]> git.proxmox.com Git - mirror_ubuntu-hirsute-kernel.git/commitdiff
liquidio: switchdev support for LiquidIO NIC
authorVijaya Mohan Guvva <vijaya.guvva@cavium.com>
Tue, 31 Oct 2017 23:04:53 +0000 (16:04 -0700)
committerDavid S. Miller <davem@davemloft.net>
Wed, 1 Nov 2017 02:39:52 +0000 (11:39 +0900)
Enable switchdev for SRIOV capable LiquidIO NIC. It registers
a representor netdev (with switchdev_ops) for each SRIOV VF created.
It also has changes to send representor interface configurations like
admin state and MTU to LiquidIO firmware and to retrieve HW counted
VF stats for VF representor.

Signed-off-by: Vijaya Mohan Guvva <vijaya.guvva@cavium.com>
Signed-off-by: Satanand Burla <satananda.burla@cavium.com>
Signed-off-by: Raghu Vatsavayi <raghu.vatsavayi@cavium.com>
Signed-off-by: Felix Manlunas <felix.manlunas@cavium.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
drivers/net/ethernet/cavium/liquidio/Makefile
drivers/net/ethernet/cavium/liquidio/lio_main.c
drivers/net/ethernet/cavium/liquidio/lio_vf_rep.c [new file with mode: 0644]
drivers/net/ethernet/cavium/liquidio/lio_vf_rep.h [new file with mode: 0644]
drivers/net/ethernet/cavium/liquidio/liquidio_common.h
drivers/net/ethernet/cavium/liquidio/octeon_device.h

index c4d411d1aa287a76ff8d8a446fa56a4c352d694a..cad4fe1ffe55e8817d3f4b3b69465d1374ef85ad 100644 (file)
@@ -17,7 +17,7 @@ liquidio-$(CONFIG_LIQUIDIO) += lio_ethtool.o \
                        octeon_droq.o      \
                        octeon_nic.o
 
-liquidio-objs := lio_main.o octeon_console.o $(liquidio-y)
+liquidio-objs := lio_main.o octeon_console.o lio_vf_rep.o $(liquidio-y)
 
 obj-$(CONFIG_LIQUIDIO_VF) += liquidio_vf.o
 
index 8ea24d68e82430b75f78cde8571b561b469c9b1e..80784122e6e9a95b7d240de6028e16d13d93c079 100644 (file)
@@ -21,6 +21,7 @@
 #include <linux/firmware.h>
 #include <net/vxlan.h>
 #include <linux/kthread.h>
+#include <net/switchdev.h>
 #include "liquidio_common.h"
 #include "octeon_droq.h"
 #include "octeon_iq.h"
@@ -3309,6 +3310,29 @@ static int liquidio_set_vf_link_state(struct net_device *netdev, int vfidx,
        return 0;
 }
 
+static int
+lio_pf_switchdev_attr_get(struct net_device *dev, struct switchdev_attr *attr)
+{
+       struct lio *lio = GET_LIO(dev);
+
+       switch (attr->id) {
+       case SWITCHDEV_ATTR_ID_PORT_PARENT_ID:
+               attr->u.ppid.id_len = ETH_ALEN;
+               ether_addr_copy(attr->u.ppid.id,
+                               (void *)&lio->linfo.hw_addr + 2);
+               break;
+
+       default:
+               return -EOPNOTSUPP;
+       }
+
+       return 0;
+}
+
+static const struct switchdev_ops lio_pf_switchdev_ops = {
+       .switchdev_port_attr_get = lio_pf_switchdev_attr_get,
+};
+
 static const struct net_device_ops lionetdevops = {
        .ndo_open               = liquidio_open,
        .ndo_stop               = liquidio_stop,
@@ -3583,6 +3607,7 @@ static int setup_nic_devices(struct octeon_device *octeon_dev)
                 * netdev tasks.
                 */
                netdev->netdev_ops = &lionetdevops;
+               SWITCHDEV_SET_OPS(netdev, &lio_pf_switchdev_ops);
 
                lio = GET_LIO(netdev);
 
diff --git a/drivers/net/ethernet/cavium/liquidio/lio_vf_rep.c b/drivers/net/ethernet/cavium/liquidio/lio_vf_rep.c
new file mode 100644 (file)
index 0000000..67ff7a1
--- /dev/null
@@ -0,0 +1,621 @@
+/**********************************************************************
+ * Author: Cavium, Inc.
