp_port->mlxsw_sp;
+       char *spms_pl;
+       int err;
+
+       spms_pl = kmalloc(MLXSW_REG_SPMS_LEN, GFP_KERNEL);
+       if (!spms_pl)
+               return -ENOMEM;
+       mlxsw_reg_spms_pack(spms_pl, mlxsw_sp_port->local_port);
+       mlxsw_reg_spms_vid_pack(spms_pl, vid, state);
+       err = mlxsw_reg_write(mlxsw_sp->core, MLXSW_REG(spms), spms_pl);
+       kfree(spms_pl);
+       return err;
+}
+
+static int mlxsw_sp_port_mtu_set(struct mlxsw_sp_port *mlxsw_sp_port, u16 mtu)
+{
+       struct mlxsw_sp *mlxsw_sp = mlxsw_sp_port->mlxsw_sp;
+       char pmtu_pl[MLXSW_REG_PMTU_LEN];
+       int max_mtu;
+       int err;
+
+       mtu += MLXSW_TXHDR_LEN + ETH_HLEN;
+       mlxsw_reg_pmtu_pack(pmtu_pl, mlxsw_sp_port->local_port, 0);
+       err = mlxsw_reg_query(mlxsw_sp->core, MLXSW_REG(pmtu), pmtu_pl);
+       if (err)
+               return err;
+       max_mtu = mlxsw_reg_pmtu_max_mtu_get(pmtu_pl);
+
+       if (mtu > max_mtu)
+               return -EINVAL;
+
+       mlxsw_reg_pmtu_pack(pmtu_pl, mlxsw_sp_port->local_port, mtu);
+       return mlxsw_reg_write(mlxsw_sp->core, MLXSW_REG(pmtu), pmtu_pl);
+}
+
+static int mlxsw_sp_port_swid_set(struct mlxsw_sp_port *mlxsw_sp_port, u8 swid)
+{
+       struct mlxsw_sp *mlxsw_sp = mlxsw_sp_port->mlxsw_sp;
+       char pspa_pl[MLXSW_REG_PSPA_LEN];
+
+       mlxsw_reg_pspa_pack(pspa_pl, swid, mlxsw_sp_port->local_port);
+       return mlxsw_reg_write(mlxsw_sp->core, MLXSW_REG(pspa), pspa_pl);
+}
+
+static int mlxsw_sp_port_vp_mode_set(struct mlxsw_sp_port *mlxsw_sp_port,
+                                    bool enable)
+{
+       struct mlxsw_sp *mlxsw_sp = mlxsw_sp_port->mlxsw_sp;
+       char svpe_pl[MLXSW_REG_SVPE_LEN];
+
+       mlxsw_reg_svpe_pack(svpe_pl, mlxsw_sp_port->local_port, enable);
+       return mlxsw_reg_write(mlxsw_sp->core, MLXSW_REG(svpe), svpe_pl);
+}
+
+int mlxsw_sp_port_vid_to_fid_set(struct mlxsw_sp_port *mlxsw_sp_port,
+                                enum mlxsw_reg_svfa_mt mt, bool valid, u16 fid,
+                                u16 vid)
+{
+       struct mlxsw_sp *mlxsw_sp = mlxsw_sp_port->mlxsw_sp;
+       char svfa_pl[MLXSW_REG_SVFA_LEN];
+
+       mlxsw_reg_svfa_pack(svfa_pl, mlxsw_sp_port->local_port, mt, valid,
+                           fid, vid);
+       return mlxsw_reg_write(mlxsw_sp->core, MLXSW_REG(svfa), svfa_pl);
+}
+
+static int mlxsw_sp_port_vid_learning_set(struct mlxsw_sp_port *mlxsw_sp_port,
+                                         u16 vid, bool learn_enable)
+{
+       struct mlxsw_sp *mlxsw_sp = mlxsw_sp_port->mlxsw_sp;
+       char *spvmlr_pl;
+       int err;
+
+       spvmlr_pl = kmalloc(MLXSW_REG_SPVMLR_LEN, GFP_KERNEL);
+       if (!spvmlr_pl)
+               return -ENOMEM;
+       mlxsw_reg_spvmlr_pack(spvmlr_pl, mlxsw_sp_port->local_port, vid, vid,
+                             learn_enable);
+       err = mlxsw_reg_write(mlxsw_sp->core, MLXSW_REG(spvmlr), spvmlr_pl);
+       kfree(spvmlr_pl);
+       return err;
+}
+
+static int
+mlxsw_sp_port_system_port_mapping_set(struct mlxsw_sp_port *mlxsw_sp_port)
+{
+       struct mlxsw_sp *mlxsw_sp = mlxsw_sp_port->mlxsw_sp;
+       char sspr_pl[MLXSW_REG_SSPR_LEN];
+
+       mlxsw_reg_sspr_pack(sspr_pl, mlxsw_sp_port->local_port);
+       return mlxsw_reg_write(mlxsw_sp->core, MLXSW_REG(sspr), sspr_pl);
+}
+
+static int mlxsw_sp_port_module_check(struct mlxsw_sp_port *mlxsw_sp_port,
+                                     bool *p_usable)
+{
+       struct mlxsw_sp *mlxsw_sp = mlxsw_sp_port->mlxsw_sp;
+       char pmlp_pl[MLXSW_REG_PMLP_LEN];
+       int err;
+
+       mlxsw_reg_pmlp_pack(pmlp_pl, mlxsw_sp_port->local_port);
+       err = mlxsw_reg_query(mlxsw_sp->core, MLXSW_REG(pmlp), pmlp_pl);
+       if (err)
+               return err;
+       *p_usable = mlxsw_reg_pmlp_width_get(pmlp_pl) ? true : false;
+       return 0;
+}
+
+static int mlxsw_sp_port_open(struct net_device *dev)
+{
+       struct mlxsw_sp_port *mlxsw_sp_port = netdev_priv(dev);
+       int err;
+
+       err = mlxsw_sp_port_admin_status_set(mlxsw_sp_port, true);
+       if (err)
+               return err;
+       netif_start_queue(dev);
+       return 0;
+}
+
+static int mlxsw_sp_port_stop(struct net_device *dev)
+{
+       struct mlxsw_sp_port *mlxsw_sp_port = netdev_priv(dev);
+
+       netif_stop_queue(dev);
+       return mlxsw_sp_port_admin_status_set(mlxsw_sp_port, false);
+}
+
+static netdev_tx_t mlxsw_sp_port_xmit(struct sk_buff *skb,
+                                     struct net_device *dev)
+{
+       struct mlxsw_sp_port *mlxsw_sp_port = netdev_priv(dev);
+       struct mlxsw_sp *mlxsw_sp = mlxsw_sp_port->mlxsw_sp;
+       struct mlxsw_sp_port_pcpu_stats *pcpu_stats;
+       const struct mlxsw_tx_info tx_info = {
+               .local_port = mlxsw_sp_port->local_port,
+               .is_emad = false,
+       };
+       u64 len;
+       int err;
+
+       if (mlxsw_core_skb_transmit_busy(mlxsw_sp, &tx_info))
+               return NETDEV_TX_BUSY;
+
+       if (unlikely(skb_headroom(skb) < MLXSW_TXHDR_LEN)) {
+               struct sk_buff *skb_orig = skb;
+
+               skb = skb_realloc_headroom(skb, MLXSW_TXHDR_LEN);
+               if (!skb) {
+                       this_cpu_inc(mlxsw_sp_port->pcpu_stats->tx_dropped);
+                       dev_kfree_skb_any(skb_orig);
+                       return NETDEV_TX_OK;
+               }
+       }
+
+       if (eth_skb_pad(skb)) {
+               this_cpu_inc(mlxsw_sp_port->pcpu_stats->tx_dropped);
+               return NETDEV_TX_OK;
+       }
+
+       mlxsw_sp_txhdr_construct(skb, &tx_info);
+       len = skb->len;
+       /* Due to a race we might fail here because of a full queue. In that
+        * unlikely case we simply drop the packet.
+        */
+       err = mlxsw_core_skb_transmit(mlxsw_sp, skb, &tx_info);
+
+       if (!err) {
+               pcpu_stats = this_cpu_ptr(mlxsw_sp_port->pcpu_stats);
+               u64_stats_update_begin(&pcpu_stats->syncp);
+               pcpu_stats->tx_packets++;
+               pcpu_stats->tx_bytes += len;
+               u64_stats_update_end(&pcpu_stats->syncp);
+       } else {
+               this_cpu_inc(mlxsw_sp_port->pcpu_stats->tx_dropped);
+               dev_kfree_skb_any(skb);
+       }
+       return NETDEV_TX_OK;
+}
+
+static int mlxsw_sp_port_set_mac_address(struct net_device *dev, void *p)
+{
+       struct mlxsw_sp_port *mlxsw_sp_port = netdev_priv(dev);
+       struct sockaddr *addr = p;
+       int err;
+
+       if (!is_valid_ether_addr(addr->sa_data))
+               return -EADDRNOTAVAIL;
+
+       err = mlxsw_sp_port_dev_addr_set(mlxsw_sp_port, addr->sa_data);
+       if (err)
+               return err;
+       memcpy(dev->dev_addr, addr->sa_data, dev->addr_len);
+       return 0;
+}
+
+static int mlxsw_sp_port_change_mtu(struct net_device *dev, int mtu)
+{
+       struct mlxsw_sp_port *mlxsw_sp_port = netdev_priv(dev);
+       int err;
+
+       err = mlxsw_sp_port_mtu_set(mlxsw_sp_port, mtu);
+       if (err)
+               return err;
+       dev->mtu = mtu;
+       return 0;
+}
+
+static struct rtnl_link_stats64 *
+mlxsw_sp_port_get_stats64(struct net_device *dev,
+                         struct rtnl_link_stats64 *stats)
+{
+       struct mlxsw_sp_port *mlxsw_sp_port = netdev_priv(dev);
+       struct mlxsw_sp_port_pcpu_stats *p;
+       u64 rx_packets, rx_bytes, tx_packets, tx_bytes;
+       u32 tx_dropped = 0;
+       unsigned int start;
+       int i;
+
+       for_each_possible_cpu(i) {
+               p = per_cpu_ptr(mlxsw_sp_port->pcpu_stats, i);
+               do {
+                       start = u64_stats_fetch_begin_irq(&p->syncp);
+                       rx_packets      = p->rx_packets;
+                       rx_bytes        = p->rx_bytes;
+                       tx_packets      = p->tx_packets;
+                       tx_bytes        = p->tx_bytes;
+               } while (u64_stats_fetch_retry_irq(&p->syncp, start));
+
+               stats->rx_packets       += rx_packets;
+               stats->rx_bytes         += rx_bytes;
+               stats->tx_packets       += tx_packets;
+               stats->tx_bytes         += tx_bytes;
+               /* tx_dropped is u32, updated without syncp protection. */
+               tx_dropped      += p->tx_dropped;
+       }
+       stats->tx_dropped       = tx_dropped;
+       return stats;
+}
+
+int mlxsw_sp_port_vlan_set(struct mlxsw_sp_port *mlxsw_sp_port, u16 vid_begin,
+                          u16 vid_end, bool is_member, bool untagged)
+{
+       struct mlxsw_sp *mlxsw_sp = mlxsw_sp_port->mlxsw_sp;
+       char *spvm_pl;
+       int err;
+
+       spvm_pl = kmalloc(MLXSW_REG_SPVM_LEN, GFP_KERNEL);
+       if (!spvm_pl)
+               return -ENOMEM;
+
+       mlxsw_reg_spvm_pack(spvm_pl, mlxsw_sp_port->local_port, vid_begin,
+                           vid_end, is_member, untagged);
+       err = mlxsw_reg_write(mlxsw_sp->core, MLXSW_REG(spvm), spvm_pl);
+       kfree(spvm_pl);
+       return err;
+}
+
+static int mlxsw_sp_port_vp_mode_trans(struct mlxsw_sp_port *mlxsw_sp_port)
+{
+       enum mlxsw_reg_svfa_mt mt = MLXSW_REG_SVFA_MT_PORT_VID_TO_FID;
+       u16 vid, last_visited_vid;
+       int err;
+
+       for_each_set_bit(vid, mlxsw_sp_port->active_vlans, VLAN_N_VID) {
+               err = mlxsw_sp_port_vid_to_fid_set(mlxsw_sp_port, mt, true, vid,
+                                                  vid);
+               if (err) {
+                       last_visited_vid = vid;
+                       goto err_port_vid_to_fid_set;
+               }
+       }
+
+       err = mlxsw_sp_port_vp_mode_set(mlxsw_sp_port, true);
+       if (err) {
+               last_visited_vid = VLAN_N_VID;
+               goto err_port_vid_to_fid_set;
+       }
+
+       return 0;
+
+err_port_vid_to_fid_set:
+       for_each_set_bit(vid, mlxsw_sp_port->active_vlans, last_visited_vid)
+               mlxsw_sp_port_vid_to_fid_set(mlxsw_sp_port, mt, false, vid,
+                                            vid);
+       return err;
+}
+
+static int mlxsw_sp_port_vlan_mode_trans(struct mlxsw_sp_port *mlxsw_sp_port)
+{
+       enum mlxsw_reg_svfa_mt mt = MLXSW_REG_SVFA_MT_PORT_VID_TO_FID;
+       u16 vid;
+       int err;
+
+       err = mlxsw_sp_port_vp_mode_set(mlxsw_sp_port, false);
+       if (err)
+               return err;
+
+       for_each_set_bit(vid, mlxsw_sp_port->active_vlans, VLAN_N_VID) {
+               err = mlxsw_sp_port_vid_to_fid_set(mlxsw_sp_port, mt, false,
+                                                  vid, vid);
+               if (err)
+                       return err;
+       }
+
+       return 0;
+}
+
+int mlxsw_sp_port_add_vid(struct net_device *dev, __be16 __always_unused proto,
+                         u16 vid)
+{
+       struct mlxsw_sp_port *mlxsw_sp_port = netdev_priv(dev);
+       struct mlxsw_sp *mlxsw_sp = mlxsw_sp_port->mlxsw_sp;
+       char *sftr_pl;
+       int err;
+
+       /* VLAN 0 is added to HW filter when device goes up, but it is
+        * reserved in our case, so simply return.
+        */
+       if (!vid)
+               return 0;
+
+       if (test_bit(vid, mlxsw_sp_port->active_vfids)) {
+               netdev_warn(dev, "VID=%d already configured\n", vid);
+               return 0;
+       }
+
+       if (!test_bit(vid, mlxsw_sp->active_vfids)) {
+               err = mlxsw_sp_vfid_create(mlxsw_sp, vid);
+               if (err) {
+                       netdev_err(dev, "Failed to create vFID=%d\n",
+                                  MLXSW_SP_VFID_BASE + vid);
+                       return err;
+               }
+
+               sftr_pl = kmalloc(MLXSW_REG_SFTR_LEN, GFP_KERNEL);
+               if (!sftr_pl) {
+                       err = -ENOMEM;
+                       goto err_flood_table_alloc;
+               }
+               mlxsw_reg_sftr_pack(sftr_pl, 0, vid,
+                                   MLXSW_REG_SFGC_TABLE_TYPE_FID, 0,
+                                   MLXSW_PORT_CPU_PORT, true);
+               err = mlxsw_reg_write(mlxsw_sp->core, MLXSW_REG(sftr), sftr_pl);
+               kfree(sftr_pl);
+               if (err) {
+                       netdev_err(dev, "Failed to configure flood table\n");
+                       goto err_flood_table_config;
+               }
+       }
+
+       /* In case we fail in the following steps, we intentionally do not
+        * destroy the associated vFID.
+        */
+
+       /* When adding the first VLAN interface on a bridged port we need to
+        * transition all the active 802.1Q bridge VLANs to use explicit
+        * {Port, VID} to FID mappings and set the port's mode to Virtual mode.
