1// SPDX-License-Identifier: GPL-2.0
2/*
3 * Copyright (C) 2015-2019 Jason A. Donenfeld <Jason@zx2c4.com>. All Rights Reserved.
4 */
5
6#include "device.h"
7#include "peer.h"
8#include "socket.h"
9#include "queueing.h"
10#include "messages.h"
11
12#include <linux/ctype.h>
13#include <linux/net.h>
14#include <linux/if_vlan.h>
15#include <linux/if_ether.h>
16#include <linux/inetdevice.h>
17#include <net/udp_tunnel.h>
18#include <net/ipv6.h>
19
20static int send4(struct wg_device *wg, struct sk_buff *skb,
21		 struct endpoint *endpoint, u8 ds, struct dst_cache *cache)
22{
23	struct flowi4 fl = {
24		.saddr = endpoint->src4.s_addr,
25		.daddr = endpoint->addr4.sin_addr.s_addr,
26		.fl4_dport = endpoint->addr4.sin_port,
27		.flowi4_mark = wg->fwmark,
28		.flowi4_proto = IPPROTO_UDP
29	};
30	struct rtable *rt = NULL;
31	struct sock *sock;
32	int ret = 0;
33
34	skb_mark_not_on_list(skb);
35	skb->dev = wg->dev;
36	skb->mark = wg->fwmark;
37
38	rcu_read_lock_bh();
39	sock = rcu_dereference_bh(wg->sock4);
40
41	if (unlikely(!sock)) {
42		ret = -ENONET;
43		goto err;
44	}
45
46	fl.fl4_sport = inet_sk(sock)->inet_sport;
47
48	if (cache)
49		rt = dst_cache_get_ip4(cache, &fl.saddr);
50
51	if (!rt) {
52		security_sk_classify_flow(sock, flowi4_to_flowi_common(&fl));
53		if (unlikely(!inet_confirm_addr(sock_net(sock), NULL, 0,
54						fl.saddr, RT_SCOPE_HOST))) {
55			endpoint->src4.s_addr = 0;
56			endpoint->src_if4 = 0;
57			fl.saddr = 0;
58			if (cache)
59				dst_cache_reset(cache);
60		}
61		rt = ip_route_output_flow(sock_net(sock), &fl, sock);
62		if (unlikely(endpoint->src_if4 && ((IS_ERR(rt) &&
63			     PTR_ERR(rt) == -EINVAL) || (!IS_ERR(rt) &&
64			     rt->dst.dev->ifindex != endpoint->src_if4)))) {
65			endpoint->src4.s_addr = 0;
66			endpoint->src_if4 = 0;
67			fl.saddr = 0;
68			if (cache)
69				dst_cache_reset(cache);
70			if (!IS_ERR(rt))
71				ip_rt_put(rt);
72			rt = ip_route_output_flow(sock_net(sock), &fl, sock);
73		}
74		if (IS_ERR(rt)) {
75			ret = PTR_ERR(rt);
76			net_dbg_ratelimited("%s: No route to %pISpfsc, error %d\n",
77					    wg->dev->name, &endpoint->addr, ret);
78			goto err;
79		}
80		if (cache)
81			dst_cache_set_ip4(cache, &rt->dst, fl.saddr);
82	}
83
84	skb->ignore_df = 1;
85	udp_tunnel_xmit_skb(rt, sock, skb, fl.saddr, fl.daddr, ds,
86			    ip4_dst_hoplimit(&rt->dst), 0, fl.fl4_sport,
87			    fl.fl4_dport, false, false);
88	goto out;
89
90err:
91	kfree_skb(skb);
92out:
93	rcu_read_unlock_bh();
94	return ret;
95}
96
97static int send6(struct wg_device *wg, struct sk_buff *skb,
98		 struct endpoint *endpoint, u8 ds, struct dst_cache *cache)
99{
100#if IS_ENABLED(CONFIG_IPV6)
101	struct flowi6 fl = {
102		.saddr = endpoint->src6,
103		.daddr = endpoint->addr6.sin6_addr,
104		.fl6_dport = endpoint->addr6.sin6_port,
105		.flowi6_mark = wg->fwmark,
106		.flowi6_oif = endpoint->addr6.sin6_scope_id,
107		.flowi6_proto = IPPROTO_UDP
108		/* TODO: addr->sin6_flowinfo */
109	};
110	struct dst_entry *dst = NULL;
111	struct sock *sock;
112	int ret = 0;
113
114	skb_mark_not_on_list(skb);
115	skb->dev = wg->dev;
116	skb->mark = wg->fwmark;
117
118	rcu_read_lock_bh();
