datapath/vport-netdev.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 #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
  30 #include <net/ip_tunnels.h>
  31 #include <net/rtnetlink.h>
  32
  33 #include "datapath.h"
  34 #include "gso.h"
  35 #include "vport.h"
  36 #include "vport-internal_dev.h"
  37 #include "vport-netdev.h"
  38
  39 static struct vport_ops ovs_netdev_vport_ops;
  40
  41 /* Must be called with rcu_read_lock. */
  42 void netdev_port_receive(struct sk_buff *skb, struct ip_tunnel_info *tun_info)
  43 {
  44         struct vport *vport;
  45
  46         vport = ovs_netdev_get_vport(skb->dev);
  47         if (unlikely(!vport))
  48                 goto error;
  49
  50         if (unlikely(skb_warn_if_lro(skb)))
  51                 goto error;
  52
  53         /* Make our own copy of the packet.  Otherwise we will mangle the
  54          * packet for anyone who came before us (e.g. tcpdump via AF_PACKET).
  55          */
  56         skb = skb_share_check(skb, GFP_ATOMIC);
  57         if (unlikely(!skb))
  58                 return;
  59
  60         skb_push(skb, ETH_HLEN);
  61         ovs_skb_postpush_rcsum(skb, skb->data, ETH_HLEN);
  62         ovs_vport_receive(vport, skb, tun_info);
  63         return;
  64 error:
  65         kfree_skb(skb);
  66 }
  67
  68 #ifndef HAVE_METADATA_DST
  69 #define port_receive(skb)  netdev_port_receive(skb, NULL)
  70 #else
  71 #define port_receive(skb)  netdev_port_receive(skb, skb_tunnel_info(skb))
  72 #endif
  73
  74 #if defined HAVE_RX_HANDLER_PSKB  /* 2.6.39 and above or backports */
  75 /* Called with rcu_read_lock and bottom-halves disabled. */
  76 static rx_handler_result_t netdev_frame_hook(struct sk_buff **pskb)
  77 {
  78         struct sk_buff *skb = *pskb;
  79
  80         if (unlikely(skb->pkt_type == PACKET_LOOPBACK))
  81                 return RX_HANDLER_PASS;
  82
  83         port_receive(skb);
  84         return RX_HANDLER_CONSUMED;
  85 }
  86 #elif LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,36) || \
  87       defined HAVE_RHEL_OVS_HOOK
  88 /* Called with rcu_read_lock and bottom-halves disabled. */
  89 static struct sk_buff *netdev_frame_hook(struct sk_buff *skb)
  90 {
  91         if (unlikely(skb->pkt_type == PACKET_LOOPBACK))
  92                 return skb;
  93
  94         port_receive(skb);
  95         return NULL;
  96 }
  97 #elif LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,32)
  98 /*
  99  * Used as br_handle_frame_hook.  (Cannot run bridge at the same time, even on
 100  * different set of devices!)
 101  */
 102 /* Called with rcu_read_lock and bottom-halves disabled. */
 103 static struct sk_buff *netdev_frame_hook(struct net_bridge_port *p,
 104                                          struct sk_buff *skb)
 105 {
 106         port_receive(skb);
 107         return NULL;
 108 }
 109 #else
 110 #error
 111 #endif
 112
 113 static struct net_device *get_dpdev(const struct datapath *dp)
 114 {
 115         struct vport *local;
 116
 117         local = ovs_vport_ovsl(dp, OVSP_LOCAL);
 118         BUG_ON(!local);
 119         return local->dev;
 120 }
 121
 122 struct vport *ovs_netdev_link(struct vport *vport, const char *name)
 123 {
 124         int err;
 125
 126         vport->dev = dev_get_by_name(ovs_dp_get_net(vport->dp), name);
 127         if (!vport->dev) {
 128                 err = -ENODEV;
 129                 goto error_free_vport;
 130         }
 131
 132         if (vport->dev->flags & IFF_LOOPBACK ||
 133             vport->dev->type != ARPHRD_ETHER ||
 134             ovs_is_internal_dev(vport->dev)) {
 135                 err = -EINVAL;
 136                 goto error_put;
 137         }
 138
 139         rtnl_lock();
 140         err = netdev_master_upper_dev_link(vport->dev,
 141                                            get_dpdev(vport->dp));
 142         if (err)
 143                 goto error_unlock;
 144
 145         err = netdev_rx_handler_register(vport->dev, netdev_frame_hook,
 146                                          vport);
 147         if (err)
 148                 goto error_master_upper_dev_unlink;
 149
 150         dev_disable_lro(vport->dev);
 151         dev_set_promiscuity(vport->dev, 1);
 152         vport->dev->priv_flags |= IFF_OVS_DATAPATH;
 153         rtnl_unlock();
 154
 155         return vport;
 156
 157 error_master_upper_dev_unlink:
 158         netdev_upper_dev_unlink(vport->dev, get_dpdev(vport->dp));
 159 error_unlock:
 160         rtnl_unlock();
