datapath/vport-netdev.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 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
  20
  21 #include <linux/if_arp.h>
  22 #include <linux/if_bridge.h>
  23 #include <linux/if_vlan.h>
  24 #include <linux/kernel.h>
  25 #include <linux/llc.h>
  26 #include <linux/rtnetlink.h>
  27 #include <linux/skbuff.h>
  28 #include <linux/openvswitch.h>
  29 #include <linux/netdevice.h>
  30
  31 #include <net/llc.h>
  32
  33 #include "datapath.h"
  34 #include "vlan.h"
  35 #include "vport-internal_dev.h"
  36 #include "vport-netdev.h"
  37
  38 static struct vport_ops ovs_netdev_vport_ops;
  39 static void netdev_port_receive(struct vport *vport, struct sk_buff *skb);
  40
  41 #if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,39)
  42 /* Called with rcu_read_lock and bottom-halves disabled. */
  43 static rx_handler_result_t netdev_frame_hook(struct sk_buff **pskb)
  44 {
  45         struct sk_buff *skb = *pskb;
  46         struct vport *vport;
  47
  48         if (unlikely(skb->pkt_type == PACKET_LOOPBACK))
  49                 return RX_HANDLER_PASS;
  50
  51         vport = ovs_netdev_get_vport(skb->dev);
  52
  53         netdev_port_receive(vport, skb);
  54
  55         return RX_HANDLER_CONSUMED;
  56 }
  57 #elif LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,36) || \
  58       defined HAVE_RHEL_OVS_HOOK
  59 /* Called with rcu_read_lock and bottom-halves disabled. */
  60 static struct sk_buff *netdev_frame_hook(struct sk_buff *skb)
  61 {
  62         struct vport *vport;
  63
  64         if (unlikely(skb->pkt_type == PACKET_LOOPBACK))
  65                 return skb;
  66
  67         vport = ovs_netdev_get_vport(skb->dev);
  68
  69         netdev_port_receive(vport, skb);
  70
  71         return NULL;
  72 }
  73 #elif LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,32)
  74 /*
  75  * Used as br_handle_frame_hook.  (Cannot run bridge at the same time, even on
  76  * different set of devices!)
  77  */
  78 /* Called with rcu_read_lock and bottom-halves disabled. */
  79 static struct sk_buff *netdev_frame_hook(struct net_bridge_port *p,
  80                                          struct sk_buff *skb)
  81 {
  82         netdev_port_receive((struct vport *)p, skb);
  83         return NULL;
  84 }
  85 #else
  86 #error
  87 #endif
  88
  89 static struct net_device *get_dpdev(const struct datapath *dp)
  90 {
  91         struct vport *local;
  92
  93         local = ovs_vport_ovsl(dp, OVSP_LOCAL);
  94         BUG_ON(!local);
  95         return netdev_vport_priv(local)->dev;
  96 }
  97
  98 static struct vport *netdev_create(const struct vport_parms *parms)
  99 {
 100         struct vport *vport;
 101         struct netdev_vport *netdev_vport;
 102         int err;
 103
 104         vport = ovs_vport_alloc(sizeof(struct netdev_vport),
 105                                 &ovs_netdev_vport_ops, parms);
 106         if (IS_ERR(vport)) {
 107                 err = PTR_ERR(vport);
 108                 goto error;
 109         }
 110
 111         netdev_vport = netdev_vport_priv(vport);
 112
 113         netdev_vport->dev = dev_get_by_name(ovs_dp_get_net(vport->dp), parms->name);
 114         if (!netdev_vport->dev) {
 115                 err = -ENODEV;
 116                 goto error_free_vport;
 117         }
 118
 119         if (netdev_vport->dev->flags & IFF_LOOPBACK ||
 120             netdev_vport->dev->type != ARPHRD_ETHER ||
 121             ovs_is_internal_dev(netdev_vport->dev)) {
 122                 err = -EINVAL;
 123                 goto error_put;
 124         }
 125
 126         rtnl_lock();
 127         err = netdev_master_upper_dev_link(netdev_vport->dev,
 128                                            get_dpdev(vport->dp));
 129         if (err)
 130                 goto error_unlock;
 131
 132         err = netdev_rx_handler_register(netdev_vport->dev, netdev_frame_hook,
 133                                          vport);
 134         if (err)
 135                 goto error_master_upper_dev_unlink;
 136
 137         dev_set_promiscuity(netdev_vport->dev, 1);
 138         netdev_vport->dev->priv_flags |= IFF_OVS_DATAPATH;
 139         rtnl_unlock();
 140
 141         return vport;
 142
 143 error_master_upper_dev_unlink:
 144         netdev_upper_dev_unlink(netdev_vport->dev, get_dpdev(vport->dp));
 145 error_unlock:
 146         rtnl_unlock();
 147 error_put:
 148         dev_put(netdev_vport->dev);
 149 error_free_vport:
 150         ovs_vport_free(vport);
 151 error:
 152         return ERR_PTR(err);
 153 }
 154
 155 static void free_port_rcu(struct rcu_head *rcu)
 156 {
 157         struct netdev_vport *netdev_vport = container_of(rcu,
 158                                         struct netdev_vport, rcu);
 159
 160         dev_put(netdev_vport->dev);
 161         ovs_vport_free(vport_from_priv(netdev_vport));
 162 }
 163
 164 void ovs_netdev_detach_dev(struct vport *vport)
 165 {
 166         struct netdev_vport *netdev_vport = netdev_vport_priv(vport);
 167
 168         ASSERT_RTNL();
 169         netdev_vport->dev->priv_flags &= ~IFF_OVS_DATAPATH;
 170         netdev_rx_handler_unregister(netdev_vport->dev);
 171         netdev_upper_dev_unlink(netdev_vport->dev,
 172                                 netdev_master_upper_dev_get(netdev_vport->dev));
 173         dev_set_promiscuity(netdev_vport->dev, -1);
 174 }
 175
 176 static void netdev_destroy(struct vport *vport)
 177 {
 178         struct netdev_vport *netdev_vport = netdev_vport_priv(vport);
 179
 180         rtnl_lock();
 181         if (ovs_netdev_get_vport(netdev_vport->dev))
 182                 ovs_netdev_detach_dev(vport);
 183         rtnl_unlock();
 184
 185         call_rcu(&netdev_vport->rcu, free_port_rcu);
 186 }
 187
 188 const char *ovs_netdev_get_name(const struct vport *vport)
 189 {
 190         const struct netdev_vport *netdev_vport = netdev_vport_priv(vport);
 191         return netdev_vport->dev->name;
 192 }
 193
 194 /* Must be called with rcu_read_lock. */
 195 static void netdev_port_receive(struct vport *vport, struct sk_buff *skb)
 196 {
 197         if (unlikely(!vport))
 198                 goto error;
 199
 200         if (unlikely(skb_warn_if_lro(skb)))
 201                 goto error;
 202
 203         /* Make our own copy of the packet.  Otherwise we will mangle the
 204          * packet for anyone who came before us (e.g. tcpdump via AF_PACKET).