+        */
+       if (!mlxsw_sp_port->nr_vfids) {
+               err = mlxsw_sp_port_vp_mode_trans(mlxsw_sp_port);
+               if (err) {
+                       netdev_err(dev, "Failed to set to Virtual mode\n");
+                       return err;
+               }
+       }
+
+       err = mlxsw_sp_port_vid_to_fid_set(mlxsw_sp_port,
+                                          MLXSW_REG_SVFA_MT_PORT_VID_TO_FID,
+                                          true, MLXSW_SP_VFID_BASE + vid, vid);
+       if (err) {
+               netdev_err(dev, "Failed to map {Port, VID=%d} to vFID=%d\n",
+                          vid, MLXSW_SP_VFID_BASE + vid);
+               goto err_port_vid_to_fid_set;
+       }
+
+       err = mlxsw_sp_port_vid_learning_set(mlxsw_sp_port, vid, false);
+       if (err) {
+               netdev_err(dev, "Failed to disable learning for VID=%d\n", vid);
+               goto err_port_vid_learning_set;
+       }
+
+       err = mlxsw_sp_port_vlan_set(mlxsw_sp_port, vid, vid, true, false);
+       if (err) {
+               netdev_err(dev, "Failed to set VLAN membership for VID=%d\n",
+                          vid);
+               goto err_port_add_vid;
+       }
+
+       err = mlxsw_sp_port_stp_state_set(mlxsw_sp_port, vid,
+                                         MLXSW_REG_SPMS_STATE_FORWARDING);
+       if (err) {
+               netdev_err(dev, "Failed to set STP state for VID=%d\n", vid);
+               goto err_port_stp_state_set;
+       }
+
+       mlxsw_sp_port->nr_vfids++;
+       set_bit(vid, mlxsw_sp_port->active_vfids);
+
+       return 0;
+
+err_flood_table_config:
+err_flood_table_alloc:
+       mlxsw_sp_vfid_destroy(mlxsw_sp, vid);
+       return err;
+
+err_port_stp_state_set:
+       mlxsw_sp_port_vlan_set(mlxsw_sp_port, vid, vid, false, false);
+err_port_add_vid:
+       mlxsw_sp_port_vid_learning_set(mlxsw_sp_port, vid, true);
+err_port_vid_learning_set:
+       mlxsw_sp_port_vid_to_fid_set(mlxsw_sp_port,
+                                    MLXSW_REG_SVFA_MT_PORT_VID_TO_FID, false,
+                                    MLXSW_SP_VFID_BASE + vid, vid);
+err_port_vid_to_fid_set:
+       mlxsw_sp_port_vlan_mode_trans(mlxsw_sp_port);
+       return err;
+}
+
+int mlxsw_sp_port_kill_vid(struct net_device *dev,
+                          __be16 __always_unused proto, u16 vid)
+{
+       struct mlxsw_sp_port *mlxsw_sp_port = netdev_priv(dev);
+       int err;
+
+       /* VLAN 0 is removed from HW filter when device goes down, but
+        * it is reserved in our case, so simply return.
+        */
+       if (!vid)
+               return 0;
+
+       if (!test_bit(vid, mlxsw_sp_port->active_vfids)) {
+               netdev_warn(dev, "VID=%d does not exist\n", vid);
+               return 0;
+       }
+
+       err = mlxsw_sp_port_stp_state_set(mlxsw_sp_port, vid,
+                                         MLXSW_REG_SPMS_STATE_DISCARDING);
+       if (err) {
+               netdev_err(dev, "Failed to set STP state for VID=%d\n", vid);
+               return err;
+       }
+
+       err = mlxsw_sp_port_vlan_set(mlxsw_sp_port, vid, vid, false, false);
+       if (err) {
+               netdev_err(dev, "Failed to set VLAN membership for VID=%d\n",
+                          vid);
+               return err;
+       }
+
+       err = mlxsw_sp_port_vid_learning_set(mlxsw_sp_port, vid, true);
+       if (err) {
+               netdev_err(dev, "Failed to enable learning for VID=%d\n", vid);
+               return err;
+       }
+
+       err = mlxsw_sp_port_vid_to_fid_set(mlxsw_sp_port,
+                                          MLXSW_REG_SVFA_MT_PORT_VID_TO_FID,
+                                          false, MLXSW_SP_VFID_BASE + vid,
+                                          vid);
+       if (err) {
+               netdev_err(dev, "Failed to invalidate {Port, VID=%d} to vFID=%d mapping\n",
+                          vid, MLXSW_SP_VFID_BASE + vid);
+               return err;
+       }
+
+       /* When removing the last VLAN interface on a bridged port we need to
+        * transition all active 802.1Q bridge VLANs to use VID to FID
+        * mappings and set port's mode to VLAN mode.
+        */
+       if (mlxsw_sp_port->nr_vfids == 1) {
+               err = mlxsw_sp_port_vlan_mode_trans(mlxsw_sp_port);
+               if (err) {
+                       netdev_err(dev, "Failed to set to VLAN mode\n");
+                       return err;
+               }
+       }
+
+       mlxsw_sp_port->nr_vfids--;
+       clear_bit(vid, mlxsw_sp_port->active_vfids);
+
+       return 0;
+}
+
+static const struct net_device_ops mlxsw_sp_port_netdev_ops = {
+       .ndo_open               = mlxsw_sp_port_open,
+       .ndo_stop               = mlxsw_sp_port_stop,
+       .ndo_start_xmit         = mlxsw_sp_port_xmit,
+       .ndo_set_mac_address    = mlxsw_sp_port_set_mac_address,
+       .ndo_change_mtu         = mlxsw_sp_port_change_mtu,
+       .ndo_get_stats64        = mlxsw_sp_port_get_stats64,
+       .ndo_vlan_rx_add_vid    = mlxsw_sp_port_add_vid,
+       .ndo_vlan_rx_kill_vid   = mlxsw_sp_port_kill_vid,
+       .ndo_fdb_add            = switchdev_port_fdb_add,
+       .ndo_fdb_del            = switchdev_port_fdb_del,
+       .ndo_fdb_dump           = switchdev_port_fdb_dump,
+       .ndo_bridge_setlink     = switchdev_port_bridge_setlink,
+       .ndo_bridge_getlink     = switchdev_port_bridge_getlink,
+       .ndo_bridge_dellink     = switchdev_port_bridge_dellink,
+};
+
+static void mlxsw_sp_port_get_drvinfo(struct net_device *dev,
+                                     struct ethtool_drvinfo *drvinfo)
+{
+       struct mlxsw_sp_port *mlxsw_sp_port = netdev_priv(dev);
+       struct mlxsw_sp *mlxsw_sp = mlxsw_sp_port->mlxsw_sp;
+
+       strlcpy(drvinfo->driver, mlxsw_sp_driver_name, sizeof(drvinfo->driver));
+       strlcpy(drvinfo->version, mlxsw_sp_driver_version,
+               sizeof(drvinfo->version));
+       snprintf(drvinfo->fw_version, sizeof(drvinfo->fw_version),
+                "%d.%d.%d",
+                mlxsw_sp->bus_info->fw_rev.major,
+                mlxsw_sp->bus_info->fw_rev.minor,
+                mlxsw_sp->bus_info->fw_rev.subminor);
+       strlcpy(drvinfo->bus_info, mlxsw_sp->bus_info->device_name,
+               sizeof(drvinfo->bus_info));
+}
+
+struct mlxsw_sp_port_hw_stats {
+       char str[ETH_GSTRING_LEN];
+       u64 (*getter)(char *payload);
+};
+
+static const struct mlxsw_sp_port_hw_stats mlxsw_sp_port_hw_stats[] = {
+       {
+               .str = "a_frames_transmitted_ok",
+               .getter = mlxsw_reg_ppcnt_a_frames_transmitted_ok_get,
+       },
+       {
+               .str = "a_frames_received_ok",
+               .getter = mlxsw_reg_ppcnt_a_frames_received_ok_get,
+       },
+       {
+               .str = "a_frame_check_sequence_errors",
+               .getter = mlxsw_reg_ppcnt_a_frame_check_sequence_errors_get,
+       },
+       {
+               .str = "a_alignment_errors",
+               .getter = mlxsw_reg_ppcnt_a_alignment_errors_get,
+       },
+       {
+               .str = "a_octets_transmitted_ok",
+               .getter = mlxsw_reg_ppcnt_a_octets_transmitted_ok_get,
+       },
+       {
+               .str = "a_octets_received_ok",
+               .getter = mlxsw_reg_ppcnt_a_octets_received_ok_get,
+       },
+       {
+               .str = "a_multicast_frames_xmitted_ok",
+               .getter = mlxsw_reg_ppcnt_a_multicast_frames_xmitted_ok_get,
+       },
+       {
+               .str = "a_broadcast_frames_xmitted_ok",
+               .getter = mlxsw_reg_ppcnt_a_broadcast_frames_xmitted_ok_get,
+       },
+       {
+               .str = "a_multicast_frames_received_ok",
+               .getter = mlxsw_reg_ppcnt_a_multicast_frames_received_ok_get,
+       },
+       {
+               .str = "a_broadcast_frames_received_ok",
+               .getter = mlxsw_reg_ppcnt_a_broadcast_frames_received_ok_get,
+       },
+       {
+               .str = "a_in_range_length_errors",
+               .getter = mlxsw_reg_ppcnt_a_in_range_length_errors_get,
+       },
+       {
+               .str = "a_out_of_range_length_field",
+               .getter = mlxsw_reg_ppcnt_a_out_of_range_length_field_get,
+       },
+       {
+               .str = "a_frame_too_long_errors",
+               .getter = mlxsw_reg_ppcnt_a_frame_too_long_errors_get,
+       },
+       {
+               .str = "a_symbol_error_during_carrier",
+               .getter = mlxsw_reg_ppcnt_a_symbol_error_during_carrier_get,
+       },
+       {
+               .str = "a_mac_control_frames_transmitted",
+               .getter = mlxsw_reg_ppcnt_a_mac_control_frames_transmitted_get,
+       },
+       {
+               .str = "a_mac_control_frames_received",
+               .getter = mlxsw_reg_ppcnt_a_mac_control_frames_received_get,
+       },
+       {
+               .str = "a_unsupported_opcodes_received",
+               .getter = mlxsw_reg_ppcnt_a_unsupported_opcodes_received_get,
+       },
+       {
+               .str = "a_pause_mac_ctrl_frames_received",
+               .getter = mlxsw_reg_ppcnt_a_pause_mac_ctrl_frames_received_get,
+       },
+       {
+               .str = "a_pause_mac_ctrl_frames_xmitted",
+               .getter = mlxsw_reg_ppcnt_a_pause_mac_ctrl_frames_transmitted_get,
+       },
+};
+
+#define MLXSW_SP_PORT_HW_STATS_LEN ARRAY_SIZE(mlxsw_sp_port_hw_stats)
+
+static void mlxsw_sp_port_get_strings(struct net_device *dev,
+                                     u32 stringset, u8 *data)
+{
+       u8 *p = data;
+       int i;
+
+       switch (stringset) {
+       case ETH_SS_STATS:
+               for (i = 0; i < MLXSW_SP_PORT_HW_STATS_LEN; i++) {
+                       memcpy(p, mlxsw_sp_port_hw_stats[i].str,
+                              ETH_GSTRING_LEN);
+                       p += ETH_GSTRING_LEN;
+               }
+               break;
+       }
+}
+
+static void mlxsw_sp_port_get_stats(struct net_device *dev,
+                                   struct ethtool_stats *stats, u64 *data)
+{
+       struct mlxsw_sp_port *mlxsw_sp_port = netdev_priv(dev);
+       struct mlxsw_sp *mlxsw_sp = mlxsw_sp_port->mlxsw_sp;
+       char ppcnt_pl[MLXSW_REG_PPCNT_LEN];
+       int i;
+       int err;
+
+       mlxsw_reg_ppcnt_pack(ppcnt_pl, mlxsw_sp_port->local_port);
+       err = mlxsw_reg_query(mlxsw_sp->core, MLXSW_REG(ppcnt), ppcnt_pl);
+       for (i = 0; i < MLXSW_SP_PORT_HW_STATS_LEN; i++)
+               data[i] = !err ? mlxsw_sp_port_hw_stats[i].getter(ppcnt_pl) : 0;
+}
+
+static int mlxsw_sp_port_get_sset_count(struct net_device *dev, int sset)
+{
+       switch (sset) {
+       case ETH_SS_STATS:
+               return MLXSW_SP_PORT_HW_STATS_LEN;
+       default:
+               return -EOPNOTSUPP;
+       }
+}
+
+struct mlxsw_sp_port_link_mode {
+       u32 mask;
+       u32 supported;
+       u32 advertised;
+       u32 speed;
+};
+
+static const struct mlxsw_sp_port_link_mode mlxsw_sp_port_link_mode[] = {
+       {
+               .mask           = MLXSW_REG_PTYS_ETH_SPEED_100BASE_T,
+               .supported      = SUPPORTED_100baseT_Full,
+               .advertised     = ADVERTISED_100baseT_Full,
+               .speed          = 100,
+       },
+       {
+               .mask           = MLXSW_REG_PTYS_ETH_SPEED_100BASE_TX,
+               .speed          = 100,
+       },
+       {
+               .mask           = MLXSW_REG_PTYS_ETH_SPEED_SGMII |
+                                 MLXSW_REG_PTYS_ETH_SPEED_1000BASE_KX,
+               .supported      = SUPPORTED_1000baseKX_Full,
+               .advertised     = ADVERTISED_1000baseKX_Full,
+               .speed          = 1000,
+       },
+       {
+               .mask           = MLXSW_REG_PTYS_ETH_SPEED_10GBASE_T,
+               .supported      = SUPPORTED_10000baseT_Full,
+               .advertised     = ADVERTISED_10000baseT_Full,
+               .speed          = 10000,
+       },
+       {
+               .mask           = MLXSW_REG_PTYS_ETH_SPEED_10GBASE_CX4 |
+                                 MLXSW_REG_PTYS_ETH_SPEED_10GBASE_KX4,
+               .supported      = SUPPORTED_10000baseKX4_Full,
+               .advertised     = ADVERTISED_10000baseKX4_Full,
+               .speed          = 10000,
+       },
+       {
+               .mask           = MLXSW_REG_PTYS_ETH_SPEED_10GBASE_KR |
+                                 MLXSW_REG_PTYS_ETH_SPEED_10GBASE_CR |
+                                 MLXSW_REG_PTYS_ETH_SPEED_10GBASE_SR |
+                                 MLXSW_REG_PTYS_ETH_SPEED_10GBASE_ER_LR,
+               .supported      = SUPPORTED_10000baseKR_Full,
+               .advertised     = ADVERTISED_10000baseKR_Full,
+               .speed          = 10000,
+       },
+       {
+               .mask           = MLXSW_REG_PTYS_ETH_SPEED_20GBASE_KR2,
+               .supported      = SUPPORTED_20000baseKR2_Full,
+               .advertised     = ADVERTISED_20000baseKR2_Full,
+               .speed          = 20000,
+       },
+       {
+               .mask           = MLXSW_REG_PTYS_ETH_SPEED_40GBASE_CR4,
+               .supported      = SUPPORTED_40000baseCR4_Full,
+               .advertised     = ADVERTISED_40000baseCR4_Full,
+               .speed          = 40000,
+       },
+       {
+               .mask           = MLXSW_REG_PTYS_ETH_SPEED_40GBASE_KR4,
+               .supported      = SUPPORTED_40000baseKR4_Full,
+               .advertised     = ADVERTISED_40000baseKR4_Full,
+               .speed          = 40000,
+       },
+       {
+               .mask           = MLXSW_REG_PTYS_ETH_SPEED_40GBASE_SR4,
+               .supported      = SUPPORTED_40000baseSR4_Full,
+               .advertised     = ADVERTISED_40000baseSR4_Full,
+               .speed          = 40000,
+       },
+       {
+               .mask           = MLXSW_REG_PTYS_ETH_SPEED_40GBASE_LR4_ER4,
+               .supported      = SUPPORTED_40000baseLR4_Full,
+               .advertised     = ADVERTISED_40000baseLR4_Full,
+               .speed          = 40000,
+       },
+       {
+               .mask           = MLXSW_REG_PTYS_ETH_SPEED_25GBASE_CR |
+                                 MLXSW_REG_PTYS_ETH_SPEED_25GBASE_KR |
+                                 MLXSW_REG_PTYS_ETH_SPEED_25GBASE_SR,
+               .speed          = 25000,
+       },
+       {
+               .mask           = MLXSW_REG_PTYS_ETH_SPEED_50GBASE_KR4 |
+                                 MLXSW_REG_PTYS_ETH_SPEED_50GBASE_CR2 |
+                                 MLXSW_REG_PTYS_ETH_SPEED_50GBASE_KR2,
+               .speed          = 50000,
+       },
+       {
+               .mask           = MLXSW_REG_PTYS_ETH_SPEED_56GBASE_R4,
+               .supported      = SUPPORTED_56000baseKR4_Full,
+               .advertised     = ADVERTISED_56000baseKR4_Full,
+               .speed          = 56000,
+       },