 161 error_put:
 162         dev_put(vport->dev);
 163 error_free_vport:
 164         ovs_vport_free(vport);
 165         return ERR_PTR(err);
 166 }
 167 EXPORT_SYMBOL_GPL(ovs_netdev_link);
 168
 169 static struct vport *netdev_create(const struct vport_parms *parms)
 170 {
 171         struct vport *vport;
 172
 173         vport = ovs_vport_alloc(0, &ovs_netdev_vport_ops, parms);
 174         if (IS_ERR(vport))
 175                 return vport;
 176
 177         return ovs_netdev_link(vport, parms->name);
 178 }
 179
 180 static void vport_netdev_free(struct rcu_head *rcu)
 181 {
 182         struct vport *vport = container_of(rcu, struct vport, rcu);
 183
 184         if (vport->dev)
 185                 dev_put(vport->dev);
 186         ovs_vport_free(vport);
 187 }
 188
 189 void ovs_netdev_detach_dev(struct vport *vport)
 190 {
 191         ASSERT_RTNL();
 192         vport->dev->priv_flags &= ~IFF_OVS_DATAPATH;
 193         netdev_rx_handler_unregister(vport->dev);
 194         netdev_upper_dev_unlink(vport->dev,
 195                                 netdev_master_upper_dev_get(vport->dev));
 196         dev_set_promiscuity(vport->dev, -1);
 197 }
 198 EXPORT_SYMBOL_GPL(ovs_netdev_detach_dev);
 199
 200 static void netdev_destroy(struct vport *vport)
 201 {
 202         rtnl_lock();
 203         if (vport->dev->priv_flags & IFF_OVS_DATAPATH)
 204                 ovs_netdev_detach_dev(vport);
 205         rtnl_unlock();
 206
 207         call_rcu(&vport->rcu, vport_netdev_free);
 208 }
 209
 210 void ovs_netdev_tunnel_destroy(struct vport *vport)
 211 {
 212         rtnl_lock();
 213         if (vport->dev->priv_flags & IFF_OVS_DATAPATH)
 214                 ovs_netdev_detach_dev(vport);
 215
 216         /* Early release so we can unregister the device */
 217         dev_put(vport->dev);
 218         rtnl_delete_link(vport->dev);
 219         vport->dev = NULL;
 220         rtnl_unlock();
 221
 222         call_rcu(&vport->rcu, vport_netdev_free);
 223 }
 224 EXPORT_SYMBOL_GPL(ovs_netdev_tunnel_destroy);
 225
 226 /* Returns null if this device is not attached to a datapath. */
 227 struct vport *ovs_netdev_get_vport(struct net_device *dev)
 228 {
 229 #if defined HAVE_NETDEV_RX_HANDLER_REGISTER || \
 230     defined HAVE_RHEL_OVS_HOOK
 231 #ifdef HAVE_OVS_DATAPATH
 232         if (likely(dev->priv_flags & IFF_OVS_DATAPATH))
 233 #else
 234         if (likely(rcu_access_pointer(dev->rx_handler) == netdev_frame_hook))
 235 #endif
 236 #ifdef HAVE_RHEL_OVS_HOOK
 237                 return (struct vport *)rcu_dereference_rtnl(dev->ax25_ptr);
 238 #else
 239 #ifdef HAVE_NET_DEVICE_EXTENDED
 240                 return (struct vport *)
 241                         rcu_dereference_rtnl(netdev_extended(dev)->rx_handler_data);
 242 #else
 243                 return (struct vport *)rcu_dereference_rtnl(dev->rx_handler_data);
 244 #endif
 245 #endif
 246         else
 247                 return NULL;
 248 #else
 249         return (struct vport *)rcu_dereference_rtnl(dev->br_port);
 250 #endif
 251 }
 252
 253 static struct vport_ops ovs_netdev_vport_ops = {
 254         .type           = OVS_VPORT_TYPE_NETDEV,
 255         .create         = netdev_create,
 256         .destroy        = netdev_destroy,
 257         .send           = dev_queue_xmit,
 258 };
 259
 260 int __init ovs_netdev_init(void)
 261 {
 262         return ovs_vport_ops_register(&ovs_netdev_vport_ops);
 263 }
 264
 265 void ovs_netdev_exit(void)
 266 {
 267         ovs_vport_ops_unregister(&ovs_netdev_vport_ops);
 268 }
 269
 270 #if !defined HAVE_NETDEV_RX_HANDLER_REGISTER && \
 271     !defined HAVE_RHEL_OVS_HOOK
 272 /*
 273  * Enforces, mutual exclusion with the Linux bridge module, by declaring and
 274  * exporting br_should_route_hook.  Because the bridge module also exports the
 275  * same symbol, the module loader will refuse to load both modules at the same
 276  * time (e.g. "bridge: exports duplicate symbol br_should_route_hook (owned by
 277  * openvswitch)").
 278  *
 279  * Before Linux 2.6.36, Open vSwitch cannot safely coexist with the Linux
 280  * bridge module, so openvswitch uses this macro in those versions.  In
 281  * Linux 2.6.36 and later, Open vSwitch can coexist with the bridge module.
 282  *
 283  * The use of "typeof" here avoids the need to track changes in the type of
 284  * br_should_route_hook over various kernel versions.
 285  */
 286 typeof(br_should_route_hook) br_should_route_hook;
 287 EXPORT_SYMBOL(br_should_route_hook);
 288 #endif