119	sock = rcu_dereference_bh(wg->sock6);
120
121	if (unlikely(!sock)) {
122		ret = -ENONET;
123		goto err;
124	}
125
126	fl.fl6_sport = inet_sk(sock)->inet_sport;
127
128	if (cache)
129		dst = dst_cache_get_ip6(cache, &fl.saddr);
130
131	if (!dst) {
132		security_sk_classify_flow(sock, flowi6_to_flowi_common(&fl));
133		if (unlikely(!ipv6_addr_any(&fl.saddr) &&
134			     !ipv6_chk_addr(sock_net(sock), &fl.saddr, NULL, 0))) {
135			endpoint->src6 = fl.saddr = in6addr_any;
136			if (cache)
137				dst_cache_reset(cache);
138		}
139		dst = ipv6_stub->ipv6_dst_lookup_flow(sock_net(sock), sock, &fl,
140						      NULL);
141		if (IS_ERR(dst)) {
142			ret = PTR_ERR(dst);
143			net_dbg_ratelimited("%s: No route to %pISpfsc, error %d\n",
144					    wg->dev->name, &endpoint->addr, ret);
145			goto err;
146		}
147		if (cache)
148			dst_cache_set_ip6(cache, dst, &fl.saddr);
149	}
150
151	skb->ignore_df = 1;
152	udp_tunnel6_xmit_skb(dst, sock, skb, skb->dev, &fl.saddr, &fl.daddr, ds,
153			     ip6_dst_hoplimit(dst), 0, fl.fl6_sport,
154			     fl.fl6_dport, false);
155	goto out;
156
157err:
158	kfree_skb(skb);
159out:
160	rcu_read_unlock_bh();
161	return ret;
162#else
163	kfree_skb(skb);
164	return -EAFNOSUPPORT;
165#endif
166}
167
168int wg_socket_send_skb_to_peer(struct wg_peer *peer, struct sk_buff *skb, u8 ds)
169{
170	size_t skb_len = skb->len;
171	int ret = -EAFNOSUPPORT;
172
173	read_lock_bh(&peer->endpoint_lock);
174	if (peer->endpoint.addr.sa_family == AF_INET)
175		ret = send4(peer->device, skb, &peer->endpoint, ds,
176			    &peer->endpoint_cache);
177	else if (peer->endpoint.addr.sa_family == AF_INET6)
178		ret = send6(peer->device, skb, &peer->endpoint, ds,
179			    &peer->endpoint_cache);
180	else
181		dev_kfree_skb(skb);
182	if (likely(!ret))
183		peer->tx_bytes += skb_len;
184	read_unlock_bh(&peer->endpoint_lock);
185
186	return ret;
187}
188
189int wg_socket_send_buffer_to_peer(struct wg_peer *peer, void *buffer,
190				  size_t len, u8 ds)
191{
192	struct sk_buff *skb = alloc_skb(len + SKB_HEADER_LEN, GFP_ATOMIC);
193
194	if (unlikely(!skb))
195		return -ENOMEM;
196
197	skb_reserve(skb, SKB_HEADER_LEN);
198	skb_set_inner_network_header(skb, 0);
199	skb_put_data(skb, buffer, len);
200	return wg_socket_send_skb_to_peer(peer, skb, ds);
201}
202
203int wg_socket_send_buffer_as_reply_to_skb(struct wg_device *wg,
204					  struct sk_buff *in_skb, void *buffer,
205					  size_t len)
206{
207	int ret = 0;
208	struct sk_buff *skb;
209	struct endpoint endpoint;
210
211	if (unlikely(!in_skb))
212		return -EINVAL;
213	ret = wg_socket_endpoint_from_skb(&endpoint, in_skb);
214	if (unlikely(ret < 0))
215		return ret;
216
217	skb = alloc_skb(len + SKB_HEADER_LEN, GFP_ATOMIC);
218	if (unlikely(!skb))
219		return -ENOMEM;
220	skb_reserve(skb, SKB_HEADER_LEN);
221	skb_set_inner_network_header(skb, 0);
222	skb_put_data(skb, buffer, len);
223
224	if (endpoint.addr.sa_family == AF_INET)
225		ret = send4(wg, skb, &endpoint, 0, NULL);
226	else if (endpoint.addr.sa_family == AF_INET6)
227		ret = send6(wg, skb, &endpoint, 0, NULL);
228	/* No other possibilities if the endpoint is valid, which it is,
229	 * as we checked above.