+       {
+               .mask           = MLXSW_REG_PTYS_ETH_SPEED_100GBASE_CR4 |
+                                 MLXSW_REG_PTYS_ETH_SPEED_100GBASE_SR4 |
+                                 MLXSW_REG_PTYS_ETH_SPEED_100GBASE_KR4 |
+                                 MLXSW_REG_PTYS_ETH_SPEED_100GBASE_LR4_ER4,
+               .speed          = 100000,
+       },
+};
+
+#define MLXSW_SP_PORT_LINK_MODE_LEN ARRAY_SIZE(mlxsw_sp_port_link_mode)
+
+static u32 mlxsw_sp_from_ptys_supported_port(u32 ptys_eth_proto)
+{
+       if (ptys_eth_proto & (MLXSW_REG_PTYS_ETH_SPEED_10GBASE_CR |
+                             MLXSW_REG_PTYS_ETH_SPEED_10GBASE_SR |
+                             MLXSW_REG_PTYS_ETH_SPEED_40GBASE_CR4 |
+                             MLXSW_REG_PTYS_ETH_SPEED_40GBASE_SR4 |
+                             MLXSW_REG_PTYS_ETH_SPEED_100GBASE_SR4 |
+                             MLXSW_REG_PTYS_ETH_SPEED_SGMII))
+               return SUPPORTED_FIBRE;
+
+       if (ptys_eth_proto & (MLXSW_REG_PTYS_ETH_SPEED_10GBASE_KR |
+                             MLXSW_REG_PTYS_ETH_SPEED_10GBASE_KX4 |
+                             MLXSW_REG_PTYS_ETH_SPEED_40GBASE_KR4 |
+                             MLXSW_REG_PTYS_ETH_SPEED_100GBASE_KR4 |
+                             MLXSW_REG_PTYS_ETH_SPEED_1000BASE_KX))
+               return SUPPORTED_Backplane;
+       return 0;
+}
+
+static u32 mlxsw_sp_from_ptys_supported_link(u32 ptys_eth_proto)
+{
+       u32 modes = 0;
+       int i;
+
+       for (i = 0; i < MLXSW_SP_PORT_LINK_MODE_LEN; i++) {
+               if (ptys_eth_proto & mlxsw_sp_port_link_mode[i].mask)
+                       modes |= mlxsw_sp_port_link_mode[i].supported;
+       }
+       return modes;
+}
+
+static u32 mlxsw_sp_from_ptys_advert_link(u32 ptys_eth_proto)
+{
+       u32 modes = 0;
+       int i;
+
+       for (i = 0; i < MLXSW_SP_PORT_LINK_MODE_LEN; i++) {
+               if (ptys_eth_proto & mlxsw_sp_port_link_mode[i].mask)
+                       modes |= mlxsw_sp_port_link_mode[i].advertised;
+       }
+       return modes;
+}
+
+static void mlxsw_sp_from_ptys_speed_duplex(bool carrier_ok, u32 ptys_eth_proto,
+                                           struct ethtool_cmd *cmd)
+{
+       u32 speed = SPEED_UNKNOWN;
+       u8 duplex = DUPLEX_UNKNOWN;
+       int i;
+
+       if (!carrier_ok)
+               goto out;
+
+       for (i = 0; i < MLXSW_SP_PORT_LINK_MODE_LEN; i++) {
+               if (ptys_eth_proto & mlxsw_sp_port_link_mode[i].mask) {
+                       speed = mlxsw_sp_port_link_mode[i].speed;
+                       duplex = DUPLEX_FULL;
+                       break;
+               }
+       }
+out:
+       ethtool_cmd_speed_set(cmd, speed);
+       cmd->duplex = duplex;
+}
+
+static u8 mlxsw_sp_port_connector_port(u32 ptys_eth_proto)
+{
+       if (ptys_eth_proto & (MLXSW_REG_PTYS_ETH_SPEED_10GBASE_SR |
+                             MLXSW_REG_PTYS_ETH_SPEED_40GBASE_SR4 |
+                             MLXSW_REG_PTYS_ETH_SPEED_100GBASE_SR4 |
+                             MLXSW_REG_PTYS_ETH_SPEED_SGMII))
+               return PORT_FIBRE;
+
+       if (ptys_eth_proto & (MLXSW_REG_PTYS_ETH_SPEED_10GBASE_CR |
+                             MLXSW_REG_PTYS_ETH_SPEED_40GBASE_CR4 |
+                             MLXSW_REG_PTYS_ETH_SPEED_100GBASE_CR4))
+               return PORT_DA;
+
+       if (ptys_eth_proto & (MLXSW_REG_PTYS_ETH_SPEED_10GBASE_KR |
+                             MLXSW_REG_PTYS_ETH_SPEED_10GBASE_KX4 |
+                             MLXSW_REG_PTYS_ETH_SPEED_40GBASE_KR4 |
+                             MLXSW_REG_PTYS_ETH_SPEED_100GBASE_KR4))
+               return PORT_NONE;
+
+       return PORT_OTHER;
+}
+
+static int mlxsw_sp_port_get_settings(struct net_device *dev,
+                                     struct ethtool_cmd *cmd)
+{
+       struct mlxsw_sp_port *mlxsw_sp_port = netdev_priv(dev);
+       struct mlxsw_sp *mlxsw_sp = mlxsw_sp_port->mlxsw_sp;
+       char ptys_pl[MLXSW_REG_PTYS_LEN];
+       u32 eth_proto_cap;
+       u32 eth_proto_admin;
+       u32 eth_proto_oper;
+       int err;
+
+       mlxsw_reg_ptys_pack(ptys_pl, mlxsw_sp_port->local_port, 0);
+       err = mlxsw_reg_query(mlxsw_sp->core, MLXSW_REG(ptys), ptys_pl);
+       if (err) {
+               netdev_err(dev, "Failed to get proto");
+               return err;
+       }
+       mlxsw_reg_ptys_unpack(ptys_pl, &eth_proto_cap,
+                             &eth_proto_admin, &eth_proto_oper);
+
+       cmd->supported = mlxsw_sp_from_ptys_supported_port(eth_proto_cap) |
+                        mlxsw_sp_from_ptys_supported_link(eth_proto_cap) |
+                        SUPPORTED_Pause | SUPPORTED_Asym_Pause;
+       cmd->advertising = mlxsw_sp_from_ptys_advert_link(eth_proto_admin);
+       mlxsw_sp_from_ptys_speed_duplex(netif_carrier_ok(dev),
+                                       eth_proto_oper, cmd);
+
+       eth_proto_oper = eth_proto_oper ? eth_proto_oper : eth_proto_cap;
+       cmd->port = mlxsw_sp_port_connector_port(eth_proto_oper);
+       cmd->lp_advertising = mlxsw_sp_from_ptys_advert_link(eth_proto_oper);
+
+       cmd->transceiver = XCVR_INTERNAL;
+       return 0;
+}
+
+static u32 mlxsw_sp_to_ptys_advert_link(u32 advertising)
+{
+       u32 ptys_proto = 0;
+       int i;
+
+       for (i = 0; i < MLXSW_SP_PORT_LINK_MODE_LEN; i++) {
+               if (advertising & mlxsw_sp_port_link_mode[i].advertised)
+                       ptys_proto |= mlxsw_sp_port_link_mode[i].mask;
+       }
+       return ptys_proto;
+}
+
+static u32 mlxsw_sp_to_ptys_speed(u32 speed)
+{
+       u32 ptys_proto = 0;
+       int i;
+
+       for (i = 0; i < MLXSW_SP_PORT_LINK_MODE_LEN; i++) {
+               if (speed == mlxsw_sp_port_link_mode[i].speed)
+                       ptys_proto |= mlxsw_sp_port_link_mode[i].mask;
+       }
+       return ptys_proto;
+}
+
+static int mlxsw_sp_port_set_settings(struct net_device *dev,
+                                     struct ethtool_cmd *cmd)
+{
+       struct mlxsw_sp_port *mlxsw_sp_port = netdev_priv(dev);
+       struct mlxsw_sp *mlxsw_sp = mlxsw_sp_port->mlxsw_sp;
+       char ptys_pl[MLXSW_REG_PTYS_LEN];
+       u32 speed;
+       u32 eth_proto_new;
+       u32 eth_proto_cap;
+       u32 eth_proto_admin;
+       bool is_up;
+       int err;
+
+       speed = ethtool_cmd_speed(cmd);
+
+       eth_proto_new = cmd->autoneg == AUTONEG_ENABLE ?
+               mlxsw_sp_to_ptys_advert_link(cmd->advertising) :
+               mlxsw_sp_to_ptys_speed(speed);
+
+       mlxsw_reg_ptys_pack(ptys_pl, mlxsw_sp_port->local_port, 0);
+       err = mlxsw_reg_query(mlxsw_sp->core, MLXSW_REG(ptys), ptys_pl);
+       if (err) {
+               netdev_err(dev, "Failed to get proto");
+               return err;
+       }
+       mlxsw_reg_ptys_unpack(ptys_pl, &eth_proto_cap, &eth_proto_admin, NULL);
+
+       eth_proto_new = eth_proto_new & eth_proto_cap;
+       if (!eth_proto_new) {
+               netdev_err(dev, "Not supported proto admin requested");
+               return -EINVAL;
+       }
+       if (eth_proto_new == eth_proto_admin)
+               return 0;
+
+       mlxsw_reg_ptys_pack(ptys_pl, mlxsw_sp_port->local_port, eth_proto_new);
+       err = mlxsw_reg_write(mlxsw_sp->core, MLXSW_REG(ptys), ptys_pl);
+       if (err) {
+               netdev_err(dev, "Failed to set proto admin");
+               return err;
+       }
+
+       err = mlxsw_sp_port_oper_status_get(mlxsw_sp_port, &is_up);
+       if (err) {
+               netdev_err(dev, "Failed to get oper status");
+               return err;
+       }
+       if (!is_up)
+               return 0;
+
+       err = mlxsw_sp_port_admin_status_set(mlxsw_sp_port, false);
+       if (err) {
+               netdev_err(dev, "Failed to set admin status");
+               return err;
+       }
+
+       err = mlxsw_sp_port_admin_status_set(mlxsw_sp_port, true);
+       if (err) {
+               netdev_err(dev, "Failed to set admin status");
+               return err;
+       }
+
+       return 0;
+}
+
+static const struct ethtool_ops mlxsw_sp_port_ethtool_ops = {
+       .get_drvinfo            = mlxsw_sp_port_get_drvinfo,
+       .get_link               = ethtool_op_get_link,
+       .get_strings            = mlxsw_sp_port_get_strings,
+       .get_ethtool_stats      = mlxsw_sp_port_get_stats,
+       .get_sset_count         = mlxsw_sp_port_get_sset_count,
+       .get_settings           = mlxsw_sp_port_get_settings,
+       .set_settings           = mlxsw_sp_port_set_settings,
+};
+
+static int mlxsw_sp_port_create(struct mlxsw_sp *mlxsw_sp, u8 local_port)
+{
+       struct mlxsw_sp_port *mlxsw_sp_port;
+       struct net_device *dev;
+       bool usable;
+       int err;
+
+       dev = alloc_etherdev(sizeof(struct mlxsw_sp_port));
+       if (!dev)
+               return -ENOMEM;
+       mlxsw_sp_port = netdev_priv(dev);
+       mlxsw_sp_port->dev = dev;
+       mlxsw_sp_port->mlxsw_sp = mlxsw_sp;
+       mlxsw_sp_port->local_port = local_port;
+       mlxsw_sp_port->learning = 1;
+       mlxsw_sp_port->learning_sync = 1;
+       mlxsw_sp_port->pvid = 1;
+
+       mlxsw_sp_port->pcpu_stats =
+               netdev_alloc_pcpu_stats(struct mlxsw_sp_port_pcpu_stats);
+       if (!mlxsw_sp_port->pcpu_stats) {
+               err = -ENOMEM;
+               goto err_alloc_stats;
+       }
+
+       dev->netdev_ops = &mlxsw_sp_port_netdev_ops;
+       dev->ethtool_ops = &mlxsw_sp_port_ethtool_ops;
+
+       err = mlxsw_sp_port_dev_addr_init(mlxsw_sp_port);
+       if (err) {
+               dev_err(mlxsw_sp->bus_info->dev, "Port %d: Unable to init port mac address\n",
+                       mlxsw_sp_port->local_port);
+               goto err_dev_addr_init;
+       }
+
+       netif_carrier_off(dev);
+
+       dev->features |= NETIF_F_NETNS_LOCAL | NETIF_F_LLTX | NETIF_F_SG |
+                        NETIF_F_HW_VLAN_CTAG_FILTER;
+
+       /* Each packet needs to have a Tx header (metadata) on top all other
+        * headers.
+        */
+       dev->hard_header_len += MLXSW_TXHDR_LEN;
+
+       err = mlxsw_sp_port_module_check(mlxsw_sp_port, &usable);
+       if (err) {
+               dev_err(mlxsw_sp->bus_info->dev, "Port %d: Failed to check module\n",
+                       mlxsw_sp_port->local_port);
+               goto err_port_module_check;
+       }
+
+       if (!usable) {
+               dev_dbg(mlxsw_sp->bus_info->dev, "Port %d: Not usable, skipping initialization\n",
+                       mlxsw_sp_port->local_port);
+               goto port_not_usable;
+       }
+
+       err = mlxsw_sp_port_system_port_mapping_set(mlxsw_sp_port);
+       if (err) {
+               dev_err(mlxsw_sp->bus_info->dev, "Port %d: Failed to set system port mapping\n",
+                       mlxsw_sp_port->local_port);
+               goto err_port_system_port_mapping_set;
+       }
+
+       err = mlxsw_sp_port_swid_set(mlxsw_sp_port, 0);
+       if (err) {
+               dev_err(mlxsw_sp->bus_info->dev, "Port %d: Failed to set SWID\n",
+                       mlxsw_sp_port->local_port);
+               goto err_port_swid_set;
+       }
+
+       err = mlxsw_sp_port_mtu_set(mlxsw_sp_port, ETH_DATA_LEN);
+       if (err) {
+               dev_err(mlxsw_sp->bus_info->dev, "Port %d: Failed to set MTU\n",
+                       mlxsw_sp_port->local_port);
+               goto err_port_mtu_set;
+       }
+
+       err = mlxsw_sp_port_admin_status_set(mlxsw_sp_port, false);
+       if (err)
+               goto err_port_admin_status_set;
+
+       err = mlxsw_sp_port_buffers_init(mlxsw_sp_port);
+       if (err) {
+               dev_err(mlxsw_sp->bus_info->dev, "Port %d: Failed to initialize buffers\n",
+                       mlxsw_sp_port->local_port);
+               goto err_port_buffers_init;
+       }
+
+       mlxsw_sp_port_switchdev_init(mlxsw_sp_port);
+       err = register_netdev(dev);
+       if (err) {
+               dev_err(mlxsw_sp->bus_info->dev, "Port %d: Failed to register netdev\n",
+                       mlxsw_sp_port->local_port);
+               goto err_register_netdev;
+       }
+
+       err = mlxsw_sp_port_vlan_init(mlxsw_sp_port);
+       if (err)
+               goto err_port_vlan_init;
+
+       mlxsw_sp->ports[local_port] = mlxsw_sp_port;
+       return 0;
+
+err_port_vlan_init:
+       unregister_netdev(dev);
+err_register_netdev:
+err_port_buffers_init:
+err_port_admin_status_set:
+err_port_mtu_set:
+err_port_swid_set:
+err_port_system_port_mapping_set:
+port_not_usable:
+err_port_module_check:
+err_dev_addr_init:
+       free_percpu(mlxsw_sp_port->pcpu_stats);
+err_alloc_stats:
+       free_netdev(dev);
+       return err;
+}
+
+static void mlxsw_sp_vfids_fini(struct mlxsw_sp *mlxsw_sp)
+{
+       u16 vfid;
+
+       for_each_set_bit(vfid, mlxsw_sp->active_vfids, VLAN_N_VID)
+               mlxsw_sp_vfid_destroy(mlxsw_sp, vfid);
+}
+
+static void mlxsw_sp_port_remove(struct mlxsw_sp *mlxsw_sp, u8 local_port)
+{
+       struct mlxsw_sp_port *mlxsw_sp_port = mlxsw_sp->ports[local_port];
+
+       if (!mlxsw_sp_port)
+               return;
+       mlxsw_sp_port_kill_vid(mlxsw_sp_port->dev, 0, 1);
+       unregister_netdev(mlxsw_sp_port->dev); /* This calls ndo_stop */
+       mlxsw_sp_port_switchdev_fini(mlxsw_sp_port);
+       free_percpu(mlxsw_sp_port->pcpu_stats);
+       free_netdev(mlxsw_sp_port->dev);
+}
+
+static void mlxsw_sp_ports_remove(struct mlxsw_sp *mlxsw_sp)
+{
+       int i;
+
+       for (i = 1; i < MLXSW_PORT_MAX_PORTS; i++)
+               mlxsw_sp_port_remove(mlxsw_sp, i);
+       kfree(mlxsw_sp->ports);
+}
+
+static int mlxsw_sp_ports_create(struct mlxsw_sp *mlxsw_sp)
+{
+       size_t alloc_size;
+       int i;
+       int err;
+
+       alloc_size = sizeof(struct mlxsw_sp_port *) * MLXSW_PORT_MAX_PORTS;
+       mlxsw_sp->ports = kzalloc(alloc_size, GFP_KERNEL);
+       if (!mlxsw_sp->ports)
+               return -ENOMEM;
+
+       for (i = 1; i < MLXSW_PORT_MAX_PORTS; i++) {
+               err = mlxsw_sp_port_create(mlxsw_sp, i);
+               if (err)
+                       goto err_port_create;
+       }
+       return 0;
+
+err_port_create:
+       for (i--; i >= 1; i--)
+               mlxsw_sp_port_remove(mlxsw_sp, i);
+       kfree(mlxsw_sp->ports);
+       return err;
+}
+
+static void mlxsw_sp_pude_event_func(const struct mlxsw_reg_info *reg,
+                                    char *pude_pl, void *priv)
+{
+       struct mlxsw_sp *mlxsw_sp = priv;
+       struct mlxsw_sp_port *mlxsw_sp_port;
+       enum mlxsw_reg_pude_oper_status status;
+       u8 local_port;
+
+       local_port = mlxsw_reg_pude_local_port_get(pude_pl);
+       mlxsw_sp_port = mlxsw_sp->ports[local_port];
+       if (!mlxsw_sp_port) {
+               dev_warn(mlxsw_sp->bus_info->dev, "Port %d: Link event received for non-existent port\n",
+                        local_port);
+               return;
+       }
+
