1 /******************************************************************************
4 * A two-interface loopback device to emulate a local netfront-netback
5 * connection. This ensures that local packet delivery looks identical
6 * to inter-domain delivery. Most importantly, packets delivered locally
7 * originating from other domains will get *copied* when they traverse this
8 * driver. This prevents unbounded delays in socket-buffer queues from
9 * causing the netback driver to "seize up".
11 * This driver creates a symmetric pair of loopback interfaces with names
12 * vif0.0 and veth0. The intention is that 'vif0.0' is bound to an Ethernet
13 * bridge, just like a proper netback interface, while a local IP interface
14 * is configured on 'veth0'.
16 * As with a real netback interface, vif0.0 is configured with a suitable
17 * dummy MAC address. No default is provided for veth0: a reasonable strategy
18 * is to transfer eth0's MAC address to veth0, and give eth0 a dummy address
19 * (to avoid confusing the Etherbridge).
21 * Copyright (c) 2005 K A Fraser
23 * This program is free software; you can redistribute it and/or
24 * modify it under the terms of the GNU General Public License version 2
25 * as published by the Free Software Foundation; or, when distributed
26 * separately from the Linux kernel or incorporated into other
27 * software packages, subject to the following license:
29 * Permission is hereby granted, free of charge, to any person obtaining a copy
30 * of this source file (the "Software"), to deal in the Software without
31 * restriction, including without limitation the rights to use, copy, modify,
32 * merge, publish, distribute, sublicense, and/or sell copies of the Software,
33 * and to permit persons to whom the Software is furnished to do so, subject to
34 * the following conditions:
36 * The above copyright notice and this permission notice shall be included in
37 * all copies or substantial portions of the Software.
39 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
40 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
41 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
42 * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
43 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
44 * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
48 #include <linux/module.h>
49 #include <linux/netdevice.h>
50 #include <linux/inetdevice.h>
51 #include <linux/etherdevice.h>
52 #include <linux/skbuff.h>
53 #include <linux/ethtool.h>
55 #include <net/xfrm.h> /* secpath_reset() */
56 #include <asm/hypervisor.h> /* is_initial_xendomain() */
58 #include "../../../net/core/kmap_skb.h"
60 static int nloopbacks = -1;
61 module_param(nloopbacks, int, 0);
62 MODULE_PARM_DESC(nloopbacks, "Number of netback-loopback devices to create");
65 struct net_device *loopback_dev;
66 struct net_device_stats stats;
69 static int loopback_open(struct net_device *dev)
71 struct net_private *np = netdev_priv(dev);
72 memset(&np->stats, 0, sizeof(np->stats));
73 netif_start_queue(dev);
77 static int loopback_close(struct net_device *dev)
79 netif_stop_queue(dev);
84 static int is_foreign(unsigned long pfn)
86 /* NB. Play it safe for auto-translation mode. */
87 return (xen_feature(XENFEAT_auto_translated_physmap) ||
88 (phys_to_machine_mapping[pfn] & FOREIGN_FRAME_BIT));
91 /* How to detect a foreign mapping? Play it safe. */
92 #define is_foreign(pfn) (1)
95 static int skb_remove_foreign_references(struct sk_buff *skb)
102 BUG_ON(skb_shinfo(skb)->frag_list);
104 for (i = 0; i < skb_shinfo(skb)->nr_frags; i++) {
105 pfn = page_to_pfn(skb_shinfo(skb)->frags[i].page);
106 if (!is_foreign(pfn))
109 page = alloc_page(GFP_ATOMIC | __GFP_NOWARN);
113 vaddr = kmap_skb_frag(&skb_shinfo(skb)->frags[i]);
114 off = skb_shinfo(skb)->frags[i].page_offset;
115 memcpy(page_address(page) + off,
117 skb_shinfo(skb)->frags[i].size);
118 kunmap_skb_frag(vaddr);
120 put_page(skb_shinfo(skb)->frags[i].page);
121 skb_shinfo(skb)->frags[i].page = page;
127 static int loopback_start_xmit(struct sk_buff *skb, struct net_device *dev)
129 struct net_private *np = netdev_priv(dev);
131 if (!skb_remove_foreign_references(skb)) {
132 np->stats.tx_dropped++;
137 dst_release(skb->dst);
142 np->stats.tx_bytes += skb->len;
143 np->stats.tx_packets++;
145 /* Switch to loopback context. */
146 dev = np->loopback_dev;
147 np = netdev_priv(dev);
149 np->stats.rx_bytes += skb->len;
150 np->stats.rx_packets++;
152 if (skb->ip_summed == CHECKSUM_PARTIAL) {
153 /* Defer checksum calculation. */
154 skb->proto_csum_blank = 1;
155 /* Must be a local packet: assert its integrity. */
156 skb->proto_data_valid = 1;
159 skb->ip_summed = skb->proto_data_valid ?
