projects/linux·blob v5.3 · refs

275 lines · 7.1 KiB

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, <[email protected]>
13 * Donald Becker, <[email protected]>
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/sock.h>
48#include <net/checksum.h>
49#include <linux/if_ether.h> /* For the statistics structure. */
50#include <linux/if_arp.h> /* For ARPHRD_ETHER */
51#include <linux/ip.h>
52#include <linux/tcp.h>
53#include <linux/percpu.h>
54#include <linux/net_tstamp.h>
55#include <net/net_namespace.h>
56#include <linux/u64_stats_sync.h>
57
58/* blackhole_netdev - a device used for dsts that are marked expired!
59 * This is global device (instead of per-net-ns) since it's not needed
60 * to be per-ns and gets initialized at boot time.
61 */
62struct net_device *blackhole_netdev;
63EXPORT_SYMBOL(blackhole_netdev);
64
65/* The higher levels take care of making this non-reentrant (it's
66 * called with bh's disabled).
67 */
68static netdev_tx_t loopback_xmit(struct sk_buff *skb,
69 struct net_device *dev)
70{
71 struct pcpu_lstats *lb_stats;
72 int len;
73
74 skb_tx_timestamp(skb);
75
76 /* do not fool net_timestamp_check() with various clock bases */
77 skb->tstamp = 0;
78
79 skb_orphan(skb);
80
81 /* Before queueing this packet to netif_rx(),
82 * make sure dst is refcounted.
83 */
84 skb_dst_force(skb);
85
86 skb->protocol = eth_type_trans(skb, dev);
87
88 /* it's OK to use per_cpu_ptr() because BHs are off */
89 lb_stats = this_cpu_ptr(dev->lstats);
90
91 len = skb->len;
92 if (likely(netif_rx(skb) == NET_RX_SUCCESS)) {
93 u64_stats_update_begin(&lb_stats->syncp);
94 lb_stats->bytes += len;
95 lb_stats->packets++;
96 u64_stats_update_end(&lb_stats->syncp);
97 }
98
99 return NETDEV_TX_OK;
100}
101
102static void loopback_get_stats64(struct net_device *dev,
103 struct rtnl_link_stats64 *stats)
104{
105 u64 bytes = 0;
106 u64 packets = 0;
107 int i;
108
109 for_each_possible_cpu(i) {
110 const struct pcpu_lstats *lb_stats;
111 u64 tbytes, tpackets;
112 unsigned int start;
113
114 lb_stats = per_cpu_ptr(dev->lstats, i);
115 do {
116 start = u64_stats_fetch_begin_irq(&lb_stats->syncp);
117 tbytes = lb_stats->bytes;
118 tpackets = lb_stats->packets;
119 } while (u64_stats_fetch_retry_irq(&lb_stats->syncp, start));
120 bytes += tbytes;
121 packets += tpackets;
122 }
123 stats->rx_packets = packets;
124 stats->tx_packets = packets;
125 stats->rx_bytes = bytes;
126 stats->tx_bytes = bytes;
127}
128
129static u32 always_on(struct net_device *dev)
130{
131 return 1;
132}
133
134static const struct ethtool_ops loopback_ethtool_ops = {
135 .get_link = always_on,
136 .get_ts_info = ethtool_op_get_ts_info,
137};
138
139static int loopback_dev_init(struct net_device *dev)
140{
141 dev->lstats = netdev_alloc_pcpu_stats(struct pcpu_lstats);
142 if (!dev->lstats)
143 return -ENOMEM;
144 return 0;
145}
146
147static void loopback_dev_free(struct net_device *dev)
148{
149 dev_net(dev)->loopback_dev = NULL;
150 free_percpu(dev->lstats);
151}
152
153static const struct net_device_ops loopback_ops = {
154 .ndo_init = loopback_dev_init,
155 .ndo_start_xmit = loopback_xmit,
156 .ndo_get_stats64 = loopback_get_stats64,
157 .ndo_set_mac_address = eth_mac_addr,
158};
159
160static void gen_lo_setup(struct net_device *dev,