+       status = mlxsw_reg_pude_oper_status_get(pude_pl);
+       if (status == MLXSW_PORT_OPER_STATUS_UP) {
+               netdev_info(mlxsw_sp_port->dev, "link up\n");
+               netif_carrier_on(mlxsw_sp_port->dev);
+       } else {
+               netdev_info(mlxsw_sp_port->dev, "link down\n");
+               netif_carrier_off(mlxsw_sp_port->dev);
+       }
+}
+
+static struct mlxsw_event_listener mlxsw_sp_pude_event = {
+       .func = mlxsw_sp_pude_event_func,
+       .trap_id = MLXSW_TRAP_ID_PUDE,
+};
+
+static int mlxsw_sp_event_register(struct mlxsw_sp *mlxsw_sp,
+                                  enum mlxsw_event_trap_id trap_id)
+{
+       struct mlxsw_event_listener *el;
+       char hpkt_pl[MLXSW_REG_HPKT_LEN];
+       int err;
+
+       switch (trap_id) {
+       case MLXSW_TRAP_ID_PUDE:
+               el = &mlxsw_sp_pude_event;
+               break;
+       }
+       err = mlxsw_core_event_listener_register(mlxsw_sp->core, el, mlxsw_sp);
+       if (err)
+               return err;
+
+       mlxsw_reg_hpkt_pack(hpkt_pl, MLXSW_REG_HPKT_ACTION_FORWARD, trap_id);
+       err = mlxsw_reg_write(mlxsw_sp->core, MLXSW_REG(hpkt), hpkt_pl);
+       if (err)
+               goto err_event_trap_set;
+
+       return 0;
+
+err_event_trap_set:
+       mlxsw_core_event_listener_unregister(mlxsw_sp->core, el, mlxsw_sp);
+       return err;
+}
+
+static void mlxsw_sp_event_unregister(struct mlxsw_sp *mlxsw_sp,
+                                     enum mlxsw_event_trap_id trap_id)
+{
+       struct mlxsw_event_listener *el;
+
+       switch (trap_id) {
+       case MLXSW_TRAP_ID_PUDE:
+               el = &mlxsw_sp_pude_event;
+               break;
+       }
+       mlxsw_core_event_listener_unregister(mlxsw_sp->core, el, mlxsw_sp);
+}
+
+static void mlxsw_sp_rx_listener_func(struct sk_buff *skb, u8 local_port,
+                                     void *priv)
+{
+       struct mlxsw_sp *mlxsw_sp = priv;
+       struct mlxsw_sp_port *mlxsw_sp_port = mlxsw_sp->ports[local_port];
+       struct mlxsw_sp_port_pcpu_stats *pcpu_stats;
+
+       if (unlikely(!mlxsw_sp_port)) {
+               dev_warn_ratelimited(mlxsw_sp->bus_info->dev, "Port %d: skb received for non-existent port\n",
+                                    local_port);
+               return;
+       }
+
+       skb->dev = mlxsw_sp_port->dev;
+
+       pcpu_stats = this_cpu_ptr(mlxsw_sp_port->pcpu_stats);
+       u64_stats_update_begin(&pcpu_stats->syncp);
+       pcpu_stats->rx_packets++;
+       pcpu_stats->rx_bytes += skb->len;
+       u64_stats_update_end(&pcpu_stats->syncp);
+
+       skb->protocol = eth_type_trans(skb, skb->dev);
+       netif_receive_skb(skb);
+}
+
+static const struct mlxsw_rx_listener mlxsw_sp_rx_listener[] = {
+       {
+               .func = mlxsw_sp_rx_listener_func,
+               .local_port = MLXSW_PORT_DONT_CARE,
+               .trap_id = MLXSW_TRAP_ID_FDB_MC,
+       },
+       /* Traps for specific L2 packet types, not trapped as FDB MC */
+       {
+               .func = mlxsw_sp_rx_listener_func,
+               .local_port = MLXSW_PORT_DONT_CARE,
+               .trap_id = MLXSW_TRAP_ID_STP,
+       },
+       {
+               .func = mlxsw_sp_rx_listener_func,
+               .local_port = MLXSW_PORT_DONT_CARE,
+               .trap_id = MLXSW_TRAP_ID_LACP,
+       },
+       {
+               .func = mlxsw_sp_rx_listener_func,
+               .local_port = MLXSW_PORT_DONT_CARE,
+               .trap_id = MLXSW_TRAP_ID_EAPOL,
+       },
+       {
+               .func = mlxsw_sp_rx_listener_func,
+               .local_port = MLXSW_PORT_DONT_CARE,
+               .trap_id = MLXSW_TRAP_ID_LLDP,
+       },
+       {
+               .func = mlxsw_sp_rx_listener_func,
+               .local_port = MLXSW_PORT_DONT_CARE,
+               .trap_id = MLXSW_TRAP_ID_MMRP,
+       },
+       {
+               .func = mlxsw_sp_rx_listener_func,
+               .local_port = MLXSW_PORT_DONT_CARE,
+               .trap_id = MLXSW_TRAP_ID_MVRP,
+       },
+       {
+               .func = mlxsw_sp_rx_listener_func,
+               .local_port = MLXSW_PORT_DONT_CARE,
+               .trap_id = MLXSW_TRAP_ID_RPVST,
+       },
+       {
+               .func = mlxsw_sp_rx_listener_func,
+               .local_port = MLXSW_PORT_DONT_CARE,
+               .trap_id = MLXSW_TRAP_ID_DHCP,
+       },
+       {
+               .func = mlxsw_sp_rx_listener_func,
+               .local_port = MLXSW_PORT_DONT_CARE,
+               .trap_id = MLXSW_TRAP_ID_IGMP_QUERY,
+       },
+       {
+               .func = mlxsw_sp_rx_listener_func,
+               .local_port = MLXSW_PORT_DONT_CARE,
+               .trap_id = MLXSW_TRAP_ID_IGMP_V1_REPORT,
+       },
+       {
+               .func = mlxsw_sp_rx_listener_func,
+               .local_port = MLXSW_PORT_DONT_CARE,
+               .trap_id = MLXSW_TRAP_ID_IGMP_V2_REPORT,
+       },
+       {
+               .func = mlxsw_sp_rx_listener_func,
+               .local_port = MLXSW_PORT_DONT_CARE,
+               .trap_id = MLXSW_TRAP_ID_IGMP_V2_LEAVE,
+       },
+       {
+               .func = mlxsw_sp_rx_listener_func,
+               .local_port = MLXSW_PORT_DONT_CARE,
+               .trap_id = MLXSW_TRAP_ID_IGMP_V3_REPORT,
+       },
+};
+
+static int mlxsw_sp_traps_init(struct mlxsw_sp *mlxsw_sp)
+{
+       char htgt_pl[MLXSW_REG_HTGT_LEN];
+       char hpkt_pl[MLXSW_REG_HPKT_LEN];
+       int i;
+       int err;
+
+       mlxsw_reg_htgt_pack(htgt_pl, MLXSW_REG_HTGT_TRAP_GROUP_RX);
+       err = mlxsw_reg_write(mlxsw_sp->core, MLXSW_REG(htgt), htgt_pl);
+       if (err)
+               return err;
+
+       mlxsw_reg_htgt_pack(htgt_pl, MLXSW_REG_HTGT_TRAP_GROUP_CTRL);
+       err = mlxsw_reg_write(mlxsw_sp->core, MLXSW_REG(htgt), htgt_pl);
+       if (err)
+               return err;
+
+       for (i = 0; i < ARRAY_SIZE(mlxsw_sp_rx_listener); i++) {
+               err = mlxsw_core_rx_listener_register(mlxsw_sp->core,
+                                                     &mlxsw_sp_rx_listener[i],
+                                                     mlxsw_sp);
+               if (err)
+                       goto err_rx_listener_register;
+
+               mlxsw_reg_hpkt_pack(hpkt_pl, MLXSW_REG_HPKT_ACTION_TRAP_TO_CPU,
+                                   mlxsw_sp_rx_listener[i].trap_id);
+               err = mlxsw_reg_write(mlxsw_sp->core, MLXSW_REG(hpkt), hpkt_pl);
+               if (err)
+                       goto err_rx_trap_set;
+       }
+       return 0;
+
+err_rx_trap_set:
+       mlxsw_core_rx_listener_unregister(mlxsw_sp->core,
+                                         &mlxsw_sp_rx_listener[i],
+                                         mlxsw_sp);
+err_rx_listener_register:
+       for (i--; i >= 0; i--) {
+               mlxsw_reg_hpkt_pack(hpkt_pl, MLXSW_REG_HPKT_ACTION_FORWARD,
+                                   mlxsw_sp_rx_listener[i].trap_id);
+               mlxsw_reg_write(mlxsw_sp->core, MLXSW_REG(hpkt), hpkt_pl);
+
+               mlxsw_core_rx_listener_unregister(mlxsw_sp->core,
+                                                 &mlxsw_sp_rx_listener[i],
+                                                 mlxsw_sp);
+       }
+       return err;
+}
+
+static void mlxsw_sp_traps_fini(struct mlxsw_sp *mlxsw_sp)
+{
+       char hpkt_pl[MLXSW_REG_HPKT_LEN];
+       int i;
+
+       for (i = 0; i < ARRAY_SIZE(mlxsw_sp_rx_listener); i++) {
+               mlxsw_reg_hpkt_pack(hpkt_pl, MLXSW_REG_HPKT_ACTION_FORWARD,
+                                   mlxsw_sp_rx_listener[i].trap_id);
+               mlxsw_reg_write(mlxsw_sp->core, MLXSW_REG(hpkt), hpkt_pl);
+
+               mlxsw_core_rx_listener_unregister(mlxsw_sp->core,
+                                                 &mlxsw_sp_rx_listener[i],
+                                                 mlxsw_sp);
+       }
+}
+
+static int __mlxsw_sp_flood_init(struct mlxsw_core *mlxsw_core,
+                                enum mlxsw_reg_sfgc_type type,
+                                enum mlxsw_reg_sfgc_bridge_type bridge_type)
+{
+       enum mlxsw_flood_table_type table_type;
+       enum mlxsw_sp_flood_table flood_table;
+       char sfgc_pl[MLXSW_REG_SFGC_LEN];
+
+       if (bridge_type == MLXSW_REG_SFGC_BRIDGE_TYPE_VFID) {
+               table_type = MLXSW_REG_SFGC_TABLE_TYPE_FID;
+               flood_table = 0;
+       } else {
+               table_type = MLXSW_REG_SFGC_TABLE_TYPE_FID_OFFEST;
+               if (type == MLXSW_REG_SFGC_TYPE_UNKNOWN_UNICAST)
+                       flood_table = MLXSW_SP_FLOOD_TABLE_UC;
+               else
+                       flood_table = MLXSW_SP_FLOOD_TABLE_BM;
+       }
+
+       mlxsw_reg_sfgc_pack(sfgc_pl, type, bridge_type, table_type,
+                           flood_table);
+       return mlxsw_reg_write(mlxsw_core, MLXSW_REG(sfgc), sfgc_pl);
+}
+
+static int mlxsw_sp_flood_init(struct mlxsw_sp *mlxsw_sp)
+{
+       int type, err;
+
+       /* For non-offloaded netdevs, flood all traffic types to CPU
+        * port.
+        */
+       for (type = 0; type < MLXSW_REG_SFGC_TYPE_MAX; type++) {
+               if (type == MLXSW_REG_SFGC_TYPE_RESERVED)
+                       continue;
+
+               err = __mlxsw_sp_flood_init(mlxsw_sp->core, type,
+                                           MLXSW_REG_SFGC_BRIDGE_TYPE_VFID);
+               if (err)
+                       return err;
+       }
+
+       /* For bridged ports, use one flooding table for unknown unicast
+        * traffic and a second table for unregistered multicast and
+        * broadcast.
+        */
+       for (type = 0; type < MLXSW_REG_SFGC_TYPE_MAX; type++) {
+               if (type == MLXSW_REG_SFGC_TYPE_RESERVED)
+                       continue;
+
+               err = __mlxsw_sp_flood_init(mlxsw_sp->core, type,
+                                           MLXSW_REG_SFGC_BRIDGE_TYPE_1Q_FID);
+               if (err)
+                       return err;
+       }
+
+       return 0;
+}
+
+static int mlxsw_sp_init(void *priv, struct mlxsw_core *mlxsw_core,
+                        const struct mlxsw_bus_info *mlxsw_bus_info)
+{
+       struct mlxsw_sp *mlxsw_sp = priv;
+       int err;
+
+       mlxsw_sp->core = mlxsw_core;
+       mlxsw_sp->bus_info = mlxsw_bus_info;
+
+       err = mlxsw_sp_base_mac_get(mlxsw_sp);
+       if (err) {
+               dev_err(mlxsw_sp->bus_info->dev, "Failed to get base mac\n");
+               return err;
+       }
+
+       err = mlxsw_sp_ports_create(mlxsw_sp);
+       if (err) {
+               dev_err(mlxsw_sp->bus_info->dev, "Failed to create ports\n");
+               goto err_ports_create;
+       }
+
+       err = mlxsw_sp_event_register(mlxsw_sp, MLXSW_TRAP_ID_PUDE);
+       if (err) {
+               dev_err(mlxsw_sp->bus_info->dev, "Failed to register for PUDE events\n");
+               goto err_event_register;
+       }
+
+       err = mlxsw_sp_traps_init(mlxsw_sp);
+       if (err) {
+               dev_err(mlxsw_sp->bus_info->dev, "Failed to set traps for RX\n");
+               goto err_rx_listener_register;
+       }
+
+       err = mlxsw_sp_flood_init(mlxsw_sp);
+       if (err) {
+               dev_err(mlxsw_sp->bus_info->dev, "Failed to initialize flood tables\n");
+               goto err_flood_init;
+       }
+
+       err = mlxsw_sp_buffers_init(mlxsw_sp);
+       if (err) {
+               dev_err(mlxsw_sp->bus_info->dev, "Failed to initialize buffers\n");
+               goto err_buffers_init;
+       }
+
+       err = mlxsw_sp_switchdev_init(mlxsw_sp);
+       if (err) {
+               dev_err(mlxsw_sp->bus_info->dev, "Failed to initialize switchdev\n");
+               goto err_switchdev_init;
+       }
+
+       return 0;
+
+err_switchdev_init:
+err_buffers_init:
+err_flood_init:
+       mlxsw_sp_traps_fini(mlxsw_sp);
+err_rx_listener_register:
+       mlxsw_sp_event_unregister(mlxsw_sp, MLXSW_TRAP_ID_PUDE);
+err_event_register:
+       mlxsw_sp_ports_remove(mlxsw_sp);
+err_ports_create:
+       mlxsw_sp_vfids_fini(mlxsw_sp);
+       return err;
+}
+
+static void mlxsw_sp_fini(void *priv)
+{
+       struct mlxsw_sp *mlxsw_sp = priv;
+
+       mlxsw_sp_switchdev_fini(mlxsw_sp);
+       mlxsw_sp_traps_fini(mlxsw_sp);
+       mlxsw_sp_event_unregister(mlxsw_sp, MLXSW_TRAP_ID_PUDE);
+       mlxsw_sp_ports_remove(mlxsw_sp);
+       mlxsw_sp_vfids_fini(mlxsw_sp);
+}
+
+static struct mlxsw_config_profile mlxsw_sp_config_profile = {
+       .used_max_vepa_channels         = 1,
+       .max_vepa_channels              = 0,
+       .used_max_lag                   = 1,
+       .max_lag                        = 64,
+       .used_max_port_per_lag          = 1,
+       .max_port_per_lag               = 16,
+       .used_max_mid                   = 1,
+       .max_mid                        = 7000,
+       .used_max_pgt                   = 1,
+       .max_pgt                        = 0,
+       .used_max_system_port           = 1,
+       .max_system_port                = 64,
+       .used_max_vlan_groups           = 1,
+       .max_vlan_groups                = 127,
+       .used_max_regions               = 1,
+       .max_regions                    = 400,
+       .used_flood_tables              = 1,
+       .used_flood_mode                = 1,
+       .flood_mode                     = 3,
+       .max_fid_offset_flood_tables    = 2,
+       .fid_offset_flood_table_size    = VLAN_N_VID - 1,
+       .max_fid_flood_tables           = 1,
+       .fid_flood_table_size           = VLAN_N_VID,
+       .used_max_ib_mc                 = 1,
+       .max_ib_mc                      = 0,
+       .used_max_pkey                  = 1,
+       .max_pkey                       = 0,
+       .swid_config                    = {
+               {
+                       .used_type      = 1,
+                       .type           = MLXSW_PORT_SWID_TYPE_ETH,
+               }
+       },
+};
+
+static struct mlxsw_driver mlxsw_sp_driver = {
+       .kind                   = MLXSW_DEVICE_KIND_SPECTRUM,
+       .owner                  = THIS_MODULE,
+       .priv_size              = sizeof(struct mlxsw_sp),
+       .init                   = mlxsw_sp_init,
+       .fini                   = mlxsw_sp_fini,
+       .txhdr_construct        = mlxsw_sp_txhdr_construct,
+       .txhdr_len              = MLXSW_TXHDR_LEN,
+       .profile                = &mlxsw_sp_config_profile,
+};
+
+static bool mlxsw_sp_port_dev_check(const struct net_device *dev)
+{
+       return dev->netdev_ops == &mlxsw_sp_port_netdev_ops;
+}
+
+static int mlxsw_sp_port_bridge_join(struct mlxsw_sp_port *mlxsw_sp_port)
+{
+       struct net_device *dev = mlxsw_sp_port->dev;
+       int err;
+
+       /* When port is not bridged untagged packets are tagged with
+        * PVID=VID=1, thereby creating an implicit VLAN interface in
+        * the device. Remove it and let bridge code take care of its
+        * own VLANs.
