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/if_vlan.h>
  20 #include <linux/kernel.h>
  21 #include <linux/netdevice.h>
  22 #include <linux/etherdevice.h>
  23 #include <linux/ethtool.h>
  24 #include <linux/skbuff.h>
  25
  26 #include <net/dst.h>
  27 #include <net/xfrm.h>
  28 #include <net/rtnetlink.h>
  29
  30 #include "datapath.h"
  31 #include "vport-internal_dev.h"
  32 #include "vport-netdev.h"
  33
  34 struct internal_dev {
  35         struct vport *vport;
  36 };
  37
  38 static struct vport_ops ovs_internal_vport_ops;
  39
  40 static struct internal_dev *internal_dev_priv(struct net_device *netdev)
  41 {
  42         return netdev_priv(netdev);
  43 }
  44
  45 /* Called with rcu_read_lock_bh. */
  46 static netdev_tx_t
  47 internal_dev_xmit(struct sk_buff *skb, struct net_device *netdev)
  48 {
  49         int len, err;
  50
  51         len = skb->len;
  52         rcu_read_lock();
  53         err = ovs_vport_receive(internal_dev_priv(netdev)->vport, skb, NULL);
  54         rcu_read_unlock();
  55
  56         if (likely(!err)) {
  57                 struct pcpu_sw_netstats *tstats = this_cpu_ptr(netdev->tstats);
  58
  59                 u64_stats_update_begin(&tstats->syncp);
  60                 tstats->tx_bytes += len;
  61                 tstats->tx_packets++;
  62                 u64_stats_update_end(&tstats->syncp);
  63         } else {
  64                 netdev->stats.tx_errors++;
  65         }
  66         return NETDEV_TX_OK;
  67 }
  68
  69 static int internal_dev_open(struct net_device *netdev)
  70 {
  71         netif_start_queue(netdev);
  72         return 0;
  73 }
  74
  75 static int internal_dev_stop(struct net_device *netdev)
  76 {
  77         netif_stop_queue(netdev);
  78         return 0;
  79 }
  80
  81 static void internal_dev_getinfo(struct net_device *netdev,
  82                                  struct ethtool_drvinfo *info)
  83 {
  84         strlcpy(info->driver, "openvswitch", sizeof(info->driver));
  85 }
  86
  87 static const struct ethtool_ops internal_dev_ethtool_ops = {
  88         .get_drvinfo    = internal_dev_getinfo,
  89         .get_link       = ethtool_op_get_link,
  90 };
  91
  92 #if     !defined(HAVE_NET_DEVICE_WITH_MAX_MTU) && !defined(HAVE_RHEL7_MAX_MTU)
  93 static int internal_dev_change_mtu(struct net_device *dev, int new_mtu)
  94 {
  95         if (new_mtu < ETH_MIN_MTU) {
  96                 net_err_ratelimited("%s: Invalid MTU %d requested, hw min %d\n",
  97                                     dev->name, new_mtu, ETH_MIN_MTU);
  98                 return -EINVAL;
  99         }
 100
 101         if (new_mtu > ETH_MAX_MTU) {
 102                 net_err_ratelimited("%s: Invalid MTU %d requested, hw max %d\n",
 103                                     dev->name, new_mtu, ETH_MAX_MTU);
 104                 return -EINVAL;
 105         }
 106
 107         dev->mtu = new_mtu;
 108         return 0;
 109 }
 110 #endif
 111
 112 static void internal_dev_destructor(struct net_device *dev)
 113 {
 114         struct vport *vport = ovs_internal_dev_get_vport(dev);
 115
 116         ovs_vport_free(vport);
 117 #ifndef HAVE_NEEDS_FREE_NETDEV
 118         free_netdev(dev);
 119 #endif
 120 }
 121
 122 static void
 123 internal_get_stats(struct net_device *dev, struct rtnl_link_stats64 *stats)
 124 {
 125         int i;
 126
 127         memset(stats, 0, sizeof(*stats));
 128         stats->rx_errors  = dev->stats.rx_errors;
 129         stats->tx_errors  = dev->stats.tx_errors;
 130         stats->tx_dropped = dev->stats.tx_dropped;
 131         stats->rx_dropped = dev->stats.rx_dropped;
