1 // SPDX-License-Identifier: GPL-2.0-or-later
2 /*
3 * INET An implementation of the TCP/IP protocol suite for the LINUX
4 * operating system. INET is implemented using the BSD Socket
5 * interface as the means of communication with the user level.
6 *
7 * Pseudo-driver for the loopback interface.
8 *
9 * Version: @(#)loopback.c 1.0.4b 08/16/93
10 *
11 * Authors: Ross Biro
12 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
13 * Donald Becker, <becker@scyld.com>
14 *
15 * Alan Cox : Fixed oddments for NET3.014
16 * Alan Cox : Rejig for NET3.029 snap #3
17 * Alan Cox : Fixed NET3.029 bugs and sped up
18 * Larry McVoy : Tiny tweak to double performance
19 * Alan Cox : Backed out LMV's tweak - the linux mm
20 * can't take it...
21 * Michael Griffith: Don't bother computing the checksums
22 * on packets received on the loopback
23 * interface.
24 * Alexey Kuznetsov: Potential hang under some extreme
25 * cases removed.
26 */
27 #include <linux/kernel.h>
28 #include <linux/jiffies.h>
29 #include <linux/module.h>
30 #include <linux/interrupt.h>
31 #include <linux/fs.h>
32 #include <linux/types.h>
33 #include <linux/string.h>
34 #include <linux/socket.h>
35 #include <linux/errno.h>
36 #include <linux/fcntl.h>
37 #include <linux/in.h>
38
39 #include <linux/uaccess.h>
40 #include <linux/io.h>
41
42 #include <linux/inet.h>
43 #include <linux/netdevice.h>
44 #include <linux/etherdevice.h>
45 #include <linux/skbuff.h>
46 #include <linux/ethtool.h>
47 #include <net/sch_generic.h>
48 #include <net/sock.h>
49 #include <net/checksum.h>
50 #include <linux/if_ether.h> /* For the statistics structure. */
51 #include <linux/if_arp.h> /* For ARPHRD_ETHER */
52 #include <linux/ip.h>
53 #include <linux/tcp.h>
54 #include <linux/percpu.h>
55 #include <linux/net_tstamp.h>
56 #include <net/net_namespace.h>
57 #include <net/netdev_lock.h>
58 #include <linux/u64_stats_sync.h>
59
60 /* blackhole_netdev - a device used for dsts that are marked expired!
61 * This is global device (instead of per-net-ns) since it's not needed
62 * to be per-ns and gets initialized at boot time.
63 */
64 struct net_device *blackhole_netdev;
65 EXPORT_SYMBOL(blackhole_netdev);
66
67 /* The higher levels take care of making this non-reentrant (it's
68 * called with bh's disabled).
69 */
loopback_xmit(struct sk_buff * skb,struct net_device * dev)70 static netdev_tx_t loopback_xmit(struct sk_buff *skb,
71 struct net_device *dev)
72 {
73 int len;
74
75 skb_tx_timestamp(skb);
76
77 /* do not fool net_timestamp_check() with various clock bases */
78 skb_clear_tstamp(skb);
79
80 skb_orphan(skb);
81
82 /* Before queueing this packet to __netif_rx(),
83 * make sure dst is refcounted.
