datapath/vport-internal_dev.c

   1 /*
   2  * Copyright (c) 2007-2015 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/percpu.h>
  27 #include <linux/u64_stats_sync.h>
  28 #include <linux/netdev_features.h>
  29
  30 #include <net/dst.h>
  31 #include <net/xfrm.h>
  32 #include <net/rtnetlink.h>
  33
  34 #include "datapath.h"
  35 #include "vport-internal_dev.h"
  36 #include "vport-netdev.h"
  37
  38 struct internal_dev {
  39         struct vport *vport;
  40 };
  41
  42 static struct vport_ops ovs_internal_vport_ops;
  43
  44 static struct internal_dev *internal_dev_priv(struct net_device *netdev)
  45 {
  46         return netdev_priv(netdev);
  47 }
  48
  49 /* Called with rcu_read_lock_bh. */
  50 static int internal_dev_xmit(struct sk_buff *skb, struct net_device *netdev)
  51 {
  52         int len, err;
  53
  54         len = skb->len;
  55         rcu_read_lock();
  56         err = ovs_vport_receive(internal_dev_priv(netdev)->vport, skb, NULL);
  57         rcu_read_unlock();
  58
  59         if (likely(!err)) {
  60 #ifdef HAVE_DEV_TSTATS
  61                 struct pcpu_sw_netstats *tstats;
  62
  63                 tstats = this_cpu_ptr((struct pcpu_sw_netstats __percpu *)netdev->tstats);
  64
  65                 u64_stats_update_begin(&tstats->syncp);
  66                 tstats->tx_bytes += len;
  67                 tstats->tx_packets++;
  68                 u64_stats_update_end(&tstats->syncp);
  69 #endif
  70         } else {
  71                 netdev->stats.tx_errors++;
  72         }
  73         return 0;
  74 }
  75
  76 static int internal_dev_open(struct net_device *netdev)
  77 {
  78         netif_start_queue(netdev);
  79         return 0;
  80 }
  81
  82 static int internal_dev_stop(struct net_device *netdev)
  83 {
  84         netif_stop_queue(netdev);
  85         return 0;
  86 }
  87
  88 static void internal_dev_getinfo(struct net_device *netdev,
  89                                  struct ethtool_drvinfo *info)
  90 {
  91         strlcpy(info->driver, "openvswitch", sizeof(info->driver));
  92 }
  93
  94 static const struct ethtool_ops internal_dev_ethtool_ops = {
  95         .get_drvinfo    = internal_dev_getinfo,
  96         .get_link       = ethtool_op_get_link,
  97 #if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,39)
  98         .get_sg         = ethtool_op_get_sg,
  99         .set_sg         = ethtool_op_set_sg,
 100         .get_tx_csum    = ethtool_op_get_tx_csum,
 101         .set_tx_csum    = ethtool_op_set_tx_hw_csum,
 102         .get_tso        = ethtool_op_get_tso,
 103         .set_tso        = ethtool_op_set_tso,
 104 #endif
 105 };
 106
 107 static int internal_dev_change_mtu(struct net_device *netdev, int new_mtu)
 108 {
 109         if (new_mtu < 68)
 110                 return -EINVAL;
 111
 112         netdev->mtu = new_mtu;
 113         return 0;
 114 }
 115
 116 static void internal_dev_destructor(struct net_device *dev)
 117 {
 118         struct vport *vport = ovs_internal_dev_get_vport(dev);
 119
 120         ovs_vport_free(vport);
 121         free_netdev(dev);
 122 }
 123
 124 #ifdef HAVE_DEV_TSTATS
 125 static int internal_dev_init(struct net_device *dev)
 126 {
 127         dev->tstats = (typeof(dev->tstats)) netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
 128         if (!dev->tstats)
 129                 return -ENOMEM;
 130         return 0;
 131 }
 132
 133 static void internal_dev_uninit(struct net_device *dev)