+        */
+       err = mlxsw_sp_port_kill_vid(dev, 0, 1);
+       if (err)
+               netdev_err(dev, "Failed to remove VID 1\n");
+
+       return err;
+}
+
+static int mlxsw_sp_port_bridge_leave(struct mlxsw_sp_port *mlxsw_sp_port)
+{
+       struct net_device *dev = mlxsw_sp_port->dev;
+       int err;
+
+       /* Add implicit VLAN interface in the device, so that untagged
+        * packets will be classified to the default vFID.
+        */
+       err = mlxsw_sp_port_add_vid(dev, 0, 1);
+       if (err)
+               netdev_err(dev, "Failed to add VID 1\n");
+
+       return err;
+}
+
+static bool mlxsw_sp_master_bridge_check(struct mlxsw_sp *mlxsw_sp,
+                                        struct net_device *br_dev)
+{
+       return !mlxsw_sp->master_bridge.dev ||
+              mlxsw_sp->master_bridge.dev == br_dev;
+}
+
+static void mlxsw_sp_master_bridge_inc(struct mlxsw_sp *mlxsw_sp,
+                                      struct net_device *br_dev)
+{
+       mlxsw_sp->master_bridge.dev = br_dev;
+       mlxsw_sp->master_bridge.ref_count++;
+}
+
+static void mlxsw_sp_master_bridge_dec(struct mlxsw_sp *mlxsw_sp,
+                                      struct net_device *br_dev)
+{
+       if (--mlxsw_sp->master_bridge.ref_count == 0)
+               mlxsw_sp->master_bridge.dev = NULL;
+}
+
+static int mlxsw_sp_netdevice_event(struct notifier_block *unused,
+                                   unsigned long event, void *ptr)
+{
+       struct net_device *dev = netdev_notifier_info_to_dev(ptr);
+       struct netdev_notifier_changeupper_info *info;
+       struct mlxsw_sp_port *mlxsw_sp_port;
+       struct net_device *upper_dev;
+       struct mlxsw_sp *mlxsw_sp;
+       int err;
+
+       if (!mlxsw_sp_port_dev_check(dev))
+               return NOTIFY_DONE;
+
+       mlxsw_sp_port = netdev_priv(dev);
+       mlxsw_sp = mlxsw_sp_port->mlxsw_sp;
+       info = ptr;
+
+       switch (event) {
+       case NETDEV_PRECHANGEUPPER:
+               upper_dev = info->upper_dev;
+               /* HW limitation forbids to put ports to multiple bridges. */
+               if (info->master && info->linking &&
+                   netif_is_bridge_master(upper_dev) &&
+                   !mlxsw_sp_master_bridge_check(mlxsw_sp, upper_dev))
+                       return NOTIFY_BAD;
+               break;
+       case NETDEV_CHANGEUPPER:
+               upper_dev = info->upper_dev;
+               if (info->master &&
+                   netif_is_bridge_master(upper_dev)) {
+                       if (info->linking) {
+                               err = mlxsw_sp_port_bridge_join(mlxsw_sp_port);
+                               if (err)
+                                       netdev_err(dev, "Failed to join bridge\n");
+                               mlxsw_sp_master_bridge_inc(mlxsw_sp, upper_dev);
+                               mlxsw_sp_port->bridged = true;
+                       } else {
+                               err = mlxsw_sp_port_bridge_leave(mlxsw_sp_port);
+                               if (err)
+                                       netdev_err(dev, "Failed to leave bridge\n");
+                               mlxsw_sp_port->bridged = false;
+                               mlxsw_sp_master_bridge_dec(mlxsw_sp, upper_dev);
+                       }
+               }
+               break;
+       }
+
+       return NOTIFY_DONE;
+}
+
+static struct notifier_block mlxsw_sp_netdevice_nb __read_mostly = {
+       .notifier_call = mlxsw_sp_netdevice_event,
+};
+
+static int __init mlxsw_sp_module_init(void)
+{
+       int err;
+
+       register_netdevice_notifier(&mlxsw_sp_netdevice_nb);
+       err = mlxsw_core_driver_register(&mlxsw_sp_driver);
+       if (err)
+               goto err_core_driver_register;
+       return 0;
+
+err_core_driver_register:
+       unregister_netdevice_notifier(&mlxsw_sp_netdevice_nb);
+       return err;
+}
+
+static void __exit mlxsw_sp_module_exit(void)
+{
+       mlxsw_core_driver_unregister(&mlxsw_sp_driver);
+       unregister_netdevice_notifier(&mlxsw_sp_netdevice_nb);
+}
+
+module_init(mlxsw_sp_module_init);
+module_exit(mlxsw_sp_module_exit);
+
+MODULE_LICENSE("Dual BSD/GPL");
+MODULE_AUTHOR("Jiri Pirko <jiri@mellanox.com>");
+MODULE_DESCRIPTION("Mellanox Spectrum driver");
+MODULE_MLXSW_DRIVER_ALIAS(MLXSW_DEVICE_KIND_SPECTRUM);
diff --git a/drivers/net/ethernet/mellanox/mlxsw/spectrum.h b/drivers/net/ethernet/mellanox/mlxsw/spectrum.h
new file mode 100644 (file)
index 0000000..fc00749
--- /dev/null
@@ -0,0 +1,121 @@
+/*
+ * drivers/net/ethernet/mellanox/mlxsw/spectrum.h
+ * Copyright (c) 2015 Mellanox Technologies. All rights reserved.
+ * Copyright (c) 2015 Jiri Pirko <jiri@mellanox.com>
+ * Copyright (c) 2015 Ido Schimmel <idosch@mellanox.com>
+ * Copyright (c) 2015 Elad Raz <eladr@mellanox.com>
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions are met:
+ *
+ * 1. Redistributions of source code must retain the above copyright
+ *    notice, this list of conditions and the following disclaimer.
+ * 2. 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.
+ * 3. Neither the names of the copyright holders nor the names of its
+ *    contributors may be used to endorse or promote products derived from
+ *    this software without specific prior written permission.
+ *
+ * Alternatively, this software may be distributed under the terms of the
+ * GNU General Public License ("GPL") version 2 as published by the Free
+ * Software Foundation.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+ * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
+ * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
+ * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
+ * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
+ * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
+ * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
+ * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
+ * POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#ifndef _MLXSW_SPECTRUM_H
+#define _MLXSW_SPECTRUM_H
+
+#include <linux/types.h>
+#include <linux/netdevice.h>
+#include <linux/bitops.h>
+#include <linux/if_vlan.h>
+#include <net/switchdev.h>
+
+#include "core.h"
+
+#define MLXSW_SP_VFID_BASE VLAN_N_VID
+
+struct mlxsw_sp_port;
+
+struct mlxsw_sp {
+       unsigned long active_vfids[BITS_TO_LONGS(VLAN_N_VID)];
+       unsigned long active_fids[BITS_TO_LONGS(VLAN_N_VID)];
+       struct mlxsw_sp_port **ports;
+       struct mlxsw_core *core;
+       const struct mlxsw_bus_info *bus_info;
+       unsigned char base_mac[ETH_ALEN];
+       struct {
+               struct delayed_work dw;
+#define MLXSW_SP_DEFAULT_LEARNING_INTERVAL 100
+               unsigned int interval; /* ms */
+       } fdb_notify;
+#define MLXSW_SP_DEFAULT_AGEING_TIME 300
+       u32 ageing_time;
+       struct {
+               struct net_device *dev;
+               unsigned int ref_count;
+       } master_bridge;
+};
+
+struct mlxsw_sp_port_pcpu_stats {
+       u64                     rx_packets;
+       u64                     rx_bytes;
+       u64                     tx_packets;
+       u64                     tx_bytes;
+       struct u64_stats_sync   syncp;
+       u32                     tx_dropped;
+};
+
+struct mlxsw_sp_port {
+       struct net_device *dev;
+       struct mlxsw_sp_port_pcpu_stats __percpu *pcpu_stats;
+       struct mlxsw_sp *mlxsw_sp;
+       u8 local_port;
+       u8 stp_state;
+       u8 learning:1;
+       u8 learning_sync:1;
+       u16 pvid;
+       bool bridged;
+       /* 802.1Q bridge VLANs */
+       unsigned long active_vlans[BITS_TO_LONGS(VLAN_N_VID)];
+       /* VLAN interfaces */
+       unsigned long active_vfids[BITS_TO_LONGS(VLAN_N_VID)];
+       u16 nr_vfids;
+};
+
+enum mlxsw_sp_flood_table {
+       MLXSW_SP_FLOOD_TABLE_UC,
+       MLXSW_SP_FLOOD_TABLE_BM,
+};
+
+int mlxsw_sp_buffers_init(struct mlxsw_sp *mlxsw_sp);
+int mlxsw_sp_port_buffers_init(struct mlxsw_sp_port *mlxsw_sp_port);
+
+int mlxsw_sp_switchdev_init(struct mlxsw_sp *mlxsw_sp);
+void mlxsw_sp_switchdev_fini(struct mlxsw_sp *mlxsw_sp);
+int mlxsw_sp_port_vlan_init(struct mlxsw_sp_port *mlxsw_sp_port);
+void mlxsw_sp_port_switchdev_init(struct mlxsw_sp_port *mlxsw_sp_port);
+void mlxsw_sp_port_switchdev_fini(struct mlxsw_sp_port *mlxsw_sp_port);
+int mlxsw_sp_port_vid_to_fid_set(struct mlxsw_sp_port *mlxsw_sp_port,
+                                enum mlxsw_reg_svfa_mt mt, bool valid, u16 fid,
+                                u16 vid);
+int mlxsw_sp_port_vlan_set(struct mlxsw_sp_port *mlxsw_sp_port, u16 vid_begin,
+                          u16 vid_end, bool is_member, bool untagged);
+int mlxsw_sp_port_add_vid(struct net_device *dev, __be16 __always_unused proto,
+                         u16 vid);
+int mlxsw_sp_port_kill_vid(struct net_device *dev,
+                          __be16 __always_unused proto, u16 vid);
+
+#endif
diff --git a/drivers/net/ethernet/mellanox/mlxsw/spectrum_buffers.c b/drivers/net/ethernet/mellanox/mlxsw/spectrum_buffers.c
new file mode 100644 (file)
index 0000000..d59195e
--- /dev/null
@@ -0,0 +1,422 @@
+/*
+ * drivers/net/ethernet/mellanox/mlxsw/spectrum_buffers.c
+ * Copyright (c) 2015 Mellanox Technologies. All rights reserved.
+ * Copyright (c) 2015 Jiri Pirko <jiri@mellanox.com>
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions are met:
+ *
+ * 1. Redistributions of source code must retain the above copyright
+ *    notice, this list of conditions and the following disclaimer.
+ * 2. 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.
+ * 3. Neither the names of the copyright holders nor the names of its
+ *    contributors may be used to endorse or promote products derived from
+ *    this software without specific prior written permission.
+ *
+ * Alternatively, this software may be distributed under the terms of the
+ * GNU General Public License ("GPL") version 2 as published by the Free
+ * Software Foundation.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+ * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
+ * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
+ * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
+ * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
+ * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
+ * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
+ * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
+ * POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#include <linux/kernel.h>
+#include <linux/types.h>
+
+#include "spectrum.h"
+#include "core.h"
+#include "port.h"
+#include "reg.h"
+
+struct mlxsw_sp_pb {
+       u8 index;
+       u16 size;
+};
+
+#define MLXSW_SP_PB(_index, _size)     \
+       {                               \
+               .index = _index,        \
+               .size = _size,          \
+       }
+
+static const struct mlxsw_sp_pb mlxsw_sp_pbs[] = {
+       MLXSW_SP_PB(0, 208),
+       MLXSW_SP_PB(1, 208),
+       MLXSW_SP_PB(2, 208),
+       MLXSW_SP_PB(3, 208),
+       MLXSW_SP_PB(4, 208),
+       MLXSW_SP_PB(5, 208),
+       MLXSW_SP_PB(6, 208),
+       MLXSW_SP_PB(7, 208),
+       MLXSW_SP_PB(9, 208),
+};
+
+#define MLXSW_SP_PBS_LEN ARRAY_SIZE(mlxsw_sp_pbs)
+
+static int mlxsw_sp_port_pb_init(struct mlxsw_sp_port *mlxsw_sp_port)
+{
+       char pbmc_pl[MLXSW_REG_PBMC_LEN];
+       int i;
+
+       mlxsw_reg_pbmc_pack(pbmc_pl, mlxsw_sp_port->local_port,
+                           0xffff, 0xffff / 2);
+       for (i = 0; i < MLXSW_SP_PBS_LEN; i++) {
+               const struct mlxsw_sp_pb *pb;
+
+               pb = &mlxsw_sp_pbs[i];
+               mlxsw_reg_pbmc_lossy_buffer_pack(pbmc_pl, pb->index, pb->size);
+       }
+       return mlxsw_reg_write(mlxsw_sp_port->mlxsw_sp->core,
+                              MLXSW_REG(pbmc), pbmc_pl);
+}
+
+#define MLXSW_SP_SB_BYTES_PER_CELL 96
+
+struct mlxsw_sp_sb_pool {
+       u8 pool;
+       enum mlxsw_reg_sbpr_dir dir;
+       enum mlxsw_reg_sbpr_mode mode;
+       u32 size;
+};
+
+#define MLXSW_SP_SB_POOL_INGRESS_SIZE                          \
+       ((15000000 - (2 * 20000 * MLXSW_PORT_MAX_PORTS)) /      \
+        MLXSW_SP_SB_BYTES_PER_CELL)
+#define MLXSW_SP_SB_POOL_EGRESS_SIZE                           \
+       ((14000000 - (8 * 1500 * MLXSW_PORT_MAX_PORTS)) /       \
+        MLXSW_SP_SB_BYTES_PER_CELL)
+
+#define MLXSW_SP_SB_POOL(_pool, _dir, _mode, _size)            \
+       {                                                       \
+               .pool = _pool,                                  \
+               .dir = _dir,                                    \
+               .mode = _mode,                                  \
+               .size = _size,                                  \
+       }
+
+#define MLXSW_SP_SB_POOL_INGRESS(_pool, _size)                 \
+       MLXSW_SP_SB_POOL(_pool, MLXSW_REG_SBPR_DIR_INGRESS,     \
+                        MLXSW_REG_SBPR_MODE_DYNAMIC, _size)