+ *
+ * Contact: support@cavium.com
+ *          Please include "LiquidIO" in the subject.
+ *
+ * Copyright (c) 2003-2017 Cavium, Inc.
+ *
+ * This file is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License, Version 2, as
+ * published by the Free Software Foundation.
+ *
+ * This file is distributed in the hope that it will be useful, but
+ * AS-IS and WITHOUT ANY WARRANTY; without even the implied warranty
+ * of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE, TITLE, or
+ * NONINFRINGEMENT.  See the GNU General Public License for more details.
+ ***********************************************************************/
+#include <linux/pci.h>
+#include <linux/if_vlan.h>
+#include "liquidio_common.h"
+#include "octeon_droq.h"
+#include "octeon_iq.h"
+#include "response_manager.h"
+#include "octeon_device.h"
+#include "octeon_nic.h"
+#include "octeon_main.h"
+#include "octeon_network.h"
+#include <net/switchdev.h>
+#include "lio_vf_rep.h"
+#include "octeon_network.h"
+
+static int lio_vf_rep_open(struct net_device *ndev);
+static int lio_vf_rep_stop(struct net_device *ndev);
+static int lio_vf_rep_pkt_xmit(struct sk_buff *skb, struct net_device *ndev);
+static void lio_vf_rep_tx_timeout(struct net_device *netdev);
+static int lio_vf_rep_phys_port_name(struct net_device *dev,
+                                    char *buf, size_t len);
+static void lio_vf_rep_get_stats64(struct net_device *dev,
+                                  struct rtnl_link_stats64 *stats64);
+static int lio_vf_rep_change_mtu(struct net_device *ndev, int new_mtu);
+
+static const struct net_device_ops lio_vf_rep_ndev_ops = {
+       .ndo_open = lio_vf_rep_open,
+       .ndo_stop = lio_vf_rep_stop,
+       .ndo_start_xmit = lio_vf_rep_pkt_xmit,
+       .ndo_tx_timeout = lio_vf_rep_tx_timeout,
+       .ndo_get_phys_port_name = lio_vf_rep_phys_port_name,
+       .ndo_get_stats64 = lio_vf_rep_get_stats64,
+       .ndo_change_mtu = lio_vf_rep_change_mtu,
+};
+
+static void
+lio_vf_rep_send_sc_complete(struct octeon_device *oct,
+                           u32 status, void *ptr)
+{
+       struct octeon_soft_command *sc = (struct octeon_soft_command *)ptr;
+       struct lio_vf_rep_sc_ctx *ctx =
+               (struct lio_vf_rep_sc_ctx *)sc->ctxptr;
+       struct lio_vf_rep_resp *resp =
+               (struct lio_vf_rep_resp *)sc->virtrptr;
+
+       if (status != OCTEON_REQUEST_TIMEOUT && READ_ONCE(resp->status))
+               WRITE_ONCE(resp->status, 0);
+
+       complete(&ctx->complete);
+}
+
+static int
+lio_vf_rep_send_soft_command(struct octeon_device *oct,
+                            void *req, int req_size,
+                            void *resp, int resp_size)
+{
+       int tot_resp_size = sizeof(struct lio_vf_rep_resp) + resp_size;
+       int ctx_size = sizeof(struct lio_vf_rep_sc_ctx);
+       struct octeon_soft_command *sc = NULL;
+       struct lio_vf_rep_resp *rep_resp;
+       struct lio_vf_rep_sc_ctx *ctx;
+       void *sc_req;
+       int err;
+
+       sc = (struct octeon_soft_command *)
+               octeon_alloc_soft_command(oct, req_size,
+                                         tot_resp_size, ctx_size);
+       if (!sc)
+               return -ENOMEM;
+
+       ctx = (struct lio_vf_rep_sc_ctx *)sc->ctxptr;
+       memset(ctx, 0, ctx_size);
+       init_completion(&ctx->complete);
+
+       sc_req = (struct lio_vf_rep_req *)sc->virtdptr;
+       memcpy(sc_req, req, req_size);
+
+       rep_resp = (struct lio_vf_rep_resp *)sc->virtrptr;
+       memset(rep_resp, 0, tot_resp_size);
+       WRITE_ONCE(rep_resp->status, 1);
+
+       sc->iq_no = 0;
+       octeon_prepare_soft_command(oct, sc, OPCODE_NIC,
