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 "checksum.h"
  29 #include "datapath.h"
  30 #include "vlan.h"
  31 #include "vport-generic.h"
  32 #include "vport-internal_dev.h"
  33 #include "vport-netdev.h"
  34
  35 #if LINUX_VERSION_CODE >= KERNEL_VERSION(3,1,0)
  36 #define HAVE_NET_DEVICE_OPS
  37 #endif
  38
  39 struct internal_dev {
  40         struct vport *vport;
  41 #if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,22)
  42         struct net_device_stats stats;
  43 #endif
  44 };
  45
  46 static struct internal_dev *internal_dev_priv(struct net_device *netdev)
  47 {
  48         return netdev_priv(netdev);
  49 }
  50
  51 /* This function is only called by the kernel network layer.*/
  52 #if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,36)
  53 static struct rtnl_link_stats64 *internal_dev_get_stats(struct net_device *netdev,
  54                                                         struct rtnl_link_stats64 *stats)
  55 {
  56 #else
  57 static struct net_device_stats *internal_dev_sys_stats(struct net_device *netdev)
  58 {
  59 #if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,22)
  60         struct net_device_stats *stats = &internal_dev_priv(netdev)->stats;
  61 #else
  62         struct net_device_stats *stats = &netdev->stats;
  63 #endif
  64 #endif
  65         struct vport *vport = ovs_internal_dev_get_vport(netdev);
  66         struct ovs_vport_stats vport_stats;
  67
  68         ovs_vport_get_stats(vport, &vport_stats);
  69
  70         /* The tx and rx stats need to be swapped because the
  71          * switch and host OS have opposite perspectives. */
  72         stats->rx_packets       = vport_stats.tx_packets;
  73         stats->tx_packets       = vport_stats.rx_packets;
  74         stats->rx_bytes         = vport_stats.tx_bytes;
  75         stats->tx_bytes         = vport_stats.rx_bytes;
  76         stats->rx_errors        = vport_stats.tx_errors;
  77         stats->tx_errors        = vport_stats.rx_errors;
  78         stats->rx_dropped       = vport_stats.tx_dropped;
  79         stats->tx_dropped       = vport_stats.rx_dropped;
  80
  81         return stats;
  82 }
  83
  84 static int internal_dev_mac_addr(struct net_device *dev, void *p)
  85 {
  86         struct sockaddr *addr = p;
  87
  88         if (!is_valid_ether_addr(addr->sa_data))
  89                 return -EADDRNOTAVAIL;
  90 #ifdef NET_ADDR_RANDOM
  91         dev->addr_assign_type &= ~NET_ADDR_RANDOM;
  92 #endif
  93         memcpy(dev->dev_addr, addr->sa_data, dev->addr_len);
  94         return 0;
  95 }
  96
  97 /* Called with rcu_read_lock_bh. */
  98 static int internal_dev_xmit(struct sk_buff *skb, struct net_device *netdev)
  99 {
 100         if (unlikely(compute_ip_summed(skb, true))) {
 101                 kfree_skb(skb);
 102                 return 0;
 103         }
 104
 105         vlan_copy_skb_tci(skb);
 106         OVS_CB(skb)->flow = NULL;
 107
 108         rcu_read_lock();
 109         ovs_vport_receive(internal_dev_priv(netdev)->vport, skb);
 110         rcu_read_unlock();
 111         return 0;
 112 }
 113
 114 static int internal_dev_open(struct net_device *netdev)
 115 {
 116         netif_start_queue(netdev);
 117         return 0;
 118 }
 119
 120 static int internal_dev_stop(struct net_device *netdev)
 121 {
 122         netif_stop_queue(netdev);
 123         return 0;
 124 }
 125
 126 static void internal_dev_getinfo(struct net_device *netdev,
 127                                  struct ethtool_drvinfo *info)
 128 {
 129         strcpy(info->driver, "openvswitch");
 130 }
 131
 132 static const struct ethtool_ops internal_dev_ethtool_ops = {
