datapath/vport-internal_dev.c

   1 /*
   2  * Copyright (c) 2007-2012 Nicira, Inc.
   3  *
   4  * This program is free software; you can redistribute it and/or
   5  * modify it under the terms of version 2 of the GNU General Public
   6  * License as published by the Free Software Foundation.
   7  *
   8  * This program is distributed in the hope that it will be useful, but
   9  * WITHOUT ANY WARRANTY; without even the implied warranty of
  10  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
  11  * General Public License for more details.
  12  *
  13  * You should have received a copy of the GNU General Public License
  14  * along with this program; if not, write to the Free Software
  15  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA
  16  * 02110-1301, USA
  17  */
  18
  19 #include <linux/hardirq.h>
  20 #include <linux/if_vlan.h>
  21 #include <linux/kernel.h>
  22 #include <linux/netdevice.h>
  23 #include <linux/etherdevice.h>
  24 #include <linux/ethtool.h>
  25 #include <linux/skbuff.h>
  26 #include <linux/version.h>
  27
  28 #include <net/dst.h>
  29 #include <net/xfrm.h>
  30
  31 #include "checksum.h"
  32 #include "datapath.h"
  33 #include "vlan.h"
  34 #include "vport-internal_dev.h"
  35 #include "vport-netdev.h"
  36
  37 #if LINUX_VERSION_CODE >= KERNEL_VERSION(3,1,0)
  38 #define HAVE_NET_DEVICE_OPS
  39 #endif
  40
  41 struct internal_dev {
  42         struct vport *vport;
  43 #if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,22)
  44         struct net_device_stats stats;
  45 #endif
  46 };
  47
  48 static struct internal_dev *internal_dev_priv(struct net_device *netdev)
  49 {
  50         return netdev_priv(netdev);
  51 }
  52
  53 /* This function is only called by the kernel network layer.*/
  54 #if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,36)
  55 static struct rtnl_link_stats64 *internal_dev_get_stats(struct net_device *netdev,
  56                                                         struct rtnl_link_stats64 *stats)
  57 {
  58 #else
  59 static struct net_device_stats *internal_dev_sys_stats(struct net_device *netdev)
  60 {
  61 #if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,22)
  62         struct net_device_stats *stats = &internal_dev_priv(netdev)->stats;
  63 #else
  64         struct net_device_stats *stats = &netdev->stats;
  65 #endif
  66 #endif
  67         struct vport *vport = ovs_internal_dev_get_vport(netdev);
  68         struct ovs_vport_stats vport_stats;
  69
  70         ovs_vport_get_stats(vport, &vport_stats);
  71
  72         /* The tx and rx stats need to be swapped because the
  73          * switch and host OS have opposite perspectives. */
  74         stats->rx_packets       = vport_stats.tx_packets;
  75         stats->tx_packets       = vport_stats.rx_packets;
  76         stats->rx_bytes         = vport_stats.tx_bytes;
  77         stats->tx_bytes         = vport_stats.rx_bytes;
  78         stats->rx_errors        = vport_stats.tx_errors;
  79         stats->tx_errors        = vport_stats.rx_errors;
  80         stats->rx_dropped       = vport_stats.tx_dropped;
  81         stats->tx_dropped       = vport_stats.rx_dropped;
  82
  83         return stats;
  84 }
  85
  86 /* Called with rcu_read_lock_bh. */
  87 static int internal_dev_xmit(struct sk_buff *skb, struct net_device *netdev)
  88 {
  89         if (unlikely(compute_ip_summed(skb, true))) {
  90                 kfree_skb(skb);
  91                 return 0;
  92         }
  93
  94         vlan_copy_skb_tci(skb);
  95         OVS_CB(skb)->flow = NULL;
  96
  97         rcu_read_lock();
  98         ovs_vport_receive(internal_dev_priv(netdev)->vport, skb);
  99         rcu_read_unlock();
