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
  27 #include <net/dst.h>
  28 #include <net/xfrm.h>
  29 #include <net/rtnetlink.h>
  30
  31 #include "datapath.h"
  32 #include "vport-internal_dev.h"
  33 #include "vport-netdev.h"
  34
  35 struct internal_dev {
  36         struct vport *vport;
  37 };
  38
  39 static struct vport_ops ovs_internal_vport_ops;
  40
  41 static struct internal_dev *internal_dev_priv(struct net_device *netdev)
  42 {
  43         return netdev_priv(netdev);
  44 }
  45
  46 /* Called with rcu_read_lock_bh. */
  47 static int 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 0;
  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 #ifndef HAVE_NET_DEVICE_WITH_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         free_netdev(dev);
 118 }
 119
 120 static void
 121 internal_get_stats(struct net_device *dev, struct rtnl_link_stats64 *stats)
 122 {
 123         int i;
 124
 125         memset(stats, 0, sizeof(*stats));
 126         stats->rx_errors  = dev->stats.rx_errors;
 127         stats->tx_errors  = dev->stats.tx_errors;
 128         stats->tx_dropped = dev->stats.tx_dropped;
 129         stats->rx_dropped = dev->stats.rx_dropped;
 130
 131         for_each_possible_cpu(i) {
 132                 const struct pcpu_sw_netstats *percpu_stats;
 133                 struct pcpu_sw_netstats local_stats;
 134                 unsigned int start;
 135
 136                 percpu_stats = per_cpu_ptr(dev->tstats, i);
 137
 138                 do {
 139                         start = u64_stats_fetch_begin_irq(&percpu_stats->syncp);
 140                         local_stats = *percpu_stats;
 141                 } while (u64_stats_fetch_retry_irq(&percpu_stats->syncp, start));
 142
 143                 stats->rx_bytes         += local_stats.rx_bytes;
 144                 stats->rx_packets       += local_stats.rx_packets;
 145                 stats->tx_bytes         += local_stats.tx_bytes;
 146                 stats->tx_packets       += local_stats.tx_packets;
 147         }
 148 }
 149
 150 #ifdef HAVE_IFF_PHONY_HEADROOM
 151 static void internal_set_rx_headroom(struct net_device *dev, int new_hr)
 152 {
 153         dev->needed_headroom = new_hr < 0 ? 0 : new_hr;
 154 }
 155 #endif
 156
 157 static const struct net_device_ops internal_dev_netdev_ops = {
 158         .ndo_open = internal_dev_open,
 159         .ndo_stop = internal_dev_stop,
 160         .ndo_start_xmit = internal_dev_xmit,
 161         .ndo_set_mac_address = eth_mac_addr,
 162 #ifndef HAVE_NET_DEVICE_WITH_MAX_MTU
 163         .ndo_change_mtu = internal_dev_change_mtu,
 164 #endif
 165         .ndo_get_stats64 = (void *)internal_get_stats,
 166 #ifdef HAVE_IFF_PHONY_HEADROOM
 167 #ifndef HAVE_NET_DEVICE_OPS_WITH_EXTENDED
 168         .ndo_set_rx_headroom = internal_set_rx_headroom,
 169 #else
 170         .extended.ndo_set_rx_headroom = internal_set_rx_headroom,
 171 #endif
 172 #endif
 173 };
 174
 175 static struct rtnl_link_ops internal_dev_link_ops __read_mostly = {
 176         .kind = "openvswitch",
 177 };
 178
 179 static void do_setup(struct net_device *netdev)
 180 {
 181         ether_setup(netdev);
 182
 183 #ifdef HAVE_NET_DEVICE_WITH_MAX_MTU
 184         netdev->max_mtu = ETH_MAX_MTU;
 185 #endif
 186         netdev->netdev_ops = &internal_dev_netdev_ops;
 187
 188         netdev->priv_flags &= ~IFF_TX_SKB_SHARING;
 189         netdev->priv_flags |= IFF_LIVE_ADDR_CHANGE | IFF_OPENVSWITCH |
 190                               IFF_PHONY_HEADROOM | IFF_NO_QUEUE;
 191         netdev->destructor = internal_dev_destructor;
 192         netdev->ethtool_ops = &internal_dev_ethtool_ops;
 193         netdev->rtnl_link_ops = &internal_dev_link_ops;
 194
 195 #ifndef HAVE_IFF_NO_QUEUE
 196         netdev->tx_queue_len = 0;
 197 #endif
 198
 199         netdev->features = NETIF_F_LLTX | NETIF_F_SG | NETIF_F_FRAGLIST |
 200                            NETIF_F_HIGHDMA | NETIF_F_HW_CSUM |
 201                            NETIF_F_GSO_SOFTWARE | NETIF_F_GSO_ENCAP_ALL;
 202
 203         netdev->vlan_features = netdev->features;
 204         netdev->hw_enc_features = netdev->features;
 205         netdev->features |= NETIF_F_HW_VLAN_CTAG_TX | NETIF_F_HW_VLAN_STAG_TX;
 206         netdev->hw_features = netdev->features & ~NETIF_F_LLTX;
 207
 208         eth_hw_addr_random(netdev);
 209 }
 210
 211 static struct vport *internal_dev_create(const struct vport_parms *parms)
