1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
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
35struct internal_dev {
36 struct vport *vport;
37};
38
39static struct vport_ops ovs_internal_vport_ops;
40
41static struct internal_dev *internal_dev_priv(struct net_device *netdev)
42{
43 return netdev_priv(netdev);
44}
45
46
47static 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
69static int internal_dev_open(struct net_device *netdev)
70{
71 netif_start_queue(netdev);
72 return 0;
73}
74
75static int internal_dev_stop(struct net_device *netdev)
76{
77 netif_stop_queue(netdev);
78 return 0;
79}
80
81static void internal_dev_getinfo(struct net_device *netdev,
82 struct ethtool_drvinfo *info)
83{
84 strlcpy(info->driver, "openvswitch", sizeof(info->driver));
85}
86
87static const struct ethtool_ops internal_dev_ethtool_ops = {
88 .get_drvinfo = internal_dev_getinfo,
89 .get_link = ethtool_op_get_link,
90};
91
92static void internal_dev_destructor(struct net_device *dev)
93{
94 struct vport *vport = ovs_internal_dev_get_vport(dev);
95
96 ovs_vport_free(vport);
97}
98
99static void
100internal_get_stats(struct net_device *dev, struct rtnl_link_stats64 *stats)
101{
102 int i;
103
104 memset(stats, 0, sizeof(*stats));
105 stats->rx_errors = dev->stats.rx_errors;
106 stats->tx_errors = dev->stats.tx_errors;
107 stats->tx_dropped = dev->stats.tx_dropped;
108 stats->rx_dropped = dev->stats.rx_dropped;
109
110 for_each_possible_cpu(i) {
111 const struct pcpu_sw_netstats *percpu_stats;
112 struct pcpu_sw_netstats local_stats;
113 unsigned int start;
114
115 percpu_stats = per_cpu_ptr(dev->tstats, i);
116
117 do {
118 start = u64_stats_fetch_begin_irq(&percpu_stats->syncp);
119 local_stats = *percpu_stats;
120 } while (u64_stats_fetch_retry_irq(&percpu_stats->syncp, start));
121
122 stats->rx_bytes += local_stats.rx_bytes;
123 stats->rx_packets += local_stats.rx_packets;
124 stats->tx_bytes += local_stats.tx_bytes;
125 stats->tx_packets += local_stats.tx_packets;
126 }
127}
128
129static const struct net_device_ops internal_dev_netdev_ops = {
130 .ndo_open = internal_dev_open,
131 .ndo_stop = internal_dev_stop,
132 .ndo_start_xmit = internal_dev_xmit,
133 .ndo_set_mac_address = eth_mac_addr,
134 .ndo_get_stats64 = internal_get_stats,
135 .ndo_size = sizeof(struct net_device_ops),
136};
137
138static struct rtnl_link_ops internal_dev_link_ops __read_mostly = {
139 .kind = "openvswitch",
140};
141
142static void do_setup(struct net_device *netdev)
143{
144 ether_setup(netdev);
145
146 netdev->extended->max_mtu = ETH_MAX_MTU;
147
148 netdev->netdev_ops = &internal_dev_netdev_ops;
149
150 netdev->priv_flags &= ~IFF_TX_SKB_SHARING;
151 netdev->priv_flags |= IFF_LIVE_ADDR_CHANGE | IFF_OPENVSWITCH |
152 IFF_NO_QUEUE;
153 netdev->extended->needs_free_netdev = true;
154 netdev->extended->priv_destructor = internal_dev_destructor;
155 netdev->ethtool_ops = &internal_dev_ethtool_ops;
156 netdev->rtnl_link_ops = &internal_dev_link_ops;
157
158 netdev->features = NETIF_F_LLTX | NETIF_F_SG | NETIF_F_FRAGLIST |
159 NETIF_F_HIGHDMA | NETIF_F_HW_CSUM |
160 NETIF_F_GSO_SOFTWARE | NETIF_F_GSO_ENCAP_ALL;
161
162 netdev->vlan_features = netdev->features;
163 netdev->hw_enc_features = netdev->features;
164 netdev->features |= NETIF_F_HW_VLAN_CTAG_TX | NETIF_F_HW_VLAN_STAG_TX;
165 netdev->hw_features = netdev->features & ~NETIF_F_LLTX;