160 CHECKSUM_UNNECESSARY : CHECKSUM_NONE;
162 skb->pkt_type = PACKET_HOST; /* overridden by eth_type_trans() */
163 skb->protocol = eth_type_trans(skb, dev);
165 dev->last_rx = jiffies;
167 /* Flush netfilter context: rx'ed skbuffs not expected to have any. */
176 static struct net_device_stats *loopback_get_stats(struct net_device *dev)
178 struct net_private *np = netdev_priv(dev);
182 static struct ethtool_ops network_ethtool_ops =
184 .get_tx_csum = ethtool_op_get_tx_csum,
185 .set_tx_csum = ethtool_op_set_tx_csum,
186 .get_sg = ethtool_op_get_sg,
187 .set_sg = ethtool_op_set_sg,
188 .get_tso = ethtool_op_get_tso,
189 .set_tso = ethtool_op_set_tso,
190 .get_link = ethtool_op_get_link,
194 * Nothing to do here. Virtual interface is point-to-point and the
195 * physical interface is probably promiscuous anyway.
197 static void loopback_set_multicast_list(struct net_device *dev)
201 static void loopback_construct(struct net_device *dev, struct net_device *lo)
203 struct net_private *np = netdev_priv(dev);
205 np->loopback_dev = lo;
207 dev->open = loopback_open;
208 dev->stop = loopback_close;
209 dev->hard_start_xmit = loopback_start_xmit;
210 dev->get_stats = loopback_get_stats;
211 dev->set_multicast_list = loopback_set_multicast_list;
212 dev->change_mtu = NULL; /* allow arbitrary mtu */
214 dev->tx_queue_len = 0;
216 dev->features = (NETIF_F_HIGHDMA |
222 SET_ETHTOOL_OPS(dev, &network_ethtool_ops);
225 * We do not set a jumbo MTU on the interface. Otherwise the network
226 * stack will try to send large packets that will get dropped by the
227 * Ethernet bridge (unless the physical Ethernet interface is
228 * configured to transfer jumbo packets). If a larger MTU is desired
229 * then the system administrator can specify it using the 'ifconfig'
232 /*dev->mtu = 16*1024;*/
235 static int __init make_loopback(int i)
237 struct net_device *dev1, *dev2;
238 char dev_name[IFNAMSIZ];
241 sprintf(dev_name, "vif0.%d", i);
242 dev1 = alloc_netdev(sizeof(struct net_private), dev_name, ether_setup);
246 sprintf(dev_name, "veth%d", i);
247 dev2 = alloc_netdev(sizeof(struct net_private), dev_name, ether_setup);
251 loopback_construct(dev1, dev2);
252 loopback_construct(dev2, dev1);
255 * Initialise a dummy MAC address for the 'dummy backend' interface. We
256 * choose the numerically largest non-broadcast address to prevent the
257 * address getting stolen by an Ethernet bridge for STP purposes.
259 memset(dev1->dev_addr, 0xFF, ETH_ALEN);
260 dev1->dev_addr[0] &= ~0x01;
262 if ((err = register_netdev(dev1)) != 0)
265 if ((err = register_netdev(dev2)) != 0) {
266 unregister_netdev(dev1);
279 static void __exit clean_loopback(int i)
281 struct net_device *dev1, *dev2;
282 char dev_name[IFNAMSIZ];
284 sprintf(dev_name, "vif0.%d", i);
285 dev1 = dev_get_by_name(dev_name);
286 sprintf(dev_name, "veth%d", i);
287 dev2 = dev_get_by_name(dev_name);
289 unregister_netdev(dev2);
290 unregister_netdev(dev1);
296 static int __init loopback_init(void)
300 if (nloopbacks == -1)
301 nloopbacks = is_initial_xendomain() ? 4 : 0;
303 for (i = 0; i < nloopbacks; i++)
304 if ((err = make_loopback(i)) != 0)
310 module_init(loopback_init);
312 static void __exit loopback_exit(void)
316 for (i = nloopbacks; i-- > 0; )
320 module_exit(loopback_exit);
322 MODULE_LICENSE("Dual BSD/GPL");