 100         return 0;
 101 }
 102
 103 static int internal_dev_open(struct net_device *netdev)
 104 {
 105         netif_start_queue(netdev);
 106         return 0;
 107 }
 108
 109 static int internal_dev_stop(struct net_device *netdev)
 110 {
 111         netif_stop_queue(netdev);
 112         return 0;
 113 }
 114
 115 static void internal_dev_getinfo(struct net_device *netdev,
 116                                  struct ethtool_drvinfo *info)
 117 {
 118         strcpy(info->driver, "openvswitch");
 119 }
 120
 121 static const struct ethtool_ops internal_dev_ethtool_ops = {
 122         .get_drvinfo    = internal_dev_getinfo,
 123         .get_link       = ethtool_op_get_link,
 124 #if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,39)
 125         .get_sg         = ethtool_op_get_sg,
 126         .set_sg         = ethtool_op_set_sg,
 127         .get_tx_csum    = ethtool_op_get_tx_csum,
 128         .set_tx_csum    = ethtool_op_set_tx_hw_csum,
 129         .get_tso        = ethtool_op_get_tso,
 130         .set_tso        = ethtool_op_set_tso,
 131 #endif
 132 };
 133
 134 static int internal_dev_change_mtu(struct net_device *netdev, int new_mtu)
 135 {
 136         if (new_mtu < 68)
 137                 return -EINVAL;
 138
 139         netdev->mtu = new_mtu;
 140         return 0;
 141 }
 142
 143 static void internal_dev_destructor(struct net_device *dev)
 144 {
 145         struct vport *vport = ovs_internal_dev_get_vport(dev);
 146
 147         ovs_vport_free(vport);
 148         free_netdev(dev);
 149 }
 150
 151 #ifdef HAVE_NET_DEVICE_OPS
 152 static const struct net_device_ops internal_dev_netdev_ops = {
 153         .ndo_open = internal_dev_open,
 154         .ndo_stop = internal_dev_stop,
 155         .ndo_start_xmit = internal_dev_xmit,
 156         .ndo_set_mac_address = eth_mac_addr,
 157         .ndo_change_mtu = internal_dev_change_mtu,
 158 #if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,36)
 159         .ndo_get_stats64 = internal_dev_get_stats,
 160 #else
 161         .ndo_get_stats = internal_dev_sys_stats,
 162 #endif
 163 };
 164 #endif
 165
 166 static void do_setup(struct net_device *netdev)
 167 {
 168         ether_setup(netdev);
 169
 170 #ifdef HAVE_NET_DEVICE_OPS
 171         netdev->netdev_ops = &internal_dev_netdev_ops;
 172 #else
 173         netdev->get_stats = internal_dev_sys_stats;
 174         netdev->hard_start_xmit = internal_dev_xmit;
 175         netdev->open = internal_dev_open;
 176         netdev->stop = internal_dev_stop;
 177         netdev->set_mac_address = internal_dev_mac_addr;
 178         netdev->change_mtu = internal_dev_change_mtu;
 179 #endif
 180
 181         netdev->priv_flags &= ~IFF_TX_SKB_SHARING;
 182         netdev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
 183         netdev->destructor = internal_dev_destructor;
 184         SET_ETHTOOL_OPS(netdev, &internal_dev_ethtool_ops);
 185         netdev->tx_queue_len = 0;
 186
 187         netdev->features = NETIF_F_LLTX | NETIF_F_SG | NETIF_F_FRAGLIST |
 188                            NETIF_F_HIGHDMA | NETIF_F_HW_CSUM | NETIF_F_TSO;
 189
 190 #if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,27)
 191         netdev->vlan_features = netdev->features;
 192         netdev->features |= NETIF_F_HW_VLAN_TX;
 193 #endif
 194
 195 #if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,39)
 196         netdev->hw_features = netdev->features & ~NETIF_F_LLTX;
 197 #endif
 198         eth_hw_addr_random(netdev);
 199 }
 200
 201 static struct vport *internal_dev_create(const struct vport_parms *parms)
 202 {
 203         struct vport *vport;
 204         struct netdev_vport *netdev_vport;
 205         struct internal_dev *internal_dev;