 205          * (No one comes after us, since we tell handle_bridge() that we took
 206          * the packet.)
 207          */
 208         skb = skb_share_check(skb, GFP_ATOMIC);
 209         if (unlikely(!skb))
 210                 return;
 211
 212         skb_push(skb, ETH_HLEN);
 213         ovs_skb_postpush_rcsum(skb, skb->data, ETH_HLEN);
 214
 215         ovs_vport_receive(vport, skb, NULL);
 216         return;
 217
 218 error:
 219         kfree_skb(skb);
 220 }
 221
 222 static unsigned int packet_length(const struct sk_buff *skb)
 223 {
 224         unsigned int length = skb->len - ETH_HLEN;
 225
 226         if (skb->protocol == htons(ETH_P_8021Q))
 227                 length -= VLAN_HLEN;
 228
 229         return length;
 230 }
 231
 232 static int netdev_send(struct vport *vport, struct sk_buff *skb)
 233 {
 234         struct netdev_vport *netdev_vport = netdev_vport_priv(vport);
 235         int mtu = netdev_vport->dev->mtu;
 236         int len;
 237
 238         if (unlikely(packet_length(skb) > mtu && !skb_is_gso(skb))) {
 239                 net_warn_ratelimited("%s: dropped over-mtu packet: %d > %d\n",
 240                                      netdev_vport->dev->name,
 241                                      packet_length(skb), mtu);
 242                 goto drop;
 243         }
 244
 245         skb->dev = netdev_vport->dev;
 246         len = skb->len;
 247         dev_queue_xmit(skb);
 248
 249         return len;
 250
 251 drop:
 252         kfree_skb(skb);
 253         return 0;
 254 }
 255
 256 /* Returns null if this device is not attached to a datapath. */
 257 struct vport *ovs_netdev_get_vport(struct net_device *dev)
 258 {
 259 #if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,36) || \
 260     defined HAVE_RHEL_OVS_HOOK
 261 #ifdef HAVE_OVS_DATAPATH
 262         if (likely(dev->priv_flags & IFF_OVS_DATAPATH))
 263 #else
 264         if (likely(rcu_access_pointer(dev->rx_handler) == netdev_frame_hook))
 265 #endif
 266 #ifdef HAVE_RHEL_OVS_HOOK
 267                 return (struct vport *)rcu_dereference_rtnl(dev->ax25_ptr);
 268 #else
 269                 return (struct vport *)rcu_dereference_rtnl(dev->rx_handler_data);
 270 #endif
 271         else
 272                 return NULL;
 273 #else
 274         return (struct vport *)rcu_dereference_rtnl(dev->br_port);
 275 #endif
 276 }
 277
 278 static struct vport_ops ovs_netdev_vport_ops = {
 279         .type           = OVS_VPORT_TYPE_NETDEV,
 280         .create         = netdev_create,
 281         .destroy        = netdev_destroy,
 282         .get_name       = ovs_netdev_get_name,
 283         .send           = netdev_send,
 284 };
 285
 286 int __init ovs_netdev_init(void)
 287 {
 288         return ovs_vport_ops_register(&ovs_netdev_vport_ops);
 289 }
 290
 291 void ovs_netdev_exit(void)
 292 {
 293         ovs_vport_ops_unregister(&ovs_netdev_vport_ops);
 294 }
 295
 296 #if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,36) && \
 297     !defined HAVE_RHEL_OVS_HOOK
 298 /*
 299  * Enforces, mutual exclusion with the Linux bridge module, by declaring and
 300  * exporting br_should_route_hook.  Because the bridge module also exports the
 301  * same symbol, the module loader will refuse to load both modules at the same
 302  * time (e.g. "bridge: exports duplicate symbol br_should_route_hook (owned by
 303  * openvswitch)").
 304  *
 305  * Before Linux 2.6.36, Open vSwitch cannot safely coexist with the Linux
 306  * bridge module, so openvswitch uses this macro in those versions.  In
 307  * Linux 2.6.36 and later, Open vSwitch can coexist with the bridge module.
 308  *
 309  * The use of "typeof" here avoids the need to track changes in the type of
 310  * br_should_route_hook over various kernel versions.
 311  */
 312 typeof(br_should_route_hook) br_should_route_hook;
 313 EXPORT_SYMBOL(br_should_route_hook);
 314 #endif