 133         .get_drvinfo    = internal_dev_getinfo,
 134         .get_link       = ethtool_op_get_link,
 135 #if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,39)
 136         .get_sg         = ethtool_op_get_sg,
 137         .set_sg         = ethtool_op_set_sg,
 138         .get_tx_csum    = ethtool_op_get_tx_csum,
 139         .set_tx_csum    = ethtool_op_set_tx_hw_csum,
 140         .get_tso        = ethtool_op_get_tso,
 141         .set_tso        = ethtool_op_set_tso,
 142 #endif
 143 };
 144
 145 static int internal_dev_change_mtu(struct net_device *netdev, int new_mtu)
 146 {
 147         if (new_mtu < 68)
 148                 return -EINVAL;
 149
 150         netdev->mtu = new_mtu;
 151         return 0;
 152 }
 153
 154 static int internal_dev_do_ioctl(struct net_device *dev,
 155                                  struct ifreq *ifr, int cmd)
 156 {
 157         if (ovs_dp_ioctl_hook)
 158                 return ovs_dp_ioctl_hook(dev, ifr, cmd);
 159
 160         return -EOPNOTSUPP;
 161 }
 162
 163 static void internal_dev_destructor(struct net_device *dev)
 164 {
 165         struct vport *vport = ovs_internal_dev_get_vport(dev);
 166
 167         ovs_vport_free(vport);
 168         free_netdev(dev);
 169 }
 170
 171 #ifdef HAVE_NET_DEVICE_OPS
 172 static const struct net_device_ops internal_dev_netdev_ops = {
 173         .ndo_open = internal_dev_open,
 174         .ndo_stop = internal_dev_stop,
 175         .ndo_start_xmit = internal_dev_xmit,
 176         .ndo_set_mac_address = internal_dev_mac_addr,
 177         .ndo_do_ioctl = internal_dev_do_ioctl,
 178         .ndo_change_mtu = internal_dev_change_mtu,
 179 #if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,36)
 180         .ndo_get_stats64 = internal_dev_get_stats,
 181 #else
 182         .ndo_get_stats = internal_dev_sys_stats,
 183 #endif
 184 };
 185 #endif
 186
 187 static void do_setup(struct net_device *netdev)
 188 {
 189         ether_setup(netdev);
 190
 191 #ifdef HAVE_NET_DEVICE_OPS
 192         netdev->netdev_ops = &internal_dev_netdev_ops;
 193 #else
 194         netdev->do_ioctl = internal_dev_do_ioctl;
 195         netdev->get_stats = internal_dev_sys_stats;
 196         netdev->hard_start_xmit = internal_dev_xmit;
 197         netdev->open = internal_dev_open;
 198         netdev->stop = internal_dev_stop;
 199         netdev->set_mac_address = internal_dev_mac_addr;
 200         netdev->change_mtu = internal_dev_change_mtu;
 201 #endif
 202
 203         netdev->priv_flags &= ~IFF_TX_SKB_SHARING;
 204         netdev->destructor = internal_dev_destructor;
 205         SET_ETHTOOL_OPS(netdev, &internal_dev_ethtool_ops);
 206         netdev->tx_queue_len = 0;
 207
 208         netdev->features = NETIF_F_LLTX | NETIF_F_SG | NETIF_F_FRAGLIST |
 209                            NETIF_F_HIGHDMA | NETIF_F_HW_CSUM | NETIF_F_TSO;
 210
 211 #if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,27)
 212         netdev->vlan_features = netdev->features;
 213         netdev->features |= NETIF_F_HW_VLAN_TX;
 214 #endif
 215
 216 #if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,39)
 217         netdev->hw_features = netdev->features & ~NETIF_F_LLTX;
 218 #endif
 219         eth_hw_addr_random(netdev);
 220 }
 221
 222 static struct vport *internal_dev_create(const struct vport_parms *parms)
 223 {
 224         struct vport *vport;
 225         struct netdev_vport *netdev_vport;
 226         struct internal_dev *internal_dev;
 227         int err;
 228
 229         vport = ovs_vport_alloc(sizeof(struct netdev_vport),
 230                                 &ovs_internal_vport_ops, parms);
 231         if (IS_ERR(vport)) {
 232                 err = PTR_ERR(vport);