 212 {
 213         struct vport *vport;
 214         struct internal_dev *internal_dev;
 215         int err;
 216
 217         vport = ovs_vport_alloc(0, &ovs_internal_vport_ops, parms);
 218         if (IS_ERR(vport)) {
 219                 err = PTR_ERR(vport);
 220                 goto error;
 221         }
 222
 223         vport->dev = alloc_netdev(sizeof(struct internal_dev),
 224                                   parms->name, NET_NAME_USER, do_setup);
 225         if (!vport->dev) {
 226                 err = -ENOMEM;
 227                 goto error_free_vport;
 228         }
 229         vport->dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
 230         if (!vport->dev->tstats) {
 231                 err = -ENOMEM;
 232                 goto error_free_netdev;
 233         }
 234
 235 #ifdef HAVE_IFF_PHONY_HEADROOM
 236         vport->dev->needed_headroom = vport->dp->max_headroom;
 237 #endif
 238         dev_net_set(vport->dev, ovs_dp_get_net(vport->dp));
 239         internal_dev = internal_dev_priv(vport->dev);
 240         internal_dev->vport = vport;
 241
 242         /* Restrict bridge port to current netns. */
 243         if (vport->port_no == OVSP_LOCAL)
 244                 vport->dev->features |= NETIF_F_NETNS_LOCAL;
 245
 246         rtnl_lock();
 247         err = register_netdevice(vport->dev);
 248         if (err)
 249                 goto error_unlock;
 250
 251         dev_set_promiscuity(vport->dev, 1);
 252         rtnl_unlock();
 253         netif_start_queue(vport->dev);
 254
 255         return vport;
 256
 257 error_unlock:
 258         rtnl_unlock();
 259         free_percpu(vport->dev->tstats);
 260 error_free_netdev:
 261         free_netdev(vport->dev);
 262 error_free_vport:
 263         ovs_vport_free(vport);
 264 error:
 265         return ERR_PTR(err);
 266 }
 267
 268 static void internal_dev_destroy(struct vport *vport)
 269 {
 270         netif_stop_queue(vport->dev);
 271         rtnl_lock();
 272         dev_set_promiscuity(vport->dev, -1);
 273
 274         /* unregister_netdevice() waits for an RCU grace period. */
 275         unregister_netdevice(vport->dev);
 276         free_percpu(vport->dev->tstats);
 277         rtnl_unlock();
 278 }
 279
 280 static netdev_tx_t internal_dev_recv(struct sk_buff *skb)
 281 {
 282         struct net_device *netdev = skb->dev;
 283         struct pcpu_sw_netstats *stats;
 284
 285         if (unlikely(!(netdev->flags & IFF_UP))) {
 286                 kfree_skb(skb);
 287                 netdev->stats.rx_dropped++;
 288                 return NETDEV_TX_OK;
 289         }
 290
 291         skb_dst_drop(skb);
 292         nf_reset(skb);
 293         secpath_reset(skb);
 294
 295         skb->pkt_type = PACKET_HOST;
 296         skb->protocol = eth_type_trans(skb, netdev);
 297         skb_postpull_rcsum(skb, eth_hdr(skb), ETH_HLEN);
 298
 299         stats = this_cpu_ptr(netdev->tstats);
 300         u64_stats_update_begin(&stats->syncp);
 301         stats->rx_packets++;
 302         stats->rx_bytes += skb->len;
 303         u64_stats_update_end(&stats->syncp);
 304
 305         netif_rx(skb);
 306         return NETDEV_TX_OK;
 307 }
 308
 309 static struct vport_ops ovs_internal_vport_ops = {
 310         .type           = OVS_VPORT_TYPE_INTERNAL,
 311         .create         = internal_dev_create,
 312         .destroy        = internal_dev_destroy,
 313         .send           = internal_dev_recv,
 314 };
 315
 316 int ovs_is_internal_dev(const struct net_device *netdev)
 317 {
 318         return netdev->netdev_ops == &internal_dev_netdev_ops;
 319 }
 320
 321 struct vport *ovs_internal_dev_get_vport(struct net_device *netdev)
 322 {
 323         if (!ovs_is_internal_dev(netdev))
 324                 return NULL;
 325
 326         return internal_dev_priv(netdev)->vport;
 327 }
 328
 329 int ovs_internal_dev_rtnl_link_register(void)
 330 {
 331         int err;
 332
 333         err = rtnl_link_register(&internal_dev_link_ops);
 334         if (err < 0)
 335                 return err;
 336
 337         err = ovs_vport_ops_register(&ovs_internal_vport_ops);
 338         if (err < 0)
 339                 rtnl_link_unregister(&internal_dev_link_ops);
 340
 341         return err;
 342 }
 343
 344 void ovs_internal_dev_rtnl_link_unregister(void)
 345 {
 346         ovs_vport_ops_unregister(&ovs_internal_vport_ops);
 347         rtnl_link_unregister(&internal_dev_link_ops);
 348 }