+                                   OPCODE_NIC_VF_REP_CMD, 0, 0, 0);
+       sc->callback = lio_vf_rep_send_sc_complete;
+       sc->callback_arg = sc;
+       sc->wait_time = LIO_VF_REP_REQ_TMO_MS;
+
+       err = octeon_send_soft_command(oct, sc);
+       if (err == IQ_SEND_FAILED)
+               goto free_buff;
+
+       wait_for_completion_timeout(&ctx->complete,
+                                   msecs_to_jiffies
+                                   (2 * LIO_VF_REP_REQ_TMO_MS));
+       err = READ_ONCE(rep_resp->status) ? -EBUSY : 0;
+       if (err)
+               dev_err(&oct->pci_dev->dev, "VF rep send config failed\n");
+
+       if (resp)
+               memcpy(resp, (rep_resp + 1), resp_size);
+free_buff:
+       octeon_free_soft_command(oct, sc);
+
+       return err;
+}
+
+static int
+lio_vf_rep_open(struct net_device *ndev)
+{
+       struct lio_vf_rep_desc *vf_rep = netdev_priv(ndev);
+       struct lio_vf_rep_req rep_cfg;
+       struct octeon_device *oct;
+       int ret;
+
+       oct = vf_rep->oct;
+
+       memset(&rep_cfg, 0, sizeof(rep_cfg));
+       rep_cfg.req_type = LIO_VF_REP_REQ_STATE;
+       rep_cfg.ifidx = vf_rep->ifidx;
+       rep_cfg.rep_state.state = LIO_VF_REP_STATE_UP;
+
+       ret = lio_vf_rep_send_soft_command(oct, &rep_cfg,
+                                          sizeof(rep_cfg), NULL, 0);
+
+       if (ret) {
+               dev_err(&oct->pci_dev->dev,
+                       "VF_REP open failed with err %d\n", ret);
+               return -EIO;
+       }
+
+       atomic_set(&vf_rep->ifstate, (atomic_read(&vf_rep->ifstate) |
+                                     LIO_IFSTATE_RUNNING));
+
+       netif_carrier_on(ndev);
+       netif_start_queue(ndev);
+
+       return 0;
+}
+
+static int
+lio_vf_rep_stop(struct net_device *ndev)
+{
+       struct lio_vf_rep_desc *vf_rep = netdev_priv(ndev);
+       struct lio_vf_rep_req rep_cfg;
+       struct octeon_device *oct;
+       int ret;
+
+       oct = vf_rep->oct;
+
+       memset(&rep_cfg, 0, sizeof(rep_cfg));
+       rep_cfg.req_type = LIO_VF_REP_REQ_STATE;
+       rep_cfg.ifidx = vf_rep->ifidx;
+       rep_cfg.rep_state.state = LIO_VF_REP_STATE_DOWN;
+
+       ret = lio_vf_rep_send_soft_command(oct, &rep_cfg,
+                                          sizeof(rep_cfg), NULL, 0);
+
+       if (ret) {
+               dev_err(&oct->pci_dev->dev,
+                       "VF_REP dev stop failed with err %d\n", ret);
+               return -EIO;
+       }
+
+       atomic_set(&vf_rep->ifstate, (atomic_read(&vf_rep->ifstate) &
+                                     ~LIO_IFSTATE_RUNNING));
+
+       netif_tx_disable(ndev);
+       netif_carrier_off(ndev);
+
+       return 0;
+}
+
+static void
+lio_vf_rep_tx_timeout(struct net_device *ndev)
+{
+       netif_trans_update(ndev);
+
+       netif_wake_queue(ndev);
+}
+
+static void
+lio_vf_rep_get_stats64(struct net_device *dev,
+                      struct rtnl_link_stats64 *stats64)
+{
+       struct lio_vf_rep_desc *vf_rep = netdev_priv(dev);
+
+       stats64->tx_packets = vf_rep->stats.tx_packets;
+       stats64->tx_bytes   = vf_rep->stats.tx_bytes;
+       stats64->tx_dropped = vf_rep->stats.tx_dropped;
+
+       stats64->rx_packets = vf_rep->stats.rx_packets;
+       stats64->rx_bytes   = vf_rep->stats.rx_bytes;
+       stats64->rx_dropped = vf_rep->stats.rx_dropped;
+}
+
+static int
+lio_vf_rep_change_mtu(struct net_device *ndev, int new_mtu)
+{
+       struct lio_vf_rep_desc *vf_rep = netdev_priv(ndev);
+       struct lio_vf_rep_req rep_cfg;
+       struct octeon_device *oct;
+       int ret;
+
+       oct = vf_rep->oct;
+
+       memset(&rep_cfg, 0, sizeof(rep_cfg));