+
+#define MLXSW_SP_SB_POOL_EGRESS(_pool, _size)                  \
+       MLXSW_SP_SB_POOL(_pool, MLXSW_REG_SBPR_DIR_EGRESS,      \
+                        MLXSW_REG_SBPR_MODE_DYNAMIC, _size)
+
+static const struct mlxsw_sp_sb_pool mlxsw_sp_sb_pools[] = {
+       MLXSW_SP_SB_POOL_INGRESS(0, MLXSW_SP_SB_POOL_INGRESS_SIZE),
+       MLXSW_SP_SB_POOL_INGRESS(1, 0),
+       MLXSW_SP_SB_POOL_INGRESS(2, 0),
+       MLXSW_SP_SB_POOL_INGRESS(3, 0),
+       MLXSW_SP_SB_POOL_EGRESS(0, MLXSW_SP_SB_POOL_EGRESS_SIZE),
+       MLXSW_SP_SB_POOL_EGRESS(1, 0),
+       MLXSW_SP_SB_POOL_EGRESS(2, 0),
+       MLXSW_SP_SB_POOL_EGRESS(2, MLXSW_SP_SB_POOL_EGRESS_SIZE),
+};
+
+#define MLXSW_SP_SB_POOLS_LEN ARRAY_SIZE(mlxsw_sp_sb_pools)
+
+static int mlxsw_sp_sb_pools_init(struct mlxsw_sp *mlxsw_sp)
+{
+       char sbpr_pl[MLXSW_REG_SBPR_LEN];
+       int i;
+       int err;
+
+       for (i = 0; i < MLXSW_SP_SB_POOLS_LEN; i++) {
+               const struct mlxsw_sp_sb_pool *pool;
+
+               pool = &mlxsw_sp_sb_pools[i];
+               mlxsw_reg_sbpr_pack(sbpr_pl, pool->pool, pool->dir,
+                                   pool->mode, pool->size);
+               err = mlxsw_reg_write(mlxsw_sp->core, MLXSW_REG(sbpr), sbpr_pl);
+               if (err)
+                       return err;
+       }
+       return 0;
+}
+
+struct mlxsw_sp_sb_cm {
+       union {
+               u8 pg;
+               u8 tc;
+       } u;
+       enum mlxsw_reg_sbcm_dir dir;
+       u32 min_buff;
+       u32 max_buff;
+       u8 pool;
+};
+
+#define MLXSW_SP_SB_CM(_pg_tc, _dir, _min_buff, _max_buff, _pool)      \
+       {                                                               \
+               .u.pg = _pg_tc,                                         \
+               .dir = _dir,                                            \
+               .min_buff = _min_buff,                                  \
+               .max_buff = _max_buff,                                  \
+               .pool = _pool,                                          \
+       }
+
+#define MLXSW_SP_SB_CM_INGRESS(_pg, _min_buff, _max_buff)              \
+       MLXSW_SP_SB_CM(_pg, MLXSW_REG_SBCM_DIR_INGRESS,                 \
+                      _min_buff, _max_buff, 0)
+
+#define MLXSW_SP_SB_CM_EGRESS(_tc, _min_buff, _max_buff)               \
+       MLXSW_SP_SB_CM(_tc, MLXSW_REG_SBCM_DIR_EGRESS,                  \
+                      _min_buff, _max_buff, 0)
+
+#define MLXSW_SP_CPU_PORT_SB_CM_EGRESS(_tc)                            \
+       MLXSW_SP_SB_CM(_tc, MLXSW_REG_SBCM_DIR_EGRESS, 104, 2, 3)
+
+static const struct mlxsw_sp_sb_cm mlxsw_sp_sb_cms[] = {
+       MLXSW_SP_SB_CM_INGRESS(0, 10000 / MLXSW_SP_SB_BYTES_PER_CELL, 8),
+       MLXSW_SP_SB_CM_INGRESS(1, 0, 0),
+       MLXSW_SP_SB_CM_INGRESS(2, 0, 0),
+       MLXSW_SP_SB_CM_INGRESS(3, 0, 0),
+       MLXSW_SP_SB_CM_INGRESS(4, 0, 0),
+       MLXSW_SP_SB_CM_INGRESS(5, 0, 0),
+       MLXSW_SP_SB_CM_INGRESS(6, 0, 0),
+       MLXSW_SP_SB_CM_INGRESS(7, 0, 0),
+       MLXSW_SP_SB_CM_INGRESS(9, 20000 / MLXSW_SP_SB_BYTES_PER_CELL, 0xff),
+       MLXSW_SP_SB_CM_EGRESS(0, 1500 / MLXSW_SP_SB_BYTES_PER_CELL, 9),
+       MLXSW_SP_SB_CM_EGRESS(1, 1500 / MLXSW_SP_SB_BYTES_PER_CELL, 9),
+       MLXSW_SP_SB_CM_EGRESS(2, 1500 / MLXSW_SP_SB_BYTES_PER_CELL, 9),
+       MLXSW_SP_SB_CM_EGRESS(3, 1500 / MLXSW_SP_SB_BYTES_PER_CELL, 9),
+       MLXSW_SP_SB_CM_EGRESS(4, 1500 / MLXSW_SP_SB_BYTES_PER_CELL, 9),
+       MLXSW_SP_SB_CM_EGRESS(5, 1500 / MLXSW_SP_SB_BYTES_PER_CELL, 9),
+       MLXSW_SP_SB_CM_EGRESS(6, 1500 / MLXSW_SP_SB_BYTES_PER_CELL, 9),
+       MLXSW_SP_SB_CM_EGRESS(7, 1500 / MLXSW_SP_SB_BYTES_PER_CELL, 9),
+       MLXSW_SP_SB_CM_EGRESS(8, 0, 0),
+       MLXSW_SP_SB_CM_EGRESS(9, 0, 0),
+       MLXSW_SP_SB_CM_EGRESS(10, 0, 0),
+       MLXSW_SP_SB_CM_EGRESS(11, 0, 0),
+       MLXSW_SP_SB_CM_EGRESS(12, 0, 0),
+       MLXSW_SP_SB_CM_EGRESS(13, 0, 0),
+       MLXSW_SP_SB_CM_EGRESS(14, 0, 0),
+       MLXSW_SP_SB_CM_EGRESS(15, 0, 0),
+       MLXSW_SP_SB_CM_EGRESS(16, 1, 0xff),
+};
+
+#define MLXSW_SP_SB_CMS_LEN ARRAY_SIZE(mlxsw_sp_sb_cms)
+
+static const struct mlxsw_sp_sb_cm mlxsw_sp_cpu_port_sb_cms[] = {
+       MLXSW_SP_CPU_PORT_SB_CM_EGRESS(0),
+       MLXSW_SP_CPU_PORT_SB_CM_EGRESS(1),
+       MLXSW_SP_CPU_PORT_SB_CM_EGRESS(2),
+       MLXSW_SP_CPU_PORT_SB_CM_EGRESS(3),
+       MLXSW_SP_CPU_PORT_SB_CM_EGRESS(4),
+       MLXSW_SP_CPU_PORT_SB_CM_EGRESS(5),
+       MLXSW_SP_CPU_PORT_SB_CM_EGRESS(6),
+       MLXSW_SP_CPU_PORT_SB_CM_EGRESS(7),
+       MLXSW_SP_CPU_PORT_SB_CM_EGRESS(8),
+       MLXSW_SP_CPU_PORT_SB_CM_EGRESS(9),
+       MLXSW_SP_CPU_PORT_SB_CM_EGRESS(10),
+       MLXSW_SP_CPU_PORT_SB_CM_EGRESS(11),
+       MLXSW_SP_CPU_PORT_SB_CM_EGRESS(12),
+       MLXSW_SP_CPU_PORT_SB_CM_EGRESS(13),
+       MLXSW_SP_CPU_PORT_SB_CM_EGRESS(14),
+       MLXSW_SP_CPU_PORT_SB_CM_EGRESS(15),
+       MLXSW_SP_CPU_PORT_SB_CM_EGRESS(16),
+       MLXSW_SP_CPU_PORT_SB_CM_EGRESS(17),
+       MLXSW_SP_CPU_PORT_SB_CM_EGRESS(18),
+       MLXSW_SP_CPU_PORT_SB_CM_EGRESS(19),
+       MLXSW_SP_CPU_PORT_SB_CM_EGRESS(20),
+       MLXSW_SP_CPU_PORT_SB_CM_EGRESS(21),
+       MLXSW_SP_CPU_PORT_SB_CM_EGRESS(22),
+       MLXSW_SP_CPU_PORT_SB_CM_EGRESS(23),
+       MLXSW_SP_CPU_PORT_SB_CM_EGRESS(24),
+       MLXSW_SP_CPU_PORT_SB_CM_EGRESS(25),
+       MLXSW_SP_CPU_PORT_SB_CM_EGRESS(26),
+       MLXSW_SP_CPU_PORT_SB_CM_EGRESS(27),
+       MLXSW_SP_CPU_PORT_SB_CM_EGRESS(28),
+       MLXSW_SP_CPU_PORT_SB_CM_EGRESS(29),
+       MLXSW_SP_CPU_PORT_SB_CM_EGRESS(30),
+       MLXSW_SP_CPU_PORT_SB_CM_EGRESS(31),
+};
+
+#define MLXSW_SP_CPU_PORT_SB_MCS_LEN \
+       ARRAY_SIZE(mlxsw_sp_cpu_port_sb_cms)
+
+static int mlxsw_sp_sb_cms_init(struct mlxsw_sp *mlxsw_sp, u8 local_port,
+                               const struct mlxsw_sp_sb_cm *cms,
+                               size_t cms_len)
+{
+       char sbcm_pl[MLXSW_REG_SBCM_LEN];
+       int i;
+       int err;
+
+       for (i = 0; i < cms_len; i++) {
+               const struct mlxsw_sp_sb_cm *cm;
+
+               cm = &cms[i];
+               mlxsw_reg_sbcm_pack(sbcm_pl, local_port, cm->u.pg, cm->dir,
+                                   cm->min_buff, cm->max_buff, cm->pool);
+               err = mlxsw_reg_write(mlxsw_sp->core, MLXSW_REG(sbcm), sbcm_pl);
+               if (err)
+                       return err;
+       }
+       return 0;
+}
+
+static int mlxsw_sp_port_sb_cms_init(struct mlxsw_sp_port *mlxsw_sp_port)
+{
+       return mlxsw_sp_sb_cms_init(mlxsw_sp_port->mlxsw_sp,
+                                   mlxsw_sp_port->local_port, mlxsw_sp_sb_cms,
+                                   MLXSW_SP_SB_CMS_LEN);
+}
+
+static int mlxsw_sp_cpu_port_sb_cms_init(struct mlxsw_sp *mlxsw_sp)
+{
+       return mlxsw_sp_sb_cms_init(mlxsw_sp, 0, mlxsw_sp_cpu_port_sb_cms,
+                                   MLXSW_SP_CPU_PORT_SB_MCS_LEN);
+}
+
+struct mlxsw_sp_sb_pm {
+       u8 pool;
+       enum mlxsw_reg_sbpm_dir dir;
+       u32 min_buff;
+       u32 max_buff;
+};
+
+#define MLXSW_SP_SB_PM(_pool, _dir, _min_buff, _max_buff)      \
+       {                                                       \
+               .pool = _pool,                                  \
+               .dir = _dir,                                    \
+               .min_buff = _min_buff,                          \
+               .max_buff = _max_buff,                          \
+       }
+
+#define MLXSW_SP_SB_PM_INGRESS(_pool, _min_buff, _max_buff)    \
+       MLXSW_SP_SB_PM(_pool, MLXSW_REG_SBPM_DIR_INGRESS,       \
+                      _min_buff, _max_buff)
+
+#define MLXSW_SP_SB_PM_EGRESS(_pool, _min_buff, _max_buff)     \
+       MLXSW_SP_SB_PM(_pool, MLXSW_REG_SBPM_DIR_EGRESS,        \
+                      _min_buff, _max_buff)
+
+static const struct mlxsw_sp_sb_pm mlxsw_sp_sb_pms[] = {
+       MLXSW_SP_SB_PM_INGRESS(0, 0, 0xff),
+       MLXSW_SP_SB_PM_INGRESS(1, 0, 0),
+       MLXSW_SP_SB_PM_INGRESS(2, 0, 0),
+       MLXSW_SP_SB_PM_INGRESS(3, 0, 0),
+       MLXSW_SP_SB_PM_EGRESS(0, 0, 7),
+       MLXSW_SP_SB_PM_EGRESS(1, 0, 0),
+       MLXSW_SP_SB_PM_EGRESS(2, 0, 0),
+       MLXSW_SP_SB_PM_EGRESS(3, 0, 0),
+};
+
+#define MLXSW_SP_SB_PMS_LEN ARRAY_SIZE(mlxsw_sp_sb_pms)
+
+static int mlxsw_sp_port_sb_pms_init(struct mlxsw_sp_port *mlxsw_sp_port)
+{
+       char sbpm_pl[MLXSW_REG_SBPM_LEN];
+       int i;
+       int err;
+
+       for (i = 0; i < MLXSW_SP_SB_PMS_LEN; i++) {
+               const struct mlxsw_sp_sb_pm *pm;
+
+               pm = &mlxsw_sp_sb_pms[i];
+               mlxsw_reg_sbpm_pack(sbpm_pl, mlxsw_sp_port->local_port,
+                                   pm->pool, pm->dir,
+                                   pm->min_buff, pm->max_buff);
+               err = mlxsw_reg_write(mlxsw_sp_port->mlxsw_sp->core,
+                                     MLXSW_REG(sbpm), sbpm_pl);
+               if (err)
+                       return err;
+       }
+       return 0;
+}
+
+struct mlxsw_sp_sb_mm {
+       u8 prio;
+       u32 min_buff;
+       u32 max_buff;
+       u8 pool;
+};
+
+#define MLXSW_SP_SB_MM(_prio, _min_buff, _max_buff, _pool)     \
+       {                                                       \
+               .prio = _prio,                                  \
+               .min_buff = _min_buff,                          \
+               .max_buff = _max_buff,                          \
+               .pool = _pool,                                  \
+       }
+
+static const struct mlxsw_sp_sb_mm mlxsw_sp_sb_mms[] = {
+       MLXSW_SP_SB_MM(0, 20000 / MLXSW_SP_SB_BYTES_PER_CELL, 0xff, 0),
+       MLXSW_SP_SB_MM(1, 20000 / MLXSW_SP_SB_BYTES_PER_CELL, 0xff, 0),
+       MLXSW_SP_SB_MM(2, 20000 / MLXSW_SP_SB_BYTES_PER_CELL, 0xff, 0),
+       MLXSW_SP_SB_MM(3, 20000 / MLXSW_SP_SB_BYTES_PER_CELL, 0xff, 0),
+       MLXSW_SP_SB_MM(4, 20000 / MLXSW_SP_SB_BYTES_PER_CELL, 0xff, 0),
+       MLXSW_SP_SB_MM(5, 20000 / MLXSW_SP_SB_BYTES_PER_CELL, 0xff, 0),
+       MLXSW_SP_SB_MM(6, 20000 / MLXSW_SP_SB_BYTES_PER_CELL, 0xff, 0),
+       MLXSW_SP_SB_MM(7, 20000 / MLXSW_SP_SB_BYTES_PER_CELL, 0xff, 0),
+       MLXSW_SP_SB_MM(8, 20000 / MLXSW_SP_SB_BYTES_PER_CELL, 0xff, 0),
+       MLXSW_SP_SB_MM(9, 20000 / MLXSW_SP_SB_BYTES_PER_CELL, 0xff, 0),
+       MLXSW_SP_SB_MM(10, 20000 / MLXSW_SP_SB_BYTES_PER_CELL, 0xff, 0),
+       MLXSW_SP_SB_MM(11, 20000 / MLXSW_SP_SB_BYTES_PER_CELL, 0xff, 0),
+       MLXSW_SP_SB_MM(12, 20000 / MLXSW_SP_SB_BYTES_PER_CELL, 0xff, 0),
+       MLXSW_SP_SB_MM(13, 20000 / MLXSW_SP_SB_BYTES_PER_CELL, 0xff, 0),
+       MLXSW_SP_SB_MM(14, 20000 / MLXSW_SP_SB_BYTES_PER_CELL, 0xff, 0),
+};
+
+#define MLXSW_SP_SB_MMS_LEN ARRAY_SIZE(mlxsw_sp_sb_mms)
+
+static int mlxsw_sp_sb_mms_init(struct mlxsw_sp *mlxsw_sp)
+{
+       char sbmm_pl[MLXSW_REG_SBMM_LEN];
+       int i;
+       int err;
+
+       for (i = 0; i < MLXSW_SP_SB_MMS_LEN; i++) {
+               const struct mlxsw_sp_sb_mm *mc;
+
+               mc = &mlxsw_sp_sb_mms[i];
+               mlxsw_reg_sbmm_pack(sbmm_pl, mc->prio, mc->min_buff,
+                                   mc->max_buff, mc->pool);
+               err = mlxsw_reg_write(mlxsw_sp->core, MLXSW_REG(sbmm), sbmm_pl);
+               if (err)
+                       return err;
+       }
+       return 0;
+}
+
+int mlxsw_sp_buffers_init(struct mlxsw_sp *mlxsw_sp)
+{
+       int err;
+
+       err = mlxsw_sp_sb_pools_init(mlxsw_sp);
+       if (err)
+               return err;
+       err = mlxsw_sp_cpu_port_sb_cms_init(mlxsw_sp);
+       if (err)
+               return err;
+       err = mlxsw_sp_sb_mms_init(mlxsw_sp);
+
+       return err;
+}
+
+int mlxsw_sp_port_buffers_init(struct mlxsw_sp_port *mlxsw_sp_port)
+{
+       int err;
+
+       err = mlxsw_sp_port_pb_init(mlxsw_sp_port);
+       if (err)
+               return err;
+       err = mlxsw_sp_port_sb_cms_init(mlxsw_sp_port);
+       if (err)
+               return err;
+       err = mlxsw_sp_port_sb_pms_init(mlxsw_sp_port);
+
+       return err;
+}
diff --git a/drivers/net/ethernet/mellanox/mlxsw/spectrum_switchdev.c b/drivers/net/ethernet/mellanox/mlxsw/spectrum_switchdev.c
new file mode 100644 (file)
index 0000000..c39b7a1
--- /dev/null
@@ -0,0 +1,863 @@
+/*
+ * drivers/net/ethernet/mellanox/mlxsw/spectrum_switchdev.c
+ * Copyright (c) 2015 Mellanox Technologies. All rights reserved.
+ * Copyright (c) 2015 Jiri Pirko <jiri@mellanox.com>
+ * Copyright (c) 2015 Ido Schimmel <idosch@mellanox.com>
+ * Copyright (c) 2015 Elad Raz <eladr@mellanox.com>
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions are met:
+ *
+ * 1. Redistributions of source code must retain the above copyright
+ *    notice, this list of conditions and the following disclaimer.
+ * 2. 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.
+ * 3. Neither the names of the copyright holders nor the names of its
+ *    contributors may be used to endorse or promote products derived from
+ *    this software without specific prior written permission.
