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
 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/* Called with rcu_read_lock_bh. */
 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	free_netdev(dev);
 98}
 99
100static struct rtnl_link_stats64 *
101internal_get_stats(struct net_device *dev, struct rtnl_link_stats64 *stats)
102{
103	int i;
104
105	memset(stats, 0, sizeof(*stats));
106	stats->rx_errors  = dev->stats.rx_errors;
107	stats->tx_errors  = dev->stats.tx_errors;
108	stats->tx_dropped = dev->stats.tx_dropped;
109	stats->rx_dropped = dev->stats.rx_dropped;
110
111	for_each_possible_cpu(i) {
112		const struct pcpu_sw_netstats *percpu_stats;
113		struct pcpu_sw_netstats local_stats;
114		unsigned int start;
115
116		percpu_stats = per_cpu_ptr(dev->tstats, i);
117
118		do {
119			start = u64_stats_fetch_begin_irq(&percpu_stats->syncp);
120			local_stats = *percpu_stats;
121		} while (u64_stats_fetch_retry_irq(&percpu_stats->syncp, start));
122
123		stats->rx_bytes         += local_stats.rx_bytes;
124		stats->rx_packets       += local_stats.rx_packets;
125		stats->tx_bytes         += local_stats.tx_bytes;
126		stats->tx_packets       += local_stats.tx_packets;
127	}
128
129	return stats;
130}
131
132static void internal_set_rx_headroom(struct net_device *dev, int new_hr)
133{
134	dev->needed_headroom = new_hr < 0 ? 0 : new_hr;
135}
136
137static const struct net_device_ops internal_dev_netdev_ops = {
138	.ndo_open = internal_dev_open,
139	.ndo_stop = internal_dev_stop,
140	.ndo_start_xmit = internal_dev_xmit,
141	.ndo_set_mac_address = eth_mac_addr,
142	.ndo_get_stats64 = internal_get_stats,
143	.ndo_set_rx_headroom = internal_set_rx_headroom,
144};
145
146static struct rtnl_link_ops internal_dev_link_ops __read_mostly = {
147	.kind = "openvswitch",
148};
149
150static void do_setup(struct net_device *netdev)
151{
152	ether_setup(netdev);
153
154	netdev->netdev_ops = &internal_dev_netdev_ops;
155
156	netdev->priv_flags &= ~IFF_TX_SKB_SHARING;
157	netdev->priv_flags |= IFF_LIVE_ADDR_CHANGE | IFF_OPENVSWITCH |
158			      IFF_PHONY_HEADROOM | IFF_NO_QUEUE;
159	netdev->destructor = internal_dev_destructor;
160	netdev->ethtool_ops = &internal_dev_ethtool_ops;
161	netdev->rtnl_link_ops = &internal_dev_link_ops;
162
163	netdev->features = NETIF_F_LLTX | NETIF_F_SG | NETIF_F_FRAGLIST |
164			   NETIF_F_HIGHDMA | NETIF_F_HW_CSUM |
165			   NETIF_F_GSO_SOFTWARE | NETIF_F_GSO_ENCAP_ALL;
166
167	netdev->vlan_features = netdev->features;
168	netdev->hw_enc_features = netdev->features;
169	netdev->features |= NETIF_F_HW_VLAN_CTAG_TX | NETIF_F_HW_VLAN_STAG_TX;
170	netdev->hw_features = netdev->features & ~NETIF_F_LLTX;
171
172	eth_hw_addr_random(netdev);
173}
174
175static struct vport *internal_dev_create(const struct vport_parms *parms)
176{
177	struct vport *vport;
178	struct internal_dev *internal_dev;
179	int err;
180
181	vport = ovs_vport_alloc(0, &ovs_internal_vport_ops, parms);
182	if (IS_ERR(vport)) {
183		err = PTR_ERR(vport);
184		goto error;
185	}
186
187	vport->dev = alloc_netdev(sizeof(struct internal_dev),
188				  parms->name, NET_NAME_USER, do_setup);