+       rep_cfg.req_type = LIO_VF_REP_REQ_MTU;
+       rep_cfg.ifidx = vf_rep->ifidx;
+       rep_cfg.rep_mtu.mtu = cpu_to_be32(new_mtu);
+
+       ret = lio_vf_rep_send_soft_command(oct, &rep_cfg,
+                                          sizeof(rep_cfg), NULL, 0);
+       if (ret) {
+               dev_err(&oct->pci_dev->dev,
+                       "Change MTU failed with err %d\n", ret);
+               return -EIO;
+       }
+
+       ndev->mtu = new_mtu;
+
+       return 0;
+}
+
+static int
+lio_vf_rep_phys_port_name(struct net_device *dev,
+                         char *buf, size_t len)
+{
+       struct lio_vf_rep_desc *vf_rep = netdev_priv(dev);
+       struct octeon_device *oct = vf_rep->oct;
+       int ret;
+
+       ret = snprintf(buf, len, "pf%dvf%d", oct->pf_num,
+                      vf_rep->ifidx - oct->pf_num * 64 - 1);
+       if (ret >= len)
+               return -EOPNOTSUPP;
+
+       return 0;
+}
+
+static struct net_device *
+lio_vf_rep_get_ndev(struct octeon_device *oct, int ifidx)
+{
+       int vf_id, max_vfs = CN23XX_MAX_VFS_PER_PF + 1;
+       int vfid_mask = max_vfs - 1;
+
+       if (ifidx <= oct->pf_num * max_vfs ||
+           ifidx >= oct->pf_num * max_vfs + max_vfs)
+               return NULL;
+
+       /* ifidx 1-63 for PF0 VFs
+        * ifidx 65-127 for PF1 VFs
+        */
+       vf_id = (ifidx & vfid_mask) - 1;
+
+       return oct->vf_rep_list.ndev[vf_id];
+}
+
+static void
+lio_vf_rep_copy_packet(struct octeon_device *oct,
+                      struct sk_buff *skb,
+                      int len)
+{
+       if (likely(len > MIN_SKB_SIZE)) {
+               struct octeon_skb_page_info *pg_info;
+               unsigned char *va;
+
+               pg_info = ((struct octeon_skb_page_info *)(skb->cb));
+               if (pg_info->page) {
+                       va = page_address(pg_info->page) +
+                               pg_info->page_offset;
+                       memcpy(skb->data, va, MIN_SKB_SIZE);
+                       skb_put(skb, MIN_SKB_SIZE);
+               }
+
+               skb_add_rx_frag(skb, skb_shinfo(skb)->nr_frags,
+                               pg_info->page,
+                               pg_info->page_offset + MIN_SKB_SIZE,
+                               len - MIN_SKB_SIZE,
+                               LIO_RXBUFFER_SZ);
+       } else {
+               struct octeon_skb_page_info *pg_info =
+                       ((struct octeon_skb_page_info *)(skb->cb));
+
+               skb_copy_to_linear_data(skb, page_address(pg_info->page) +
+                                       pg_info->page_offset, len);
+               skb_put(skb, len);
+               put_page(pg_info->page);
+       }
+}
+
+static int
+lio_vf_rep_pkt_recv(struct octeon_recv_info *recv_info, void *buf)
+{
+       struct octeon_recv_pkt *recv_pkt = recv_info->recv_pkt;
+       struct lio_vf_rep_desc *vf_rep;
+       struct net_device *vf_ndev;
+       struct octeon_device *oct;
+       union octeon_rh *rh;
+       struct sk_buff *skb;
+       int i, ifidx;
+
+       oct = lio_get_device(recv_pkt->octeon_id);
+       if (!oct)
+               goto free_buffers;
+
+       skb = recv_pkt->buffer_ptr[0];
+       rh = &recv_pkt->rh;
+       ifidx = rh->r.ossp;
+
+       vf_ndev = lio_vf_rep_get_ndev(oct, ifidx);
+       if (!vf_ndev)
+               goto free_buffers;
+
+       vf_rep = netdev_priv(vf_ndev);
+       if (!(atomic_read(&vf_rep->ifstate) & LIO_IFSTATE_RUNNING) ||
+           recv_pkt->buffer_count > 1)
+               goto free_buffers;
+
+       skb->dev = vf_ndev;
+
+       /* Multiple buffers are not used for vf_rep packets.