 132
 133         for_each_possible_cpu(i) {
 134                 const struct pcpu_sw_netstats *percpu_stats;
 135                 struct pcpu_sw_netstats local_stats;
 136                 unsigned int start;
 137
 138                 percpu_stats = per_cpu_ptr(dev->tstats, i);
 139
 140                 do {
 141                         start = u64_stats_fetch_begin_irq(&percpu_stats->syncp);
 142                         local_stats = *percpu_stats;
 143                 } while (u64_stats_fetch_retry_irq(&percpu_stats->syncp, start));
 144
 145                 stats->rx_bytes         += local_stats.rx_bytes;
 146                 stats->rx_packets       += local_stats.rx_packets;
 147                 stats->tx_bytes         += local_stats.tx_bytes;
 148                 stats->tx_packets       += local_stats.tx_packets;
 149         }
 150 }
 151
 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 #if     !defined(HAVE_NET_DEVICE_WITH_MAX_MTU) && !defined(HAVE_RHEL7_MAX_MTU)
 158         .ndo_change_mtu = internal_dev_change_mtu,
 159 #endif
 160         .ndo_get_stats64 = (void *)internal_get_stats,
 161 };
 162
 163 static struct rtnl_link_ops internal_dev_link_ops __read_mostly = {
 164         .kind = "openvswitch",
 165 };
 166
 167 static void do_setup(struct net_device *netdev)
 168 {
 169         ether_setup(netdev);
 170
 171 #ifdef HAVE_NET_DEVICE_WITH_MAX_MTU
 172         netdev->max_mtu = ETH_MAX_MTU;
 173 #elif defined(HAVE_RHEL7_MAX_MTU)
 174         netdev->extended->max_mtu = ETH_MAX_MTU;
 175 #endif
 176         netdev->netdev_ops = &internal_dev_netdev_ops;
 177
 178         netdev->priv_flags &= ~IFF_TX_SKB_SHARING;
 179         netdev->priv_flags |= IFF_LIVE_ADDR_CHANGE | IFF_OPENVSWITCH |
 180                               IFF_NO_QUEUE;
 181 #ifndef HAVE_NEEDS_FREE_NETDEV
 182         netdev->destructor = internal_dev_destructor;
 183 #else
 184         netdev->needs_free_netdev = true;
 185         netdev->priv_destructor = internal_dev_destructor;
 186 #endif /* HAVE_NEEDS_FREE_NETDEV */
 187         netdev->ethtool_ops = &internal_dev_ethtool_ops;
 188         netdev->rtnl_link_ops = &internal_dev_link_ops;
 189
 190 #ifndef HAVE_IFF_NO_QUEUE
 191         netdev->tx_queue_len = 0;
 192 #endif
 193
 194         netdev->features = NETIF_F_LLTX | NETIF_F_SG | NETIF_F_FRAGLIST |
 195                            NETIF_F_HIGHDMA | NETIF_F_HW_CSUM |
 196                            NETIF_F_GSO_SOFTWARE | NETIF_F_GSO_ENCAP_ALL;
 197
 198         netdev->vlan_features = netdev->features;
 199         netdev->hw_enc_features = netdev->features;
 200         netdev->features |= NETIF_F_HW_VLAN_CTAG_TX | NETIF_F_HW_VLAN_STAG_TX;
 201         netdev->hw_features = netdev->features & ~NETIF_F_LLTX;
 202
 203         eth_hw_addr_random(netdev);
 204 }
 205
 206 static struct vport *internal_dev_create(const struct vport_parms *parms)
 207 {
 208         struct vport *vport;
 209         struct internal_dev *internal_dev;
 210         int err;
 211
 212         vport = ovs_vport_alloc(0, &ovs_internal_vport_ops, parms);
 213         if (IS_ERR(vport)) {
 214                 err = PTR_ERR(vport);
 215                 goto error;
 216         }
 217
 218         vport->dev = alloc_netdev(sizeof(struct internal_dev),
 219                                   parms->name, NET_NAME_USER, do_setup);
 220         if (!vport->dev) {
 221                 err = -ENOMEM;
 222                 goto error_free_vport;
 223         }
 224         vport->dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
 225         if (!vport->dev->tstats) {