84 */
85 skb_dst_force(skb);
86
87 skb->protocol = eth_type_trans(skb, dev);
88
89 len = skb->len;
90 if (likely(__netif_rx(skb) == NET_RX_SUCCESS))
91 dev_lstats_add(dev, len);
92
93 return NETDEV_TX_OK;
94 }
95
dev_lstats_read(struct net_device * dev,u64 * packets,u64 * bytes)96 void dev_lstats_read(struct net_device *dev, u64 *packets, u64 *bytes)
97 {
98 int i;
99
100 *packets = 0;
101 *bytes = 0;
102
103 for_each_possible_cpu(i) {
104 const struct pcpu_lstats *lb_stats;
105 u64 tbytes, tpackets;
106 unsigned int start;
107
108 lb_stats = per_cpu_ptr(dev->lstats, i);
109 do {
110 start = u64_stats_fetch_begin(&lb_stats->syncp);
111 tpackets = u64_stats_read(&lb_stats->packets);
112 tbytes = u64_stats_read(&lb_stats->bytes);
113 } while (u64_stats_fetch_retry(&lb_stats->syncp, start));
114 *bytes += tbytes;
115 *packets += tpackets;
116 }
117 }
118 EXPORT_SYMBOL(dev_lstats_read);
119
loopback_get_stats64(struct net_device * dev,struct rtnl_link_stats64 * stats)120 static void loopback_get_stats64(struct net_device *dev,
121 struct rtnl_link_stats64 *stats)
122 {
123 u64 packets, bytes;
124
125 dev_lstats_read(dev, &packets, &bytes);
126
127 stats->rx_packets = packets;
128 stats->tx_packets = packets;
129 stats->rx_bytes = bytes;
130 stats->tx_bytes = bytes;
131 }
132
always_on(struct net_device * dev)133 static u32 always_on(struct net_device *dev)
134 {
135 return 1;
136 }
137
138 static const struct ethtool_ops loopback_ethtool_ops = {
139 .get_link = always_on,
140 .get_ts_info = ethtool_op_get_ts_info,
141 };
142
loopback_dev_init(struct net_device * dev)143 static int loopback_dev_init(struct net_device *dev)
144 {
145 netdev_lockdep_set_classes(dev);
146 return 0;
147 }
148
loopback_dev_free(struct net_device * dev)149 static void loopback_dev_free(struct net_device *dev)
150 {
151 dev_net(dev)->loopback_dev = NULL;
152 }
153
154 static const struct net_device_ops loopback_ops = {
155 .ndo_init = loopback_dev_init,
156 .ndo_start_xmit = loopback_xmit,
157 .ndo_get_stats64 = loopback_get_stats64,
158 .ndo_set_mac_address = eth_mac_addr,
159 };
160
gen_lo_setup(struct net_device * dev,unsigned int mtu,const struct ethtool_ops * eth_ops,const struct header_ops * hdr_ops,const struct net_device_ops * dev_ops,void (* dev_destructor)(struct net_device * dev))161 static void gen_lo_setup(struct net_device *dev,
162 unsigned int mtu,
163 const struct ethtool_ops *eth_ops,
164 const struct header_ops *hdr_ops,
165 const struct net_device_ops *dev_ops,
166 void (*dev_destructor)(struct net_device *dev))
167 {
168 dev->mtu = mtu;
169 dev->hard_header_len = ETH_HLEN; /* 14 */
170 dev->min_header_len = ETH_HLEN; /* 14 */
171 dev->addr_len = ETH_ALEN; /* 6 */
172 dev->type = ARPHRD_LOOPBACK; /* 0x0001*/
173 dev->flags = IFF_LOOPBACK;
174 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE | IFF_NO_QUEUE;
175 dev->lltx = true;
176 dev->netns_immutable = true;
177 netif_keep_dst(dev);
178 dev->hw_features = NETIF_F_GSO_SOFTWARE;
179 dev->features = NETIF_F_SG | NETIF_F_FRAGLIST
180 | NETIF_F_GSO_SOFTWARE
181 | NETIF_F_HW_CSUM
182 | NETIF_F_RXCSUM
183 | NETIF_F_SCTP_CRC
184 | NETIF_F_HIGHDMA
185 | NETIF_F_VLAN_CHALLENGED
186 | NETIF_F_LOOPBACK;
187 dev->ethtool_ops = eth_ops;
188 dev->header_ops = hdr_ops;
189 dev->netdev_ops = dev_ops;
190 dev->needs_free_netdev = true;
191 dev->pcpu_stat_type = NETDEV_PCPU_STAT_LSTATS;
192 dev->priv_destructor = dev_destructor;
193
194 netif_set_tso_max_size(dev, GSO_MAX_SIZE);
195 }
196
197 /* The loopback device is special. There is only one instance
198 * per network namespace.