166
167 eth_hw_addr_random(netdev);
168}
169
170static struct vport *internal_dev_create(const struct vport_parms *parms)
171{
172 struct vport *vport;
173 struct internal_dev *internal_dev;
174 int err;
175
176 vport = ovs_vport_alloc(0, &ovs_internal_vport_ops, parms);
177 if (IS_ERR(vport)) {
178 err = PTR_ERR(vport);
179 goto error;
180 }
181
182 vport->dev = alloc_netdev(sizeof(struct internal_dev),
183 parms->name, do_setup);
184 if (!vport->dev) {
185 err = -ENOMEM;
186 goto error_free_vport;
187 }
188 vport->dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
189 if (!vport->dev->tstats) {
190 err = -ENOMEM;
191 goto error_free_netdev;
192 }
193
194 dev_net_set(vport->dev, ovs_dp_get_net(vport->dp));
195 internal_dev = internal_dev_priv(vport->dev);
196 internal_dev->vport = vport;
197
198
199 if (vport->port_no == OVSP_LOCAL)
200 vport->dev->features |= NETIF_F_NETNS_LOCAL;
201
202 rtnl_lock();
203 err = register_netdevice(vport->dev);
204 if (err)
205 goto error_unlock;
206
207 dev_set_promiscuity(vport->dev, 1);
208 rtnl_unlock();
209 netif_start_queue(vport->dev);
210
211 return vport;
212
213error_unlock:
214 rtnl_unlock();
215 free_percpu(vport->dev->tstats);
216error_free_netdev:
217 free_netdev(vport->dev);
218error_free_vport:
219 ovs_vport_free(vport);
220error:
221 return ERR_PTR(err);
222}
223
224static void internal_dev_destroy(struct vport *vport)
225{
226 netif_stop_queue(vport->dev);
227 rtnl_lock();
228 dev_set_promiscuity(vport->dev, -1);
229
230
231 unregister_netdevice(vport->dev);
232 free_percpu(vport->dev->tstats);
233 rtnl_unlock();
234}
235
236static netdev_tx_t internal_dev_recv(struct sk_buff *skb)
237{
238 struct net_device *netdev = skb->dev;
239 struct pcpu_sw_netstats *stats;
240
241 if (unlikely(!(netdev->flags & IFF_UP))) {
242 kfree_skb(skb);
243 netdev->stats.rx_dropped++;
244 return NETDEV_TX_OK;
245 }
246
247 skb_dst_drop(skb);
248 nf_reset(skb);
249 secpath_reset(skb);
250
251 skb->pkt_type = PACKET_HOST;
252 skb->protocol = eth_type_trans(skb, netdev);
253 skb_postpull_rcsum(skb, eth_hdr(skb), ETH_HLEN);
254
255 stats = this_cpu_ptr(netdev->tstats);
256 u64_stats_update_begin(&stats->syncp);
257 stats->rx_packets++;
258 stats->rx_bytes += skb->len;
259 u64_stats_update_end(&stats->syncp);
260
261 netif_rx(skb);
262 return NETDEV_TX_OK;
263}
264
265static struct vport_ops ovs_internal_vport_ops = {
266 .type = OVS_VPORT_TYPE_INTERNAL,
267 .create = internal_dev_create,
268 .destroy = internal_dev_destroy,
269 .send = internal_dev_recv,
270};
271
272int ovs_is_internal_dev(const struct net_device *netdev)
273{
274 return netdev->netdev_ops == &internal_dev_netdev_ops;
275}
276
277struct vport *ovs_internal_dev_get_vport(struct net_device *netdev)
278{
279 if (!ovs_is_internal_dev(netdev))
280 return NULL;
281
282 return internal_dev_priv(netdev)->vport;
283}
284
285int ovs_internal_dev_rtnl_link_register(void)
286{
287 int err;
288
289 err = rtnl_link_register(&internal_dev_link_ops);
290 if (err < 0)
291 return err;
292
293 err = ovs_vport_ops_register(&ovs_internal_vport_ops);
294 if (err < 0)
295 rtnl_link_unregister(&internal_dev_link_ops);
296
297 return err;
298}
299
300void ovs_internal_dev_rtnl_link_unregister(void)
301{
302 ovs_vport_ops_unregister(&ovs_internal_vport_ops);
303 rtnl_link_unregister(&internal_dev_link_ops);
304}
305