+        * So just buffer_size[0] is valid.
+        */
+       lio_vf_rep_copy_packet(oct, skb, recv_pkt->buffer_size[0]);
+
+       skb_pull(skb, rh->r_dh.len * BYTES_PER_DHLEN_UNIT);
+       skb->protocol = eth_type_trans(skb, skb->dev);
+       skb->ip_summed = CHECKSUM_NONE;
+
+       netif_rx(skb);
+
+       octeon_free_recv_info(recv_info);
+
+       return 0;
+
+free_buffers:
+       for (i = 0; i < recv_pkt->buffer_count; i++)
+               recv_buffer_free(recv_pkt->buffer_ptr[i]);
+
+       octeon_free_recv_info(recv_info);
+
+       return 0;
+}
+
+static void
+lio_vf_rep_packet_sent_callback(struct octeon_device *oct,
+                               u32 status, void *buf)
+{
+       struct octeon_soft_command *sc = (struct octeon_soft_command *)buf;
+       struct sk_buff *skb = sc->ctxptr;
+       struct net_device *ndev = skb->dev;
+
+       dma_unmap_single(&oct->pci_dev->dev, sc->dmadptr,
+                        sc->datasize, DMA_TO_DEVICE);
+       dev_kfree_skb_any(skb);
+       octeon_free_soft_command(oct, sc);
+
+       if (octnet_iq_is_full(oct, sc->iq_no))
+               return;
+
+       if (netif_queue_stopped(ndev))
+               netif_wake_queue(ndev);
+}
+
+static int
+lio_vf_rep_pkt_xmit(struct sk_buff *skb, struct net_device *ndev)
+{
+       struct lio_vf_rep_desc *vf_rep = netdev_priv(ndev);
+       struct net_device *parent_ndev = vf_rep->parent_ndev;
+       struct octeon_device *oct = vf_rep->oct;
+       struct octeon_instr_pki_ih3 *pki_ih3;
+       struct octeon_soft_command *sc;
+       struct lio *parent_lio;
+       int status;
+
+       parent_lio = GET_LIO(parent_ndev);
+
+       if (!(atomic_read(&vf_rep->ifstate) & LIO_IFSTATE_RUNNING) ||
+           skb->len <= 0)
+               goto xmit_failed;
+
+       if (octnet_iq_is_full(vf_rep->oct, parent_lio->txq)) {
+               dev_err(&oct->pci_dev->dev, "VF rep: Device IQ full\n");
+               netif_stop_queue(ndev);
+               return NETDEV_TX_BUSY;
+       }
+
+       sc = (struct octeon_soft_command *)
+               octeon_alloc_soft_command(oct, 0, 0, 0);
+       if (!sc) {
+               dev_err(&oct->pci_dev->dev, "VF rep: Soft command alloc failed\n");
+               goto xmit_failed;
+       }
+
+       /* Multiple buffers are not used for vf_rep packets. */
+       if (skb_shinfo(skb)->nr_frags != 0) {
+               dev_err(&oct->pci_dev->dev, "VF rep: nr_frags != 0. Dropping packet\n");
+               goto xmit_failed;
+       }
+
+       sc->dmadptr = dma_map_single(&oct->pci_dev->dev,
+                                    skb->data, skb->len, DMA_TO_DEVICE);
+       if (dma_mapping_error(&oct->pci_dev->dev, sc->dmadptr)) {
+               dev_err(&oct->pci_dev->dev, "VF rep: DMA mapping failed\n");
+               goto xmit_failed;
+       }
+
+       sc->virtdptr = skb->data;
+       sc->datasize = skb->len;
+       sc->ctxptr = skb;
+       sc->iq_no = parent_lio->txq;
+
+       octeon_prepare_soft_command(oct, sc, OPCODE_NIC, OPCODE_NIC_VF_REP_PKT,
+                                   vf_rep->ifidx, 0, 0);
+       pki_ih3 = (struct octeon_instr_pki_ih3 *)&sc->cmd.cmd3.pki_ih3;