 206         int err;
 207
 208         vport = ovs_vport_alloc(sizeof(struct netdev_vport),
 209                                 &ovs_internal_vport_ops, parms);
 210         if (IS_ERR(vport)) {
 211                 err = PTR_ERR(vport);
 212                 goto error;
 213         }
 214
 215         netdev_vport = netdev_vport_priv(vport);
 216
 217         netdev_vport->dev = alloc_netdev(sizeof(struct internal_dev),
 218                                          parms->name, do_setup);
 219         if (!netdev_vport->dev) {
 220                 err = -ENOMEM;
 221                 goto error_free_vport;
 222         }
 223
 224         dev_net_set(netdev_vport->dev, ovs_dp_get_net(vport->dp));
 225         internal_dev = internal_dev_priv(netdev_vport->dev);
 226         internal_dev->vport = vport;
 227
 228         /* Restrict bridge port to current netns. */
 229         if (vport->port_no == OVSP_LOCAL)
 230                 netdev_vport->dev->features |= NETIF_F_NETNS_LOCAL;
 231
 232         err = register_netdevice(netdev_vport->dev);
 233         if (err)
 234                 goto error_free_netdev;
 235
 236         dev_set_promiscuity(netdev_vport->dev, 1);
 237         netif_start_queue(netdev_vport->dev);
 238
 239         return vport;
 240
 241 error_free_netdev:
 242         free_netdev(netdev_vport->dev);
 243 error_free_vport:
 244         ovs_vport_free(vport);
 245 error:
 246         return ERR_PTR(err);
 247 }
 248
 249 static void internal_dev_destroy(struct vport *vport)
 250 {
 251         struct netdev_vport *netdev_vport = netdev_vport_priv(vport);
 252
 253         netif_stop_queue(netdev_vport->dev);
 254         dev_set_promiscuity(netdev_vport->dev, -1);
 255
 256         /* unregister_netdevice() waits for an RCU grace period. */
 257         unregister_netdevice(netdev_vport->dev);
 258 }
 259
 260 static int internal_dev_recv(struct vport *vport, struct sk_buff *skb)
 261 {
 262         struct net_device *netdev = netdev_vport_priv(vport)->dev;
 263         int len;
 264
 265 #if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,37)
 266         if (unlikely(vlan_deaccel_tag(skb)))
 267                 return 0;
 268 #endif
 269
 270         len = skb->len;
 271
 272         skb_dst_drop(skb);
 273         nf_reset(skb);
 274         secpath_reset(skb);
 275
 276         skb->dev = netdev;
 277         skb->pkt_type = PACKET_HOST;
 278         skb->protocol = eth_type_trans(skb, netdev);
 279         forward_ip_summed(skb, false);
 280
 281         netif_rx(skb);
 282
 283 #if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,29)
 284         netdev->last_rx = jiffies;
 285 #endif
 286
 287         return len;
 288 }
 289
 290 const struct vport_ops ovs_internal_vport_ops = {
 291         .type           = OVS_VPORT_TYPE_INTERNAL,
 292         .flags          = VPORT_F_REQUIRED | VPORT_F_FLOW,
 293         .create         = internal_dev_create,
 294         .destroy        = internal_dev_destroy,
 295         .set_addr       = ovs_netdev_set_addr,
 296         .get_name       = ovs_netdev_get_name,
 297         .get_addr       = ovs_netdev_get_addr,
 298         .get_ifindex    = ovs_netdev_get_ifindex,
 299         .send           = internal_dev_recv,
 300 };
 301
 302 int ovs_is_internal_dev(const struct net_device *netdev)
 303 {
 304 #ifdef HAVE_NET_DEVICE_OPS
 305         return netdev->netdev_ops == &internal_dev_netdev_ops;
 306 #else
 307         return netdev->open == internal_dev_open;
 308 #endif
 309 }
 310
 311 struct vport *ovs_internal_dev_get_vport(struct net_device *netdev)
 312 {
 313         if (!ovs_is_internal_dev(netdev))
 314                 return NULL;
 315
 316         return internal_dev_priv(netdev)->vport;
 317 }