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