+       pki_ih3->tagtype = ORDERED_TAG;
+
+       sc->callback = lio_vf_rep_packet_sent_callback;
+       sc->callback_arg = sc;
+
+       status = octeon_send_soft_command(oct, sc);
+       if (status == IQ_SEND_FAILED) {
+               dma_unmap_single(&oct->pci_dev->dev, sc->dmadptr,
+                                sc->datasize, DMA_TO_DEVICE);
+               goto xmit_failed;
+       }
+
+       if (status == IQ_SEND_STOP)
+               netif_stop_queue(ndev);
+
+       netif_trans_update(ndev);
+
+       return NETDEV_TX_OK;
+
+xmit_failed:
+       dev_kfree_skb_any(skb);
+
+       return NETDEV_TX_OK;
+}
+
+static int
+lio_vf_rep_attr_get(struct net_device *dev, struct switchdev_attr *attr)
+{
+       struct lio_vf_rep_desc *vf_rep = netdev_priv(dev);
+       struct net_device *parent_ndev = vf_rep->parent_ndev;
+       struct lio *lio = GET_LIO(parent_ndev);
+
+       switch (attr->id) {
+       case SWITCHDEV_ATTR_ID_PORT_PARENT_ID:
+               attr->u.ppid.id_len = ETH_ALEN;
+               ether_addr_copy(attr->u.ppid.id,
+                               (void *)&lio->linfo.hw_addr + 2);
+               break;
+
+       default:
+               return -EOPNOTSUPP;
+       }
+
+       return 0;
+}
+
+static const struct switchdev_ops lio_vf_rep_switchdev_ops = {
+       .switchdev_port_attr_get        = lio_vf_rep_attr_get,
+};
+
+static void
+lio_vf_rep_fetch_stats(struct work_struct *work)
+{
+       struct cavium_wk *wk = (struct cavium_wk *)work;
+       struct lio_vf_rep_desc *vf_rep = wk->ctxptr;
+       struct lio_vf_rep_stats stats;
+       struct lio_vf_rep_req rep_cfg;
+       struct octeon_device *oct;
+       int ret;
+
+       oct = vf_rep->oct;
+
+       memset(&rep_cfg, 0, sizeof(rep_cfg));
+       rep_cfg.req_type = LIO_VF_REP_REQ_STATS;
+       rep_cfg.ifidx = vf_rep->ifidx;
+
+       ret = lio_vf_rep_send_soft_command(oct, &rep_cfg, sizeof(rep_cfg),
+                                          &stats, sizeof(stats));
+
+       if (!ret) {
+               octeon_swap_8B_data((u64 *)&stats, (sizeof(stats) >> 3));
+               memcpy(&vf_rep->stats, &stats, sizeof(stats));
+       }
+
+       schedule_delayed_work(&vf_rep->stats_wk.work,
+                             msecs_to_jiffies(LIO_VF_REP_STATS_POLL_TIME_MS));
+}
+
+int
+lio_vf_rep_create(struct octeon_device *oct)
+{
+       struct lio_vf_rep_desc *vf_rep;
+       struct net_device *ndev;
+       int i, num_vfs;
+
+       if (!oct->sriov_info.sriov_enabled)
+               return 0;
+
+       num_vfs = oct->sriov_info.num_vfs_alloced;
+
+       oct->vf_rep_list.num_vfs = 0;
+       for (i = 0; i < num_vfs; i++) {
+               ndev = alloc_etherdev(sizeof(struct lio_vf_rep_desc));
+
+               if (!ndev) {
+                       dev_err(&oct->pci_dev->dev,
+                               "VF rep device %d creation failed\n", i);
+                       goto cleanup;
+               }
+
+               ndev->min_mtu = LIO_MIN_MTU_SIZE;
+               ndev->max_mtu = LIO_MAX_MTU_SIZE;
+               ndev->netdev_ops = &lio_vf_rep_ndev_ops;
+               SWITCHDEV_SET_OPS(ndev, &lio_vf_rep_switchdev_ops);
+
+               vf_rep = netdev_priv(ndev);