+ *
+ * Alternatively, this software may be distributed under the terms of the
+ * GNU General Public License ("GPL") version 2 as published by the Free
+ * Software Foundation.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+ * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
+ * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
+ * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
+ * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
+ * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
+ * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
+ * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
+ * POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#include <linux/kernel.h>
+#include <linux/types.h>
+#include <linux/netdevice.h>
+#include <linux/etherdevice.h>
+#include <linux/slab.h>
+#include <linux/device.h>
+#include <linux/skbuff.h>
+#include <linux/if_vlan.h>
+#include <linux/if_bridge.h>
+#include <linux/workqueue.h>
+#include <linux/jiffies.h>
+#include <net/switchdev.h>
+
+#include "spectrum.h"
+#include "core.h"
+#include "reg.h"
+
+static int mlxsw_sp_port_attr_get(struct net_device *dev,
+                                 struct switchdev_attr *attr)
+{
+       struct mlxsw_sp_port *mlxsw_sp_port = netdev_priv(dev);
+       struct mlxsw_sp *mlxsw_sp = mlxsw_sp_port->mlxsw_sp;
+
+       switch (attr->id) {
+       case SWITCHDEV_ATTR_ID_PORT_PARENT_ID:
+               attr->u.ppid.id_len = sizeof(mlxsw_sp->base_mac);
+               memcpy(&attr->u.ppid.id, &mlxsw_sp->base_mac,
+                      attr->u.ppid.id_len);
+               break;
+       case SWITCHDEV_ATTR_ID_PORT_BRIDGE_FLAGS:
+               attr->u.brport_flags =
+                       (mlxsw_sp_port->learning ? BR_LEARNING : 0) |
+                       (mlxsw_sp_port->learning_sync ? BR_LEARNING_SYNC : 0);
+               break;
+       default:
+               return -EOPNOTSUPP;
+       }
+
+       return 0;
+}
+
+static int mlxsw_sp_port_stp_state_set(struct mlxsw_sp_port *mlxsw_sp_port,
+                                      u8 state)
+{
+       struct mlxsw_sp *mlxsw_sp = mlxsw_sp_port->mlxsw_sp;
+       enum mlxsw_reg_spms_state spms_state;
+       char *spms_pl;
+       u16 vid;
+       int err;
+
+       switch (state) {
+       case BR_STATE_DISABLED: /* fall-through */
+       case BR_STATE_FORWARDING:
+               spms_state = MLXSW_REG_SPMS_STATE_FORWARDING;
+               break;
+       case BR_STATE_LISTENING: /* fall-through */
+       case BR_STATE_LEARNING:
+               spms_state = MLXSW_REG_SPMS_STATE_LEARNING;
+               break;
+       case BR_STATE_BLOCKING:
+               spms_state = MLXSW_REG_SPMS_STATE_DISCARDING;
+               break;
+       default:
+               BUG();
+       }
+
+       spms_pl = kmalloc(MLXSW_REG_SPMS_LEN, GFP_KERNEL);
+       if (!spms_pl)
+               return -ENOMEM;
+       mlxsw_reg_spms_pack(spms_pl, mlxsw_sp_port->local_port);
+       for_each_set_bit(vid, mlxsw_sp_port->active_vlans, VLAN_N_VID)
+               mlxsw_reg_spms_vid_pack(spms_pl, vid, spms_state);
+
+       err = mlxsw_reg_write(mlxsw_sp->core, MLXSW_REG(spms), spms_pl);
+       kfree(spms_pl);
+       return err;
+}
+
+static int mlxsw_sp_port_attr_stp_state_set(struct mlxsw_sp_port *mlxsw_sp_port,
+                                           struct switchdev_trans *trans,
+                                           u8 state)
+{
+       if (switchdev_trans_ph_prepare(trans))
+               return 0;
+
+       mlxsw_sp_port->stp_state = state;
+       return mlxsw_sp_port_stp_state_set(mlxsw_sp_port, state);
+}
+
+static int mlxsw_sp_port_attr_br_flags_set(struct mlxsw_sp_port *mlxsw_sp_port,
+                                          struct switchdev_trans *trans,
+                                          unsigned long brport_flags)
+{
+       if (switchdev_trans_ph_prepare(trans))
+               return 0;
+
+       mlxsw_sp_port->learning = brport_flags & BR_LEARNING ? 1 : 0;
+       mlxsw_sp_port->learning_sync = brport_flags & BR_LEARNING_SYNC ? 1 : 0;
+       return 0;
+}
+
+static int mlxsw_sp_ageing_set(struct mlxsw_sp *mlxsw_sp, u32 ageing_time)
+{
+       char sfdat_pl[MLXSW_REG_SFDAT_LEN];
+       int err;
+
+       mlxsw_reg_sfdat_pack(sfdat_pl, ageing_time);
+       err = mlxsw_reg_write(mlxsw_sp->core, MLXSW_REG(sfdat), sfdat_pl);
+       if (err)
+               return err;
+       mlxsw_sp->ageing_time = ageing_time;
+       return 0;
+}
+
+static int mlxsw_sp_port_attr_br_ageing_set(struct mlxsw_sp_port *mlxsw_sp_port,
+                                           struct switchdev_trans *trans,
+                                           unsigned long ageing_jiffies)
+{
+       struct mlxsw_sp *mlxsw_sp = mlxsw_sp_port->mlxsw_sp;
+       u32 ageing_time = jiffies_to_msecs(ageing_jiffies) / 1000;
+
+       if (switchdev_trans_ph_prepare(trans))
+               return 0;
+
+       return mlxsw_sp_ageing_set(mlxsw_sp, ageing_time);
+}
+
+static int mlxsw_sp_port_attr_set(struct net_device *dev,
+                                 const struct switchdev_attr *attr,
+                                 struct switchdev_trans *trans)
+{
+       struct mlxsw_sp_port *mlxsw_sp_port = netdev_priv(dev);
+       int err = 0;
+
+       switch (attr->id) {
+       case SWITCHDEV_ATTR_ID_PORT_STP_STATE:
+               err = mlxsw_sp_port_attr_stp_state_set(mlxsw_sp_port, trans,
+                                                      attr->u.stp_state);
+               break;
+       case SWITCHDEV_ATTR_ID_PORT_BRIDGE_FLAGS:
+               err = mlxsw_sp_port_attr_br_flags_set(mlxsw_sp_port, trans,
+                                                     attr->u.brport_flags);
+               break;
+       case SWITCHDEV_ATTR_ID_BRIDGE_AGEING_TIME:
+               err = mlxsw_sp_port_attr_br_ageing_set(mlxsw_sp_port, trans,
+                                                      attr->u.ageing_time);
+               break;
+       default:
+               err = -EOPNOTSUPP;
+               break;
+       }
+
+       return err;
+}
+
+static int mlxsw_sp_port_pvid_set(struct mlxsw_sp_port *mlxsw_sp_port, u16 vid)
+{
+       struct mlxsw_sp *mlxsw_sp = mlxsw_sp_port->mlxsw_sp;
+       char spvid_pl[MLXSW_REG_SPVID_LEN];
+
+       mlxsw_reg_spvid_pack(spvid_pl, mlxsw_sp_port->local_port, vid);
+       return mlxsw_reg_write(mlxsw_sp->core, MLXSW_REG(spvid), spvid_pl);
+}
+
+static int mlxsw_sp_fid_create(struct mlxsw_sp *mlxsw_sp, u16 fid)
+{
+       char sfmr_pl[MLXSW_REG_SFMR_LEN];
+       int err;
+
+       mlxsw_reg_sfmr_pack(sfmr_pl, MLXSW_REG_SFMR_OP_CREATE_FID, fid, fid);
+       err = mlxsw_reg_write(mlxsw_sp->core, MLXSW_REG(sfmr), sfmr_pl);
+
+       if (err)
+               return err;
+
+       set_bit(fid, mlxsw_sp->active_fids);
+       return 0;
+}
+
+static void mlxsw_sp_fid_destroy(struct mlxsw_sp *mlxsw_sp, u16 fid)
+{
+       char sfmr_pl[MLXSW_REG_SFMR_LEN];
+
+       clear_bit(fid, mlxsw_sp->active_fids);
+
+       mlxsw_reg_sfmr_pack(sfmr_pl, MLXSW_REG_SFMR_OP_DESTROY_FID,
+                           fid, fid);
+       mlxsw_reg_write(mlxsw_sp->core, MLXSW_REG(sfmr), sfmr_pl);
+}
+
+static int mlxsw_sp_port_fid_map(struct mlxsw_sp_port *mlxsw_sp_port, u16 fid)
+{
+       enum mlxsw_reg_svfa_mt mt;
+
+       if (mlxsw_sp_port->nr_vfids)
+               mt = MLXSW_REG_SVFA_MT_PORT_VID_TO_FID;
+       else
+               mt = MLXSW_REG_SVFA_MT_VID_TO_FID;
+
+       return mlxsw_sp_port_vid_to_fid_set(mlxsw_sp_port, mt, true, fid, fid);
+}
+
+static int mlxsw_sp_port_fid_unmap(struct mlxsw_sp_port *mlxsw_sp_port, u16 fid)
+{
+       enum mlxsw_reg_svfa_mt mt;
+
+       if (!mlxsw_sp_port->nr_vfids)
+               return 0;
+
+       mt = MLXSW_REG_SVFA_MT_PORT_VID_TO_FID;
+       return mlxsw_sp_port_vid_to_fid_set(mlxsw_sp_port, mt, false, fid, fid);
+}
+
+static int __mlxsw_sp_port_flood_set(struct mlxsw_sp_port *mlxsw_sp_port,
+                                    u16 fid, bool set, bool only_uc)
+{
+       struct mlxsw_sp *mlxsw_sp = mlxsw_sp_port->mlxsw_sp;
+       char *sftr_pl;
+       int err;
+
+       sftr_pl = kmalloc(MLXSW_REG_SFTR_LEN, GFP_KERNEL);
+       if (!sftr_pl)
+               return -ENOMEM;
+
+       mlxsw_reg_sftr_pack(sftr_pl, MLXSW_SP_FLOOD_TABLE_UC, fid,
+                           MLXSW_REG_SFGC_TABLE_TYPE_FID_OFFEST, 0,
+                           mlxsw_sp_port->local_port, set);
+       err = mlxsw_reg_write(mlxsw_sp->core, MLXSW_REG(sftr), sftr_pl);
+       if (err)
+               goto buffer_out;
+
+       /* Flooding control allows one to decide whether a given port will
+        * flood unicast traffic for which there is no FDB entry.
+        */
+       if (only_uc)
+               goto buffer_out;
+
+       mlxsw_reg_sftr_pack(sftr_pl, MLXSW_SP_FLOOD_TABLE_BM, fid,
+                           MLXSW_REG_SFGC_TABLE_TYPE_FID_OFFEST, 0,
+                           mlxsw_sp_port->local_port, set);
+       err = mlxsw_reg_write(mlxsw_sp->core, MLXSW_REG(sftr), sftr_pl);
+
+buffer_out:
+       kfree(sftr_pl);
+       return err;
+}
+
+static int mlxsw_sp_port_add_vids(struct net_device *dev, u16 vid_begin,
+                                 u16 vid_end)
+{
+       u16 vid;
+       int err;
+
+       for (vid = vid_begin; vid <= vid_end; vid++) {
+               err = mlxsw_sp_port_add_vid(dev, 0, vid);
+               if (err)
+                       goto err_port_add_vid;
+       }
+       return 0;
+
+err_port_add_vid:
+       for (vid--; vid >= vid_begin; vid--)
+               mlxsw_sp_port_kill_vid(dev, 0, vid);
+       return err;
+}
+
+static int __mlxsw_sp_port_vlans_add(struct mlxsw_sp_port *mlxsw_sp_port,
+                                    u16 vid_begin, u16 vid_end,
+                                    bool flag_untagged, bool flag_pvid)
+{
+       struct mlxsw_sp *mlxsw_sp = mlxsw_sp_port->mlxsw_sp;
+       struct net_device *dev = mlxsw_sp_port->dev;
+       enum mlxsw_reg_svfa_mt mt;
+       u16 vid, vid_e;
+       int err;
+
+       /* In case this is invoked with BRIDGE_FLAGS_SELF and port is
+        * not bridged, then packets ingressing through the port with
+        * the specified VIDs will be directed to CPU.
+        */
+       if (!mlxsw_sp_port->bridged)
+               return mlxsw_sp_port_add_vids(dev, vid_begin, vid_end);
+
+       for (vid = vid_begin; vid <= vid_end; vid++) {
+               if (!test_bit(vid, mlxsw_sp->active_fids)) {
+                       err = mlxsw_sp_fid_create(mlxsw_sp, vid);
+                       if (err) {
+                               netdev_err(dev, "Failed to create FID=%d\n",
+                                          vid);
+                               return err;
+                       }
+
+                       /* When creating a FID, we set a VID to FID mapping
+                        * regardless of the port's mode.
+                        */
+                       mt = MLXSW_REG_SVFA_MT_VID_TO_FID;
+                       err = mlxsw_sp_port_vid_to_fid_set(mlxsw_sp_port, mt,
+                                                          true, vid, vid);
+                       if (err) {
+                               netdev_err(dev, "Failed to create FID=VID=%d mapping\n",
+                                          vid);
+                               return err;
+                       }
+               }
+
+               /* Set FID mapping according to port's mode */
+               err = mlxsw_sp_port_fid_map(mlxsw_sp_port, vid);
+               if (err) {
+                       netdev_err(dev, "Failed to map FID=%d", vid);
+                       return err;
+               }
+
+               err = __mlxsw_sp_port_flood_set(mlxsw_sp_port, vid, true,
+                                               false);
+               if (err) {
+                       netdev_err(dev, "Failed to set flooding for FID=%d",
+                                  vid);
+                       return err;
+               }
+       }
+
+       for (vid = vid_begin; vid <= vid_end;
+            vid += MLXSW_REG_SPVM_REC_MAX_COUNT) {
+               vid_e = min((u16) (vid + MLXSW_REG_SPVM_REC_MAX_COUNT - 1),
+                           vid_end);
+
+               err = mlxsw_sp_port_vlan_set(mlxsw_sp_port, vid, vid_e, true,
+                                            flag_untagged);
+               if (err) {
+                       netdev_err(mlxsw_sp_port->dev, "Unable to add VIDs %d-%d\n",
+                                  vid, vid_e);
+                       return err;
+               }
+       }
+
+       vid = vid_begin;
+       if (flag_pvid && mlxsw_sp_port->pvid != vid) {
+               err = mlxsw_sp_port_pvid_set(mlxsw_sp_port, vid);
+               if (err) {
+                       netdev_err(mlxsw_sp_port->dev, "Unable to add PVID %d\n",
+                                  vid);
+                       return err;
+               }
+               mlxsw_sp_port->pvid = vid;
+       }
+
+       /* Changing activity bits only if HW operation succeded */
+       for (vid = vid_begin; vid <= vid_end; vid++)
+               set_bit(vid, mlxsw_sp_port->active_vlans);
+
+       return mlxsw_sp_port_stp_state_set(mlxsw_sp_port,
+                                          mlxsw_sp_port->stp_state);
+}
+
+static int mlxsw_sp_port_vlans_add(struct mlxsw_sp_port *mlxsw_sp_port,
+                                  const struct switchdev_obj_port_vlan *vlan,
+                                  struct switchdev_trans *trans)
+{
+       bool untagged_flag = vlan->flags & BRIDGE_VLAN_INFO_UNTAGGED;
+       bool pvid_flag = vlan->flags & BRIDGE_VLAN_INFO_PVID;
+
+       if (switchdev_trans_ph_prepare(trans))
+               return 0;
+
+       return __mlxsw_sp_port_vlans_add(mlxsw_sp_port,
+                                        vlan->vid_begin, vlan->vid_end,
+                                        untagged_flag, pvid_flag);
+}
+
+static int mlxsw_sp_port_fdb_op(struct mlxsw_sp_port *mlxsw_sp_port,
+                               const char *mac, u16 vid, bool adding,
+                               bool dynamic)
+{
+       enum mlxsw_reg_sfd_rec_policy policy;
+       enum mlxsw_reg_sfd_op op;
+       char *sfd_pl;
+       int err;
+
+       if (!vid)
+               vid = mlxsw_sp_port->pvid;
+
+       sfd_pl = kmalloc(MLXSW_REG_SFD_LEN, GFP_KERNEL);
+       if (!sfd_pl)
+               return -ENOMEM;
+
+       policy = dynamic ? MLXSW_REG_SFD_REC_POLICY_DYNAMIC_ENTRY_INGRESS :
+                          MLXSW_REG_SFD_REC_POLICY_STATIC_ENTRY;
+       op = adding ? MLXSW_REG_SFD_OP_WRITE_EDIT :
+                     MLXSW_REG_SFD_OP_WRITE_REMOVE;
+       mlxsw_reg_sfd_pack(sfd_pl, op, 0);
+       mlxsw_reg_sfd_uc_pack(sfd_pl, 0, policy,
+                             mac, vid, MLXSW_REG_SFD_REC_ACTION_NOP,
+                             mlxsw_sp_port->local_port);
+       err = mlxsw_reg_write(mlxsw_sp_port->mlxsw_sp->core, MLXSW_REG(sfd),
+                             sfd_pl);
+       kfree(sfd_pl);
+
+       return err;
+}
+
+static int
+mlxsw_sp_port_fdb_static_add(struct mlxsw_sp_port *mlxsw_sp_port,
+                            const struct switchdev_obj_port_fdb *fdb,
+                            struct switchdev_trans *trans)
+{
+       if (switchdev_trans_ph_prepare(trans))
+               return 0;
+
+       return mlxsw_sp_port_fdb_op(mlxsw_sp_port, fdb->addr, fdb->vid,
+                                   true, false);
+}
+
+static int mlxsw_sp_port_obj_add(struct net_device *dev,
+                                const struct switchdev_obj *obj,
+                                struct switchdev_trans *trans)
+{
+       struct mlxsw_sp_port *mlxsw_sp_port = netdev_priv(dev);
+       int err = 0;
+
+       switch (obj->id) {
+       case SWITCHDEV_OBJ_ID_PORT_VLAN:
+               err = mlxsw_sp_port_vlans_add(mlxsw_sp_port,
+                                             SWITCHDEV_OBJ_PORT_VLAN(obj),
+                                             trans);
+               break;
+       case SWITCHDEV_OBJ_ID_PORT_FDB:
+               err = mlxsw_sp_port_fdb_static_add(mlxsw_sp_port,
+                                                  SWITCHDEV_OBJ_PORT_FDB(obj),
+                                                  trans);
+               break;
+       default:
+               err = -EOPNOTSUPP;
+               break;
+       }
+
+       return err;
+}
+
+static int mlxsw_sp_port_kill_vids(struct net_device *dev, u16 vid_begin,
+                                  u16 vid_end)
+{
+       u16 vid;
+       int err;
+
+       for (vid = vid_begin; vid <= vid_end; vid++) {
+               err = mlxsw_sp_port_kill_vid(dev, 0, vid);
+               if (err)
+                       return err;
+       }
+
+       return 0;
+}
+
+static int __mlxsw_sp_port_vlans_del(struct mlxsw_sp_port *mlxsw_sp_port,
+                                    u16 vid_begin, u16 vid_end, bool init)
+{
+       struct net_device *dev = mlxsw_sp_port->dev;
+       u16 vid, vid_e;
+       int err;
+
+       /* In case this is invoked with BRIDGE_FLAGS_SELF and port is
+        * not bridged, then prevent packets ingressing through the
+        * port with the specified VIDs from being trapped to CPU.