199 */
loopback_setup(struct net_device * dev)200 static void loopback_setup(struct net_device *dev)
201 {
202 gen_lo_setup(dev, (64 * 1024), &loopback_ethtool_ops, ð_header_ops,
203 &loopback_ops, loopback_dev_free);
204 }
205
206 /* Setup and register the loopback device. */
loopback_net_init(struct net * net)207 static __net_init int loopback_net_init(struct net *net)
208 {
209 struct net_device *dev;
210 int err;
211
212 err = -ENOMEM;
213 dev = alloc_netdev(0, "lo", NET_NAME_PREDICTABLE, loopback_setup);
214 if (!dev)
215 goto out;
216
217 dev_net_set(dev, net);
218 err = register_netdev(dev);
219 if (err)
220 goto out_free_netdev;
221
222 BUG_ON(dev->ifindex != LOOPBACK_IFINDEX);
223 net->loopback_dev = dev;
224 return 0;
225
226 out_free_netdev:
227 free_netdev(dev);
228 out:
229 if (net_eq(net, &init_net))
230 panic("loopback: Failed to register netdevice: %d\n", err);
231 return err;
232 }
233
234 /* Registered in net/core/dev.c */
235 struct pernet_operations __net_initdata loopback_net_ops = {
236 .init = loopback_net_init,
237 };
238
239 /* blackhole netdevice */
blackhole_netdev_xmit(struct sk_buff * skb,struct net_device * dev)240 static netdev_tx_t blackhole_netdev_xmit(struct sk_buff *skb,
241 struct net_device *dev)
242 {
243 kfree_skb(skb);
244 net_warn_ratelimited("%s(): Dropping skb.\n", __func__);
245 return NETDEV_TX_OK;
246 }
247
blackhole_neigh_output(struct neighbour * n,struct sk_buff * skb)248 static int blackhole_neigh_output(struct neighbour *n, struct sk_buff *skb)
249 {
250 kfree_skb(skb);
251 return 0;
252 }
253
blackhole_neigh_construct(struct net_device * dev,struct neighbour * n)254 static int blackhole_neigh_construct(struct net_device *dev,
255 struct neighbour *n)
256 {
257 n->output = blackhole_neigh_output;
258 return 0;
259 }
260
261 static const struct net_device_ops blackhole_netdev_ops = {
262 .ndo_start_xmit = blackhole_netdev_xmit,
263 .ndo_neigh_construct = blackhole_neigh_construct,
264 };
265
266 /* This is a dst-dummy device used specifically for invalidated
267 * DSTs and unlike loopback, this is not per-ns.
268 */
blackhole_netdev_setup(struct net_device * dev)269 static void blackhole_netdev_setup(struct net_device *dev)
270 {
271 gen_lo_setup(dev, ETH_MIN_MTU, NULL, NULL, &blackhole_netdev_ops, NULL);
272 }
273
274 /* Setup and register the blackhole_netdev. */
blackhole_netdev_init(void)275 static int __init blackhole_netdev_init(void)
276 {
277 blackhole_netdev = alloc_netdev(0, "blackhole_dev", NET_NAME_UNKNOWN,
278 blackhole_netdev_setup);
279 if (!blackhole_netdev)
280 return -ENOMEM;
281
282 rtnl_net_lock(&init_net);
283 dev_init_scheduler(blackhole_netdev);
284 dev_activate(blackhole_netdev);
285 rtnl_net_unlock(&init_net);
286
287 blackhole_netdev->flags |= IFF_UP | IFF_RUNNING;
288
289 return 0;
290 }
291
292 device_initcall(blackhole_netdev_init);
293