+               memset(vf_rep, 0, sizeof(*vf_rep));
+
+               vf_rep->ndev = ndev;
+               vf_rep->oct = oct;
+               vf_rep->parent_ndev = oct->props[0].netdev;
+               vf_rep->ifidx = (oct->pf_num * 64) + i + 1;
+
+               eth_hw_addr_random(ndev);
+
+               if (register_netdev(ndev)) {
+                       dev_err(&oct->pci_dev->dev, "VF rep nerdev registration failed\n");
+
+                       free_netdev(ndev);
+                       goto cleanup;
+               }
+
+               netif_carrier_off(ndev);
+
+               INIT_DELAYED_WORK(&vf_rep->stats_wk.work,
+                                 lio_vf_rep_fetch_stats);
+               vf_rep->stats_wk.ctxptr = (void *)vf_rep;
+               schedule_delayed_work(&vf_rep->stats_wk.work,
+                                     msecs_to_jiffies
+                                     (LIO_VF_REP_STATS_POLL_TIME_MS));
+               oct->vf_rep_list.num_vfs++;
+               oct->vf_rep_list.ndev[i] = ndev;
+       }
+
+       if (octeon_register_dispatch_fn(oct, OPCODE_NIC,
+                                       OPCODE_NIC_VF_REP_PKT,
+                                       lio_vf_rep_pkt_recv, oct)) {
+               dev_err(&oct->pci_dev->dev, "VF rep Dispatch func registration failed\n");
+
+               goto cleanup;
+       }
+
+       return 0;
+
+cleanup:
+       for (i = 0; i < oct->vf_rep_list.num_vfs; i++) {
+               ndev = oct->vf_rep_list.ndev[i];
+               oct->vf_rep_list.ndev[i] = NULL;
+               if (ndev) {
+                       vf_rep = netdev_priv(ndev);
+                       cancel_delayed_work_sync
+                               (&vf_rep->stats_wk.work);
+                       unregister_netdev(ndev);
+                       free_netdev(ndev);
+               }
+       }
+
+       oct->vf_rep_list.num_vfs = 0;
+
+       return -1;
+}
+
+void
+lio_vf_rep_destroy(struct octeon_device *oct)
+{
+       struct lio_vf_rep_desc *vf_rep;
+       struct net_device *ndev;
+       int i;
+
+       if (!oct->sriov_info.sriov_enabled)
+               return;
+
+       for (i = 0; i < oct->vf_rep_list.num_vfs; i++) {
+               ndev = oct->vf_rep_list.ndev[i];
+               oct->vf_rep_list.ndev[i] = NULL;
+               if (ndev) {
+                       vf_rep = netdev_priv(ndev);
+                       cancel_delayed_work_sync
+                               (&vf_rep->stats_wk.work);
+                       netif_tx_disable(ndev);
+                       netif_carrier_off(ndev);
+
+                       unregister_netdev(ndev);
+                       free_netdev(ndev);
+               }
+       }
+
+       oct->vf_rep_list.num_vfs = 0;
+}
diff --git a/drivers/net/ethernet/cavium/liquidio/lio_vf_rep.h b/drivers/net/ethernet/cavium/liquidio/lio_vf_rep.h
new file mode 100644 (file)
index 0000000..5a9ec98
--- /dev/null
@@ -0,0 +1,47 @@
+/**********************************************************************
+ * Author: Cavium, Inc.
+ *
+ * Contact: support@cavium.com
+ *          Please include "LiquidIO" in the subject.
+ *
+ * Copyright (c) 2003-2017 Cavium, Inc.