 134 {
 135         free_percpu(dev->tstats);
 136 }
 137 #endif
 138
 139 static const struct net_device_ops internal_dev_netdev_ops = {
 140 #ifdef HAVE_DEV_TSTATS
 141         .ndo_init = internal_dev_init,
 142         .ndo_uninit = internal_dev_uninit,
 143         .ndo_get_stats64 = ip_tunnel_get_stats64,
 144 #endif
 145         .ndo_open = internal_dev_open,
 146         .ndo_stop = internal_dev_stop,
 147         .ndo_start_xmit = internal_dev_xmit,
 148         .ndo_set_mac_address = eth_mac_addr,
 149         .ndo_change_mtu = internal_dev_change_mtu,
 150 };
 151
 152 static struct rtnl_link_ops internal_dev_link_ops __read_mostly = {
 153         .kind = "openvswitch",
 154 };
 155
 156 static void do_setup(struct net_device *netdev)
 157 {
 158         ether_setup(netdev);
 159
 160         netdev->netdev_ops = &internal_dev_netdev_ops;
 161
 162         netdev->priv_flags &= ~IFF_TX_SKB_SHARING;
 163         netdev->priv_flags |= IFF_LIVE_ADDR_CHANGE | IFF_OPENVSWITCH;
 164         netdev->destructor = internal_dev_destructor;
 165         netdev->ethtool_ops = &internal_dev_ethtool_ops;
 166         netdev->rtnl_link_ops = &internal_dev_link_ops;
 167         netdev->tx_queue_len = 0;
 168
 169         netdev->features = NETIF_F_LLTX | NETIF_F_SG | NETIF_F_FRAGLIST |
 170                            NETIF_F_HIGHDMA | NETIF_F_HW_CSUM |
 171                            NETIF_F_GSO_SOFTWARE | NETIF_F_GSO_ENCAP_ALL;
 172
 173         netdev->vlan_features = netdev->features;
 174         netdev->features |= NETIF_F_HW_VLAN_CTAG_TX;
 175
 176 #if LINUX_VERSION_CODE >= KERNEL_VERSION(3,8,0)
 177         netdev->hw_enc_features = netdev->features;
 178 #endif
 179 #if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,39)
 180         netdev->hw_features = netdev->features & ~NETIF_F_LLTX;
 181 #endif
 182         eth_hw_addr_random(netdev);
 183 }
 184
 185 static struct vport *internal_dev_create(const struct vport_parms *parms)
 186 {
 187         struct vport *vport;
 188         struct internal_dev *internal_dev;
 189         int err;
 190
 191         vport = ovs_vport_alloc(0, &ovs_internal_vport_ops, parms);
 192         if (IS_ERR(vport)) {
 193                 err = PTR_ERR(vport);
 194                 goto error;
 195         }
 196
 197         vport->dev = alloc_netdev(sizeof(struct internal_dev),
 198                                   parms->name, NET_NAME_UNKNOWN, do_setup);
 199         if (!vport->dev) {
 200                 err = -ENOMEM;
 201                 goto error_free_vport;
 202         }
 203
 204         dev_net_set(vport->dev, ovs_dp_get_net(vport->dp));
 205         internal_dev = internal_dev_priv(vport->dev);
 206         internal_dev->vport = vport;
 207
 208         /* Restrict bridge port to current netns. */
 209         if (vport->port_no == OVSP_LOCAL)
 210                 vport->dev->features |= NETIF_F_NETNS_LOCAL;
 211
 212         rtnl_lock();
 213         err = register_netdevice(vport->dev);
 214         if (err)
 215                 goto error_free_netdev;
 216
 217         dev_set_promiscuity(vport->dev, 1);
 218         rtnl_unlock();
 219         netif_start_queue(vport->dev);
 220
 221         return vport;
 222
 223 error_free_netdev:
 224         rtnl_unlock();
 225         free_netdev(vport->dev);
 226 error_free_vport:
 227         ovs_vport_free(vport);
 228 error:
 229         return ERR_PTR(err);
 230 }