161 unsigned int mtu,
162 const struct ethtool_ops *eth_ops,
163 const struct header_ops *hdr_ops,
164 const struct net_device_ops *dev_ops,
165 void (*dev_destructor)(struct net_device *dev))
166{
167 dev->mtu = mtu;
168 dev->hard_header_len = ETH_HLEN; /* 14 */
169 dev->min_header_len = ETH_HLEN; /* 14 */
170 dev->addr_len = ETH_ALEN; /* 6 */
171 dev->type = ARPHRD_LOOPBACK; /* 0x0001*/
172 dev->flags = IFF_LOOPBACK;
173 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE | IFF_NO_QUEUE;
174 netif_keep_dst(dev);
175 dev->hw_features = NETIF_F_GSO_SOFTWARE;
176 dev->features = NETIF_F_SG | NETIF_F_FRAGLIST
177 | NETIF_F_GSO_SOFTWARE
178 | NETIF_F_HW_CSUM
179 | NETIF_F_RXCSUM
180 | NETIF_F_SCTP_CRC
181 | NETIF_F_HIGHDMA
182 | NETIF_F_LLTX
183 | NETIF_F_NETNS_LOCAL
184 | NETIF_F_VLAN_CHALLENGED
185 | NETIF_F_LOOPBACK;
186 dev->ethtool_ops = eth_ops;
187 dev->header_ops = hdr_ops;
188 dev->netdev_ops = dev_ops;
189 dev->needs_free_netdev = true;
190 dev->priv_destructor = dev_destructor;
191}
192
193/* The loopback device is special. There is only one instance
194 * per network namespace.
195 */
196static void loopback_setup(struct net_device *dev)
197{
198 gen_lo_setup(dev, (64 * 1024), &loopback_ethtool_ops, &eth_header_ops,
199 &loopback_ops, loopback_dev_free);
200}
201
202/* Setup and register the loopback device. */
203static __net_init int loopback_net_init(struct net *net)
204{
205 struct net_device *dev;
206 int err;
207
208 err = -ENOMEM;
209 dev = alloc_netdev(0, "lo", NET_NAME_UNKNOWN, loopback_setup);
210 if (!dev)
211 goto out;
212
213 dev_net_set(dev, net);
214 err = register_netdev(dev);
215 if (err)
216 goto out_free_netdev;
217
218 BUG_ON(dev->ifindex != LOOPBACK_IFINDEX);
219 net->loopback_dev = dev;
220 return 0;
221
222out_free_netdev:
223 free_netdev(dev);
224out:
225 if (net_eq(net, &init_net))
226 panic("loopback: Failed to register netdevice: %d\n", err);
227 return err;
228}
229
230/* Registered in net/core/dev.c */
231struct pernet_operations __net_initdata loopback_net_ops = {
232 .init = loopback_net_init,
233};
234
235/* blackhole netdevice */
236static netdev_tx_t blackhole_netdev_xmit(struct sk_buff *skb,
237 struct net_device *dev)
238{
239 kfree_skb(skb);
240 net_warn_ratelimited("%s(): Dropping skb.\n", __func__);
241 return NETDEV_TX_OK;
242}
243
244static const struct net_device_ops blackhole_netdev_ops = {
245 .ndo_start_xmit = blackhole_netdev_xmit,
246};
247
248/* This is a dst-dummy device used specifically for invalidated
249 * DSTs and unlike loopback, this is not per-ns.
250 */
251static void blackhole_netdev_setup(struct net_device *dev)
252{
253 gen_lo_setup(dev, ETH_MIN_MTU, NULL, NULL, &blackhole_netdev_ops, NULL);
254}
255
256/* Setup and register the blackhole_netdev. */
257static int __init blackhole_netdev_init(void)
258{
259 blackhole_netdev = alloc_netdev(0, "blackhole_dev", NET_NAME_UNKNOWN,
260 blackhole_netdev_setup);
261 if (!blackhole_netdev)
262 return -ENOMEM;
263
264 rtnl_lock();
265 dev_init_scheduler(blackhole_netdev);
266 dev_activate(blackhole_netdev);
267 rtnl_unlock();
268
269 blackhole_netdev->flags |= IFF_UP | IFF_RUNNING;
270 dev_net_set(blackhole_netdev, &init_net);
271
272 return 0;
273}
274
275device_initcall(blackhole_netdev_init);