189	if (!vport->dev) {
190		err = -ENOMEM;
191		goto error_free_vport;
192	}
193	vport->dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
194	if (!vport->dev->tstats) {
195		err = -ENOMEM;
196		goto error_free_netdev;
197	}
198	vport->dev->needed_headroom = vport->dp->max_headroom;
199
200	dev_net_set(vport->dev, ovs_dp_get_net(vport->dp));
201	internal_dev = internal_dev_priv(vport->dev);
202	internal_dev->vport = vport;
203
204	/* Restrict bridge port to current netns. */
205	if (vport->port_no == OVSP_LOCAL)
206		vport->dev->features |= NETIF_F_NETNS_LOCAL;
207
208	rtnl_lock();
209	err = register_netdevice(vport->dev);
210	if (err)
211		goto error_unlock;
212
213	dev_set_promiscuity(vport->dev, 1);
214	rtnl_unlock();
215	netif_start_queue(vport->dev);
216
217	return vport;
218
219error_unlock:
220	rtnl_unlock();
221	free_percpu(vport->dev->tstats);
222error_free_netdev:
223	free_netdev(vport->dev);
224error_free_vport:
225	ovs_vport_free(vport);
226error:
227	return ERR_PTR(err);
228}
229
230static void internal_dev_destroy(struct vport *vport)
231{
232	netif_stop_queue(vport->dev);
233	rtnl_lock();
234	dev_set_promiscuity(vport->dev, -1);
235
236	/* unregister_netdevice() waits for an RCU grace period. */
237	unregister_netdevice(vport->dev);
238	free_percpu(vport->dev->tstats);
239	rtnl_unlock();
240}
241
242static netdev_tx_t internal_dev_recv(struct sk_buff *skb)
243{
244	struct net_device *netdev = skb->dev;
245	struct pcpu_sw_netstats *stats;
246
247	if (unlikely(!(netdev->flags & IFF_UP))) {
248		kfree_skb(skb);
249		netdev->stats.rx_dropped++;
250		return NETDEV_TX_OK;
251	}
252
253	skb_dst_drop(skb);
254	nf_reset(skb);
255	secpath_reset(skb);
256
257	skb->pkt_type = PACKET_HOST;
258	skb->protocol = eth_type_trans(skb, netdev);
259	skb_postpull_rcsum(skb, eth_hdr(skb), ETH_HLEN);
260
261	stats = this_cpu_ptr(netdev->tstats);
262	u64_stats_update_begin(&stats->syncp);
263	stats->rx_packets++;
264	stats->rx_bytes += skb->len;
265	u64_stats_update_end(&stats->syncp);
266
267	netif_rx(skb);
268	return NETDEV_TX_OK;
269}
270
271static struct vport_ops ovs_internal_vport_ops = {
272	.type		= OVS_VPORT_TYPE_INTERNAL,
273	.create		= internal_dev_create,
274	.destroy	= internal_dev_destroy,
275	.send		= internal_dev_recv,
276};
277
278int ovs_is_internal_dev(const struct net_device *netdev)
279{
280	return netdev->netdev_ops == &internal_dev_netdev_ops;
281}
282
283struct vport *ovs_internal_dev_get_vport(struct net_device *netdev)
284{
285	if (!ovs_is_internal_dev(netdev))
286		return NULL;
287
288	return internal_dev_priv(netdev)->vport;
289}
290
291int ovs_internal_dev_rtnl_link_register(void)
292{
293	int err;
294
295	err = rtnl_link_register(&internal_dev_link_ops);
296	if (err < 0)
297		return err;
298
299	err = ovs_vport_ops_register(&ovs_internal_vport_ops);
300	if (err < 0)
301		rtnl_link_unregister(&internal_dev_link_ops);
302
303	return err;
304}
305
306void ovs_internal_dev_rtnl_link_unregister(void)
307{
308	ovs_vport_ops_unregister(&ovs_internal_vport_ops);
309	rtnl_link_unregister(&internal_dev_link_ops);
310}