 231
 232 static void internal_dev_destroy(struct vport *vport)
 233 {
 234         netif_stop_queue(vport->dev);
 235         rtnl_lock();
 236         dev_set_promiscuity(vport->dev, -1);
 237
 238         /* unregister_netdevice() waits for an RCU grace period. */
 239         unregister_netdevice(vport->dev);
 240
 241         rtnl_unlock();
 242 }
 243
 244 static netdev_tx_t internal_dev_recv(struct sk_buff *skb)
 245 {
 246         struct net_device *netdev = skb->dev;
 247 #ifdef HAVE_DEV_TSTATS
 248         struct pcpu_sw_netstats *stats;
 249 #endif
 250
 251         if (unlikely(!(netdev->flags & IFF_UP))) {
 252                 kfree_skb(skb);
 253                 netdev->stats.rx_dropped++;
 254                 return NETDEV_TX_OK;
 255         }
 256
 257 #if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,37)
 258         if (skb_vlan_tag_present(skb)) {
 259                 if (unlikely(!vlan_insert_tag_set_proto(skb,
 260                                                         skb->vlan_proto,
 261                                                         skb_vlan_tag_get(skb))))
 262                         return NETDEV_TX_OK;
 263
 264                 if (skb->ip_summed == CHECKSUM_COMPLETE)
 265                         skb->csum = csum_add(skb->csum,
 266                                              csum_partial(skb->data + (2 * ETH_ALEN),
 267                                                           VLAN_HLEN, 0));
 268
 269                 vlan_set_tci(skb, 0);
 270         }
 271 #endif
 272
 273         skb_dst_drop(skb);
 274         nf_reset(skb);
 275         secpath_reset(skb);
 276
 277         skb->pkt_type = PACKET_HOST;
 278         skb->protocol = eth_type_trans(skb, netdev);
 279         skb_postpull_rcsum(skb, eth_hdr(skb), ETH_HLEN);
 280
 281 #ifdef HAVE_DEV_TSTATS
 282         stats = this_cpu_ptr((struct pcpu_sw_netstats __percpu *)netdev->tstats);
 283         u64_stats_update_begin(&stats->syncp);
 284         stats->rx_packets++;
 285         stats->rx_bytes += skb->len;
 286         u64_stats_update_end(&stats->syncp);
 287 #endif
 288
 289         netif_rx(skb);
 290         return NETDEV_TX_OK;
 291 }
 292
 293 static struct vport_ops ovs_internal_vport_ops = {
 294         .type           = OVS_VPORT_TYPE_INTERNAL,
 295         .create         = internal_dev_create,
 296         .destroy        = internal_dev_destroy,
 297         .send           = internal_dev_recv,
 298 };
 299
 300 int ovs_is_internal_dev(const struct net_device *netdev)
 301 {
 302         return netdev->netdev_ops == &internal_dev_netdev_ops;
 303 }
 304
 305 struct vport *ovs_internal_dev_get_vport(struct net_device *netdev)
 306 {
 307         if (!ovs_is_internal_dev(netdev))
 308                 return NULL;
 309
 310         return internal_dev_priv(netdev)->vport;
 311 }
 312
 313 int ovs_internal_dev_rtnl_link_register(void)
 314 {
 315         int err;
 316
 317         err = rtnl_link_register(&internal_dev_link_ops);
 318         if (err < 0)
 319                 return err;
 320
 321         err = ovs_vport_ops_register(&ovs_internal_vport_ops);
 322         if (err < 0)
 323                 rtnl_link_unregister(&internal_dev_link_ops);
 324
 325         return err;
 326 }
 327
 328 void ovs_internal_dev_rtnl_link_unregister(void)
 329 {
 330         ovs_vport_ops_unregister(&ovs_internal_vport_ops);
 331         rtnl_link_unregister(&internal_dev_link_ops);
 332 }