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