+ *
+ * This file is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License, Version 2, as
+ * published by the Free Software Foundation.
+ *
+ * This file is distributed in the hope that it will be useful, but
+ * AS-IS and WITHOUT ANY WARRANTY; without even the implied warranty
+ * of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE, TITLE, or
+ * NONINFRINGEMENT.  See the GNU General Public License for more
+ * details.
+ *
+ * This file may also be available under a different license from Cavium.
+ * Contact Cavium, Inc. for more information
+ **********************************************************************/
+
+/*! \file octeon_vf_main.h
+ *  \brief Host Driver: This file defines vf_rep related macros and structures
+ */
+#ifndef __LIO_VF_REP_H__
+#define __LIO_VF_REP_H__
+#define LIO_VF_REP_REQ_TMO_MS 5000
+#define LIO_VF_REP_STATS_POLL_TIME_MS 200
+
+struct lio_vf_rep_desc {
+       struct net_device *parent_ndev;
+       struct net_device *ndev;
+       struct octeon_device *oct;
+       struct lio_vf_rep_stats stats;
+       struct cavium_wk stats_wk;
+       atomic_t ifstate;
+       int ifidx;
+};
+
+struct lio_vf_rep_sc_ctx {
+       struct completion complete;
+};
+
+int lio_vf_rep_create(struct octeon_device *oct);
+void lio_vf_rep_destroy(struct octeon_device *oct);
+#endif
index 2033a65cd97a31574e1542551795ded1b8e01b6c..3aceb78caa8696db93c16c72b99e7345e5f51b08 100644 (file)
@@ -89,6 +89,9 @@ enum octeon_tag_type {
 #define VF_DRV_REMOVED                -1
 #define VF_DRV_MACADDR_CHANGED         2
 
+#define OPCODE_NIC_VF_REP_PKT          0x15
+#define OPCODE_NIC_VF_REP_CMD          0x16
+
 #define CORE_DRV_TEST_SCATTER_OP    0xFFF5
 
 /* Application codes advertised by the core driver initialization packet. */
@@ -909,4 +912,50 @@ struct lio_time {
        s64 sec;   /* seconds */
        s64 nsec;  /* nanoseconds */
 };
+
+struct lio_vf_rep_stats {
+       u64 tx_packets;
+       u64 tx_bytes;
+       u64 tx_dropped;
+
+       u64 rx_packets;
+       u64 rx_bytes;
+       u64 rx_dropped;
+};
+
+enum lio_vf_rep_req_type {
+       LIO_VF_REP_REQ_NONE,
+       LIO_VF_REP_REQ_STATE,
+       LIO_VF_REP_REQ_MTU,
+       LIO_VF_REP_REQ_STATS
+};
+
+enum {
+       LIO_VF_REP_STATE_DOWN,
+       LIO_VF_REP_STATE_UP
+};
+
+struct lio_vf_rep_req {
+       u8 req_type;
+       u8 ifidx;
+       u8 rsvd[6];
+
+       union {
+               struct lio_vf_rep_mtu {
+                       u32 mtu;
+                       u32 rsvd;
+               } rep_mtu;
+
+               struct lio_vf_rep_state {
+                       u8 state;
+                       u8 rsvd[7];
+               } rep_state;
+       };
+};
+
+struct lio_vf_rep_resp {
+       u64 rh;
+       u8  status;
+       u8  rsvd[7];
+};
 #endif
index 33d19c4509bc4a2ad7b3432a8b2f1b63c336b14c..c97e067a89743804dc643ee19f7eaace85a52aaa 100644 (file)
@@ -391,6 +391,11 @@ struct octeon_ioq_vector {
        u32                     ioq_num;
 };
 
+struct lio_vf_rep_list {
+       int num_vfs;
+       struct net_device *ndev[CN23XX_MAX_VFS_PER_PF];
+};
+
 /** The Octeon device.
  *  Each Octeon device has this structure to represent all its
  *  components.
@@ -568,6 +573,8 @@ struct octeon_device {
        atomic_t *adapter_fw_state; /* per-adapter, lio_fw_state */
 
        bool ptp_enable;
+
+       struct lio_vf_rep_list vf_rep_list;
 };
 
 #define  OCT_DRV_ONLINE 1