+        */
+       if (!init && !mlxsw_sp_port->bridged)
+               return mlxsw_sp_port_kill_vids(dev, vid_begin, vid_end);
+
+       for (vid = vid_begin; vid <= vid_end;
+            vid += MLXSW_REG_SPVM_REC_MAX_COUNT) {
+               vid_e = min((u16) (vid + MLXSW_REG_SPVM_REC_MAX_COUNT - 1),
+                           vid_end);
+               err = mlxsw_sp_port_vlan_set(mlxsw_sp_port, vid, vid_e, false,
+                                            false);
+               if (err) {
+                       netdev_err(mlxsw_sp_port->dev, "Unable to del VIDs %d-%d\n",
+                                  vid, vid_e);
+                       return err;
+               }
+       }
+
+       if ((mlxsw_sp_port->pvid >= vid_begin) &&
+           (mlxsw_sp_port->pvid <= vid_end)) {
+               /* Default VLAN is always 1 */
+               mlxsw_sp_port->pvid = 1;
+               err = mlxsw_sp_port_pvid_set(mlxsw_sp_port,
+                                            mlxsw_sp_port->pvid);
+               if (err) {
+                       netdev_err(mlxsw_sp_port->dev, "Unable to del PVID %d\n",
+                                  vid);
+                       return err;
+               }
+       }
+
+       if (init)
+               goto out;
+
+       for (vid = vid_begin; vid <= vid_end; vid++) {
+               err = __mlxsw_sp_port_flood_set(mlxsw_sp_port, vid, false,
+                                               false);
+               if (err) {
+                       netdev_err(dev, "Failed to clear flooding for FID=%d",
+                                  vid);
+                       return err;
+               }
+
+               /* Remove FID mapping in case of Virtual mode */
+               err = mlxsw_sp_port_fid_unmap(mlxsw_sp_port, vid);
+               if (err) {
+                       netdev_err(dev, "Failed to unmap FID=%d", vid);
+                       return err;
+               }
+       }
+
+out:
+       /* Changing activity bits only if HW operation succeded */
+       for (vid = vid_begin; vid <= vid_end; vid++)
+               clear_bit(vid, mlxsw_sp_port->active_vlans);
+
+       return 0;
+}
+
+static int mlxsw_sp_port_vlans_del(struct mlxsw_sp_port *mlxsw_sp_port,
+                                  const struct switchdev_obj_port_vlan *vlan)
+{
+       return __mlxsw_sp_port_vlans_del(mlxsw_sp_port,
+                                        vlan->vid_begin, vlan->vid_end, false);
+}
+
+static int
+mlxsw_sp_port_fdb_static_del(struct mlxsw_sp_port *mlxsw_sp_port,
+                            const struct switchdev_obj_port_fdb *fdb)
+{
+       return mlxsw_sp_port_fdb_op(mlxsw_sp_port, fdb->addr, fdb->vid,
+                                   false, false);
+}
+
+static int mlxsw_sp_port_obj_del(struct net_device *dev,
+                                const struct switchdev_obj *obj)
+{
+       struct mlxsw_sp_port *mlxsw_sp_port = netdev_priv(dev);
+       int err = 0;
+
+       switch (obj->id) {
+       case SWITCHDEV_OBJ_ID_PORT_VLAN:
+               err = mlxsw_sp_port_vlans_del(mlxsw_sp_port,
+                                             SWITCHDEV_OBJ_PORT_VLAN(obj));
+               break;
+       case SWITCHDEV_OBJ_ID_PORT_FDB:
+               err = mlxsw_sp_port_fdb_static_del(mlxsw_sp_port,
+                                                  SWITCHDEV_OBJ_PORT_FDB(obj));
+               break;
+       default:
+               err = -EOPNOTSUPP;
+               break;
+       }
+
+       return err;
+}
+
+static int mlxsw_sp_port_fdb_dump(struct mlxsw_sp_port *mlxsw_sp_port,
+                                 struct switchdev_obj_port_fdb *fdb,
+                                 switchdev_obj_dump_cb_t *cb)
+{
+       char *sfd_pl;
+       char mac[ETH_ALEN];
+       u16 vid;
+       u8 local_port;
+       u8 num_rec;
+       int stored_err = 0;
+       int i;
+       int err;
+
+       sfd_pl = kmalloc(MLXSW_REG_SFD_LEN, GFP_KERNEL);
+       if (!sfd_pl)
+               return -ENOMEM;
+
+       mlxsw_reg_sfd_pack(sfd_pl, MLXSW_REG_SFD_OP_QUERY_DUMP, 0);
+       do {
+               mlxsw_reg_sfd_num_rec_set(sfd_pl, MLXSW_REG_SFD_REC_MAX_COUNT);
+               err = mlxsw_reg_query(mlxsw_sp_port->mlxsw_sp->core,
+                                     MLXSW_REG(sfd), sfd_pl);
+               if (err)
+                       goto out;
+
+               num_rec = mlxsw_reg_sfd_num_rec_get(sfd_pl);
+
+               /* Even in case of error, we have to run the dump to the end
+                * so the session in firmware is finished.
+                */
+               if (stored_err)
+                       continue;
+
+               for (i = 0; i < num_rec; i++) {
+                       switch (mlxsw_reg_sfd_rec_type_get(sfd_pl, i)) {
+                       case MLXSW_REG_SFD_REC_TYPE_UNICAST:
+                               mlxsw_reg_sfd_uc_unpack(sfd_pl, i, mac, &vid,
+                                                       &local_port);
+                               if (local_port == mlxsw_sp_port->local_port) {
+                                       ether_addr_copy(fdb->addr, mac);
+                                       fdb->ndm_state = NUD_REACHABLE;
+                                       fdb->vid = vid;
+                                       err = cb(&fdb->obj);
+                                       if (err)
+                                               stored_err = err;
+                               }
+                       }
+               }
+       } while (num_rec == MLXSW_REG_SFD_REC_MAX_COUNT);
+
+out:
+       kfree(sfd_pl);
+       return stored_err ? stored_err : err;
+}
+
+static int mlxsw_sp_port_vlan_dump(struct mlxsw_sp_port *mlxsw_sp_port,
+                                  struct switchdev_obj_port_vlan *vlan,
+                                  switchdev_obj_dump_cb_t *cb)
+{
+       u16 vid;
+       int err = 0;
+
+       for_each_set_bit(vid, mlxsw_sp_port->active_vlans, VLAN_N_VID) {
+               vlan->flags = 0;
+               if (vid == mlxsw_sp_port->pvid)
+                       vlan->flags |= BRIDGE_VLAN_INFO_PVID;
+               vlan->vid_begin = vid;
+               vlan->vid_end = vid;
+               err = cb(&vlan->obj);
+               if (err)
+                       break;
+       }
+       return err;
+}
+
+static int mlxsw_sp_port_obj_dump(struct net_device *dev,
+                                 struct switchdev_obj *obj,
+                                 switchdev_obj_dump_cb_t *cb)
+{
+       struct mlxsw_sp_port *mlxsw_sp_port = netdev_priv(dev);
+       int err = 0;
+
+       switch (obj->id) {
+       case SWITCHDEV_OBJ_ID_PORT_VLAN:
+               err = mlxsw_sp_port_vlan_dump(mlxsw_sp_port,
+                                             SWITCHDEV_OBJ_PORT_VLAN(obj), cb);
+               break;
+       case SWITCHDEV_OBJ_ID_PORT_FDB:
+               err = mlxsw_sp_port_fdb_dump(mlxsw_sp_port,
+                                            SWITCHDEV_OBJ_PORT_FDB(obj), cb);
+               break;
+       default:
+               err = -EOPNOTSUPP;
+               break;
+       }
+
+       return err;
+}
+
+const struct switchdev_ops mlxsw_sp_port_switchdev_ops = {
+       .switchdev_port_attr_get        = mlxsw_sp_port_attr_get,
+       .switchdev_port_attr_set        = mlxsw_sp_port_attr_set,
+       .switchdev_port_obj_add         = mlxsw_sp_port_obj_add,
+       .switchdev_port_obj_del         = mlxsw_sp_port_obj_del,
+       .switchdev_port_obj_dump        = mlxsw_sp_port_obj_dump,
+};
+
+static void mlxsw_sp_fdb_notify_mac_process(struct mlxsw_sp *mlxsw_sp,
+                                           char *sfn_pl, int rec_index,
+                                           bool adding)
+{
+       struct mlxsw_sp_port *mlxsw_sp_port;
+       char mac[ETH_ALEN];
+       u8 local_port;
+       u16 vid;
+       int err;
+
+       mlxsw_reg_sfn_mac_unpack(sfn_pl, rec_index, mac, &vid, &local_port);
+       mlxsw_sp_port = mlxsw_sp->ports[local_port];
+       if (!mlxsw_sp_port) {
+               dev_err_ratelimited(mlxsw_sp->bus_info->dev, "Incorrect local port in FDB notification\n");
+               return;
+       }
+
+       err = mlxsw_sp_port_fdb_op(mlxsw_sp_port, mac, vid,
+                                  adding && mlxsw_sp_port->learning, true);
+       if (err) {
+               if (net_ratelimit())
+                       netdev_err(mlxsw_sp_port->dev, "Failed to set FDB entry\n");
+               return;
+       }
+
+       if (mlxsw_sp_port->learning && mlxsw_sp_port->learning_sync) {
+               struct switchdev_notifier_fdb_info info;
+               unsigned long notifier_type;
+
+               info.addr = mac;
+               info.vid = vid;
+               notifier_type = adding ? SWITCHDEV_FDB_ADD : SWITCHDEV_FDB_DEL;
+               call_switchdev_notifiers(notifier_type, mlxsw_sp_port->dev,
+                                        &info.info);
+       }
+}
+
+static void mlxsw_sp_fdb_notify_rec_process(struct mlxsw_sp *mlxsw_sp,
+                                           char *sfn_pl, int rec_index)
+{
+       switch (mlxsw_reg_sfn_rec_type_get(sfn_pl, rec_index)) {
+       case MLXSW_REG_SFN_REC_TYPE_LEARNED_MAC:
+               mlxsw_sp_fdb_notify_mac_process(mlxsw_sp, sfn_pl,
+                                               rec_index, true);
+               break;
+       case MLXSW_REG_SFN_REC_TYPE_AGED_OUT_MAC:
+               mlxsw_sp_fdb_notify_mac_process(mlxsw_sp, sfn_pl,
+                                               rec_index, false);
+               break;
+       }
+}
+
+static void mlxsw_sp_fdb_notify_work_schedule(struct mlxsw_sp *mlxsw_sp)
+{
+       schedule_delayed_work(&mlxsw_sp->fdb_notify.dw,
+                             msecs_to_jiffies(mlxsw_sp->fdb_notify.interval));
+}
+
+static void mlxsw_sp_fdb_notify_work(struct work_struct *work)
+{
+       struct mlxsw_sp *mlxsw_sp;
+       char *sfn_pl;
+       u8 num_rec;
+       int i;
+       int err;
+
+       sfn_pl = kmalloc(MLXSW_REG_SFN_LEN, GFP_KERNEL);
+       if (!sfn_pl)
+               return;
+
+       mlxsw_sp = container_of(work, struct mlxsw_sp, fdb_notify.dw.work);
+
+       do {
+               mlxsw_reg_sfn_pack(sfn_pl);
+               err = mlxsw_reg_query(mlxsw_sp->core, MLXSW_REG(sfn), sfn_pl);
+               if (err) {
+                       dev_err_ratelimited(mlxsw_sp->bus_info->dev, "Failed to get FDB notifications\n");
+                       break;
+               }
+               num_rec = mlxsw_reg_sfn_num_rec_get(sfn_pl);
+               for (i = 0; i < num_rec; i++)
+                       mlxsw_sp_fdb_notify_rec_process(mlxsw_sp, sfn_pl, i);
+
+       } while (num_rec);
+
+       kfree(sfn_pl);
+       mlxsw_sp_fdb_notify_work_schedule(mlxsw_sp);
+}
+
+static int mlxsw_sp_fdb_init(struct mlxsw_sp *mlxsw_sp)
+{
+       int err;
+
+       err = mlxsw_sp_ageing_set(mlxsw_sp, MLXSW_SP_DEFAULT_AGEING_TIME);
+       if (err) {
+               dev_err(mlxsw_sp->bus_info->dev, "Failed to set default ageing time\n");
+               return err;
+       }
+       INIT_DELAYED_WORK(&mlxsw_sp->fdb_notify.dw, mlxsw_sp_fdb_notify_work);
+       mlxsw_sp->fdb_notify.interval = MLXSW_SP_DEFAULT_LEARNING_INTERVAL;
+       mlxsw_sp_fdb_notify_work_schedule(mlxsw_sp);
+       return 0;
+}
+
+static void mlxsw_sp_fdb_fini(struct mlxsw_sp *mlxsw_sp)
+{
+       cancel_delayed_work_sync(&mlxsw_sp->fdb_notify.dw);
+}
+
+static void mlxsw_sp_fids_fini(struct mlxsw_sp *mlxsw_sp)
+{
+       u16 fid;
+
+       for_each_set_bit(fid, mlxsw_sp->active_fids, VLAN_N_VID)
+               mlxsw_sp_fid_destroy(mlxsw_sp, fid);
+}
+
+int mlxsw_sp_switchdev_init(struct mlxsw_sp *mlxsw_sp)
+{
+       return mlxsw_sp_fdb_init(mlxsw_sp);
+}
+
+void mlxsw_sp_switchdev_fini(struct mlxsw_sp *mlxsw_sp)
+{
+       mlxsw_sp_fdb_fini(mlxsw_sp);
+       mlxsw_sp_fids_fini(mlxsw_sp);
+}
+
+int mlxsw_sp_port_vlan_init(struct mlxsw_sp_port *mlxsw_sp_port)
+{
+       struct net_device *dev = mlxsw_sp_port->dev;
+       int err;
+
+       /* Allow only untagged packets to ingress and tag them internally
+        * with VID 1.
+        */
+       mlxsw_sp_port->pvid = 1;
+       err = __mlxsw_sp_port_vlans_del(mlxsw_sp_port, 0, VLAN_N_VID, true);
+       if (err) {
+               netdev_err(dev, "Unable to init VLANs\n");
+               return err;
+       }
+
+       /* Add implicit VLAN interface in the device, so that untagged
+        * packets will be classified to the default vFID.
+        */
+       err = mlxsw_sp_port_add_vid(dev, 0, 1);
+       if (err)
+               netdev_err(dev, "Failed to configure default vFID\n");
+
+       return err;
+}
+
+void mlxsw_sp_port_switchdev_init(struct mlxsw_sp_port *mlxsw_sp_port)
+{
+       mlxsw_sp_port->dev->switchdev_ops = &mlxsw_sp_port_switchdev_ops;
+}
+
+void mlxsw_sp_port_switchdev_fini(struct mlxsw_sp_port *mlxsw_sp_port)
+{
+}
index 06fc46c..fdf9472 100644 (file)
@@ -38,6 +38,7 @@
 
 #define MLXSW_TXHDR_LEN 0x10
 #define MLXSW_TXHDR_VERSION_0 0
+#define MLXSW_TXHDR_VERSION_1 1
 
 enum {
        MLXSW_TXHDR_ETH_CTL,