 226                 err = -ENOMEM;
 227                 goto error_free_netdev;
 228         }
 229
 230         dev_net_set(vport->dev, ovs_dp_get_net(vport->dp));
 231         internal_dev = internal_dev_priv(vport->dev);
 232         internal_dev->vport = vport;
 233
 234         /* Restrict bridge port to current netns. */
 235         if (vport->port_no == OVSP_LOCAL)
 236                 vport->dev->features |= NETIF_F_NETNS_LOCAL;
 237
 238         rtnl_lock();
 239         err = register_netdevice(vport->dev);
 240         if (err)
 241                 goto error_unlock;
 242
 243         dev_set_promiscuity(vport->dev, 1);
 244         rtnl_unlock();
 245         netif_start_queue(vport->dev);
 246
 247         return vport;
 248
 249 error_unlock:
 250         rtnl_unlock();
 251         free_percpu(vport->dev->tstats);
 252 error_free_netdev:
 253         free_netdev(vport->dev);
 254 error_free_vport:
 255         ovs_vport_free(vport);
 256 error:
 257         return ERR_PTR(err);
 258 }
 259
 260 static void internal_dev_destroy(struct vport *vport)
 261 {
 262         netif_stop_queue(vport->dev);
 263         rtnl_lock();
 264         dev_set_promiscuity(vport->dev, -1);
 265
 266         /* unregister_netdevice() waits for an RCU grace period. */
 267         unregister_netdevice(vport->dev);
 268         free_percpu(vport->dev->tstats);
 269         rtnl_unlock();
 270 }
 271
 272 static netdev_tx_t internal_dev_recv(struct sk_buff *skb)
 273 {
 274         struct net_device *netdev = skb->dev;
 275         struct pcpu_sw_netstats *stats;
 276
 277         if (unlikely(!(netdev->flags & IFF_UP))) {
 278                 kfree_skb(skb);
 279                 netdev->stats.rx_dropped++;
 280                 return NETDEV_TX_OK;
 281         }
 282
 283         skb_dst_drop(skb);
 284         nf_reset_ct(skb);
 285         secpath_reset(skb);
 286
 287         skb->pkt_type = PACKET_HOST;
 288         skb->protocol = eth_type_trans(skb, netdev);
 289         skb_postpull_rcsum(skb, eth_hdr(skb), ETH_HLEN);
 290
 291         stats = this_cpu_ptr(netdev->tstats);
 292         u64_stats_update_begin(&stats->syncp);
 293         stats->rx_packets++;
 294         stats->rx_bytes += skb->len;
 295         u64_stats_update_end(&stats->syncp);
 296
 297         netif_rx(skb);
 298         return NETDEV_TX_OK;
 299 }
 300
 301 static struct vport_ops ovs_internal_vport_ops = {
 302         .type           = OVS_VPORT_TYPE_INTERNAL,
 303         .create         = internal_dev_create,
 304         .destroy        = internal_dev_destroy,
 305         .send           = internal_dev_recv,
 306 };
 307
 308 int ovs_is_internal_dev(const struct net_device *netdev)
 309 {
 310         return netdev->netdev_ops == &internal_dev_netdev_ops;
 311 }
 312
 313 struct vport *ovs_internal_dev_get_vport(struct net_device *netdev)
 314 {
 315         if (!ovs_is_internal_dev(netdev))
 316                 return NULL;
 317
 318         return internal_dev_priv(netdev)->vport;
 319 }
 320
 321 int ovs_internal_dev_rtnl_link_register(void)
 322 {
 323         int err;
 324
 325         err = rtnl_link_register(&internal_dev_link_ops);
 326         if (err < 0)
 327                 return err;
 328
 329         err = ovs_vport_ops_register(&ovs_internal_vport_ops);
 330         if (err < 0)
 331                 rtnl_link_unregister(&internal_dev_link_ops);
 332
 333         return err;
 334 }
 335
 336 void ovs_internal_dev_rtnl_link_unregister(void)
 337 {
 338         ovs_vport_ops_unregister(&ovs_internal_vport_ops);
 339         rtnl_link_unregister(&internal_dev_link_ops);
 340 }