xref: /kernel/linux/linux-5.10/drivers/net/loopback.c (revision 8c2ecf20)
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/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	int len;
72
73	skb_tx_timestamp(skb);
74
75	/* do not fool net_timestamp_check() with various clock bases */
76	skb->tstamp = 0;
77
78	skb_orphan(skb);
79
80	/* Before queueing this packet to netif_rx(),
81	 * make sure dst is refcounted.
82	 */
83	skb_dst_force(skb);
84
85	skb->protocol = eth_type_trans(skb, dev);
86
87	len = skb->len;
88	if (likely(netif_rx(skb) == NET_RX_SUCCESS))
89		dev_lstats_add(dev, len);
90
91	return NETDEV_TX_OK;
92}
93
94void dev_lstats_read(struct net_device *dev, u64 *packets, u64 *bytes)
95{
96	int i;
97
98	*packets = 0;
99	*bytes = 0;
100
101	for_each_possible_cpu(i) {
102		const struct pcpu_lstats *lb_stats;
103		u64 tbytes, tpackets;
104		unsigned int start;
105
106		lb_stats = per_cpu_ptr(dev->lstats, i);
107		do {
108			start = u64_stats_fetch_begin_irq(&lb_stats->syncp);
109			tpackets = u64_stats_read(&lb_stats->packets);
110			tbytes = u64_stats_read(&lb_stats->bytes);
111		} while (u64_stats_fetch_retry_irq(&lb_stats->syncp, start));
112		*bytes   += tbytes;
113		*packets += tpackets;
114	}
115}
116EXPORT_SYMBOL(dev_lstats_read);
117
118static void loopback_get_stats64(struct net_device *dev,
119				 struct rtnl_link_stats64 *stats)
120{
121	u64 packets, bytes;
122
123	dev_lstats_read(dev, &packets, &bytes);
124
125	stats->rx_packets = packets;
126	stats->tx_packets = packets;
127	stats->rx_bytes   = bytes;
128	stats->tx_bytes   = bytes;
129}
130
131static u32 always_on(struct net_device *dev)
132{
133	return 1;
134}
135
136static const struct ethtool_ops loopback_ethtool_ops = {
137	.get_link		= always_on,
138	.get_ts_info		= ethtool_op_get_ts_info,
139};
140
141static int loopback_dev_init(struct net_device *dev)
142{
143	dev->lstats = netdev_alloc_pcpu_stats(struct pcpu_lstats);
144	if (!dev->lstats)
145		return -ENOMEM;
146	return 0;
147}
148
149static void loopback_dev_free(struct net_device *dev)
150{
151	dev_net(dev)->loopback_dev = NULL;
152	free_percpu(dev->lstats);
153}
154
155static const struct net_device_ops loopback_ops = {
156	.ndo_init        = loopback_dev_init,
157	.ndo_start_xmit  = loopback_xmit,
158	.ndo_get_stats64 = loopback_get_stats64,
159	.ndo_set_mac_address = eth_mac_addr,
160};
161
162static void gen_lo_setup(struct net_device *dev,
163			 unsigned int mtu,
164			 const struct ethtool_ops *eth_ops,
165			 const struct header_ops *hdr_ops,
166			 const struct net_device_ops *dev_ops,
167			 void (*dev_destructor)(struct net_device *dev))
168{
169	dev->mtu		= mtu;
170	dev->hard_header_len	= ETH_HLEN;	/* 14	*/
171	dev->min_header_len	= ETH_HLEN;	/* 14	*/
172	dev->addr_len		= ETH_ALEN;	/* 6	*/
173	dev->type		= ARPHRD_LOOPBACK;	/* 0x0001*/
174	dev->flags		= IFF_LOOPBACK;
175	dev->priv_flags		|= IFF_LIVE_ADDR_CHANGE | IFF_NO_QUEUE;
176	netif_keep_dst(dev);
177	dev->hw_features	= NETIF_F_GSO_SOFTWARE;
178	dev->features		= NETIF_F_SG | NETIF_F_FRAGLIST
179		| NETIF_F_GSO_SOFTWARE
180		| NETIF_F_HW_CSUM
181		| NETIF_F_RXCSUM
182		| NETIF_F_SCTP_CRC
183		| NETIF_F_HIGHDMA
184		| NETIF_F_LLTX
185		| NETIF_F_NETNS_LOCAL
186		| NETIF_F_VLAN_CHALLENGED
187		| NETIF_F_LOOPBACK;
188	dev->ethtool_ops	= eth_ops;
189	dev->header_ops		= hdr_ops;
190	dev->netdev_ops		= dev_ops;
191	dev->needs_free_netdev	= true;
192	dev->priv_destructor	= dev_destructor;
193}
194
195/* The loopback device is special. There is only one instance
196 * per network namespace.
197 */
198static void loopback_setup(struct net_device *dev)
199{
200	gen_lo_setup(dev, (64 * 1024), &loopback_ethtool_ops, &eth_header_ops,
201		     &loopback_ops, loopback_dev_free);
202}
203
204/* Setup and register the loopback device. */
205static __net_init int loopback_net_init(struct net *net)
206{
207	struct net_device *dev;
208	int err;
209
210	err = -ENOMEM;
211	dev = alloc_netdev(0, "lo", NET_NAME_PREDICTABLE, loopback_setup);
212	if (!dev)
213		goto out;
214
215	dev_net_set(dev, net);
216	err = register_netdev(dev);
217	if (err)
218		goto out_free_netdev;
219
220	BUG_ON(dev->ifindex != LOOPBACK_IFINDEX);
221	net->loopback_dev = dev;
222	return 0;
223
224out_free_netdev:
225	free_netdev(dev);
226out:
227	if (net_eq(net, &init_net))
228		panic("loopback: Failed to register netdevice: %d\n", err);
229	return err;
230}
231
232/* Registered in net/core/dev.c */
233struct pernet_operations __net_initdata loopback_net_ops = {
234	.init = loopback_net_init,
235};
236
237/* blackhole netdevice */
238static netdev_tx_t blackhole_netdev_xmit(struct sk_buff *skb,
239					 struct net_device *dev)
240{
241	kfree_skb(skb);
242	net_warn_ratelimited("%s(): Dropping skb.\n", __func__);
243	return NETDEV_TX_OK;
244}
245
246static const struct net_device_ops blackhole_netdev_ops = {
247	.ndo_start_xmit = blackhole_netdev_xmit,
248};
249
250/* This is a dst-dummy device used specifically for invalidated
251 * DSTs and unlike loopback, this is not per-ns.
252 */
253static void blackhole_netdev_setup(struct net_device *dev)
254{
255	gen_lo_setup(dev, ETH_MIN_MTU, NULL, NULL, &blackhole_netdev_ops, NULL);
256}
257
258/* Setup and register the blackhole_netdev. */
259static int __init blackhole_netdev_init(void)
260{
261	blackhole_netdev = alloc_netdev(0, "blackhole_dev", NET_NAME_UNKNOWN,
262					blackhole_netdev_setup);
263	if (!blackhole_netdev)
264		return -ENOMEM;
265
266	rtnl_lock();
267	dev_init_scheduler(blackhole_netdev);
268	dev_activate(blackhole_netdev);
269	rtnl_unlock();
270
271	blackhole_netdev->flags |= IFF_UP | IFF_RUNNING;
272	dev_net_set(blackhole_netdev, &init_net);
273
274	return 0;
275}
276
277device_initcall(blackhole_netdev_init);
278