 233                 goto error;
 234         }
 235
 236         netdev_vport = netdev_vport_priv(vport);
 237
 238         netdev_vport->dev = alloc_netdev(sizeof(struct internal_dev),
 239                                          parms->name, do_setup);
 240         if (!netdev_vport->dev) {
 241                 err = -ENOMEM;
 242                 goto error_free_vport;
 243         }
 244
 245         dev_net_set(netdev_vport->dev, ovs_dp_get_net(vport->dp));
 246         internal_dev = internal_dev_priv(netdev_vport->dev);
 247         internal_dev->vport = vport;
 248
 249         /* Restrict bridge port to current netns. */
 250         if (vport->port_no == OVSP_LOCAL)
 251                 netdev_vport->dev->features |= NETIF_F_NETNS_LOCAL;
 252
 253         err = register_netdevice(netdev_vport->dev);
 254         if (err)
 255                 goto error_free_netdev;
 256
 257         dev_set_promiscuity(netdev_vport->dev, 1);
 258         netif_start_queue(netdev_vport->dev);
 259
 260         return vport;
 261
 262 error_free_netdev:
 263         free_netdev(netdev_vport->dev);
 264 error_free_vport:
 265         ovs_vport_free(vport);
 266 error:
 267         return ERR_PTR(err);
 268 }
 269
 270 static void internal_dev_destroy(struct vport *vport)
 271 {
 272         struct netdev_vport *netdev_vport = netdev_vport_priv(vport);
 273
 274         netif_stop_queue(netdev_vport->dev);
 275         dev_set_promiscuity(netdev_vport->dev, -1);
 276
 277         /* unregister_netdevice() waits for an RCU grace period. */
 278         unregister_netdevice(netdev_vport->dev);
 279 }
 280
 281 static int internal_dev_recv(struct vport *vport, struct sk_buff *skb)
 282 {
 283         struct net_device *netdev = netdev_vport_priv(vport)->dev;
 284         int len;
 285
 286 #if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,37)
 287         if (unlikely(vlan_deaccel_tag(skb)))
 288                 return 0;
 289 #endif
 290
 291         len = skb->len;
 292         skb->dev = netdev;
 293         skb->pkt_type = PACKET_HOST;
 294         skb->protocol = eth_type_trans(skb, netdev);
 295         forward_ip_summed(skb, false);
 296
 297         netif_rx(skb);
 298
 299 #if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,29)
 300         netdev->last_rx = jiffies;
 301 #endif
 302
 303         return len;
 304 }
 305
 306 const struct vport_ops ovs_internal_vport_ops = {
 307         .type           = OVS_VPORT_TYPE_INTERNAL,
 308         .flags          = VPORT_F_REQUIRED | VPORT_F_FLOW,
 309         .create         = internal_dev_create,
 310         .destroy        = internal_dev_destroy,
 311         .set_addr       = ovs_netdev_set_addr,
 312         .get_name       = ovs_netdev_get_name,
 313         .get_addr       = ovs_netdev_get_addr,
 314         .get_kobj       = ovs_netdev_get_kobj,
 315         .get_dev_flags  = ovs_netdev_get_dev_flags,
 316         .is_running     = ovs_netdev_is_running,
 317         .get_operstate  = ovs_netdev_get_operstate,
 318         .get_ifindex    = ovs_netdev_get_ifindex,
 319         .get_mtu        = ovs_netdev_get_mtu,
 320         .send           = internal_dev_recv,
 321 };
 322
 323 int ovs_is_internal_dev(const struct net_device *netdev)
 324 {
 325 #ifdef HAVE_NET_DEVICE_OPS
 326         return netdev->netdev_ops == &internal_dev_netdev_ops;
 327 #else
 328         return netdev->open == internal_dev_open;
 329 #endif
 330 }
 331
 332 struct vport *ovs_internal_dev_get_vport(struct net_device *netdev)
 333 {
 334         if (!ovs_is_internal_dev(netdev))
 335                 return NULL;
 336
 337         return internal_dev_priv(netdev)->vport;
 338 }