230	 */
231
232	return ret;
233}
234
235int wg_socket_endpoint_from_skb(struct endpoint *endpoint,
236				const struct sk_buff *skb)
237{
238	memset(endpoint, 0, sizeof(*endpoint));
239	if (skb->protocol == htons(ETH_P_IP)) {
240		endpoint->addr4.sin_family = AF_INET;
241		endpoint->addr4.sin_port = udp_hdr(skb)->source;
242		endpoint->addr4.sin_addr.s_addr = ip_hdr(skb)->saddr;
243		endpoint->src4.s_addr = ip_hdr(skb)->daddr;
244		endpoint->src_if4 = skb->skb_iif;
245	} else if (IS_ENABLED(CONFIG_IPV6) && skb->protocol == htons(ETH_P_IPV6)) {
246		endpoint->addr6.sin6_family = AF_INET6;
247		endpoint->addr6.sin6_port = udp_hdr(skb)->source;
248		endpoint->addr6.sin6_addr = ipv6_hdr(skb)->saddr;
249		endpoint->addr6.sin6_scope_id = ipv6_iface_scope_id(
250			&ipv6_hdr(skb)->saddr, skb->skb_iif);
251		endpoint->src6 = ipv6_hdr(skb)->daddr;
252	} else {
253		return -EINVAL;
254	}
255	return 0;
256}
257
258static bool endpoint_eq(const struct endpoint *a, const struct endpoint *b)
259{
260	return (a->addr.sa_family == AF_INET && b->addr.sa_family == AF_INET &&
261		a->addr4.sin_port == b->addr4.sin_port &&
262		a->addr4.sin_addr.s_addr == b->addr4.sin_addr.s_addr &&
263		a->src4.s_addr == b->src4.s_addr && a->src_if4 == b->src_if4) ||
264	       (a->addr.sa_family == AF_INET6 &&
265		b->addr.sa_family == AF_INET6 &&
266		a->addr6.sin6_port == b->addr6.sin6_port &&
267		ipv6_addr_equal(&a->addr6.sin6_addr, &b->addr6.sin6_addr) &&
268		a->addr6.sin6_scope_id == b->addr6.sin6_scope_id &&
269		ipv6_addr_equal(&a->src6, &b->src6)) ||
270	       unlikely(!a->addr.sa_family && !b->addr.sa_family);
271}
272
273void wg_socket_set_peer_endpoint(struct wg_peer *peer,
274				 const struct endpoint *endpoint)
275{
276	/* First we check unlocked, in order to optimize, since it's pretty rare
277	 * that an endpoint will change. If we happen to be mid-write, and two
278	 * CPUs wind up writing the same thing or something slightly different,
279	 * it doesn't really matter much either.
280	 */
281	if (endpoint_eq(endpoint, &peer->endpoint))
282		return;
283	write_lock_bh(&peer->endpoint_lock);
284	if (endpoint->addr.sa_family == AF_INET) {
285		peer->endpoint.addr4 = endpoint->addr4;
286		peer->endpoint.src4 = endpoint->src4;
287		peer->endpoint.src_if4 = endpoint->src_if4;
288	} else if (IS_ENABLED(CONFIG_IPV6) && endpoint->addr.sa_family == AF_INET6) {
289		peer->endpoint.addr6 = endpoint->addr6;
290		peer->endpoint.src6 = endpoint->src6;
291	} else {
292		goto out;
293	}
294	dst_cache_reset(&peer->endpoint_cache);
295out:
296	write_unlock_bh(&peer->endpoint_lock);
297}
298
299void wg_socket_set_peer_endpoint_from_skb(struct wg_peer *peer,
300					  const struct sk_buff *skb)
301{
302	struct endpoint endpoint;
303
304	if (!wg_socket_endpoint_from_skb(&endpoint, skb))
305		wg_socket_set_peer_endpoint(peer, &endpoint);
306}
307
308void wg_socket_clear_peer_endpoint_src(struct wg_peer *peer)
309{
310	write_lock_bh(&peer->endpoint_lock);
311	memset(&peer->endpoint.src6, 0, sizeof(peer->endpoint.src6));
312	dst_cache_reset_now(&peer->endpoint_cache);
313	write_unlock_bh(&peer->endpoint_lock);
314}
315
316static int wg_receive(struct sock *sk, struct sk_buff *skb)
317{
318	struct wg_device *wg;
319
320	if (unlikely(!sk))
321		goto err;
322	wg = sk->sk_user_data;
323	if (unlikely(!wg))
324		goto err;
325	skb_mark_not_on_list(skb);
326	wg_packet_receive(wg, skb);
327	return 0;
328
329err:
330	kfree_skb(skb);
331	return 0;
332}
333
334static void sock_free(struct sock *sock)
335{
336	if (unlikely(!sock))
337		return;
338	sk_clear_memalloc(sock);
339	udp_tunnel_sock_release(sock->sk_socket);
340}
341
342static void set_sock_opts(struct socket *sock)
343{
344	sock->sk->sk_allocation = GFP_ATOMIC;
345	sock->sk->sk_sndbuf = INT_MAX;
346	sk_set_memalloc(sock->sk);
347}
348
349int wg_socket_init(struct wg_device *wg, u16 port)
350{
351	struct net *net;
352	int ret;
353	struct udp_tunnel_sock_cfg cfg = {
354		.sk_user_data = wg,
355		.encap_type = 1,
356		.encap_rcv = wg_receive
357	};
358	struct socket *new4 = NULL, *new6 = NULL;
359	struct udp_port_cfg port4 = {
360		.family = AF_INET,
361		.local_ip.s_addr = htonl(INADDR_ANY),
362		.local_udp_port = htons(port),
363		.use_udp_checksums = true
364	};
365#if IS_ENABLED(CONFIG_IPV6)
366	int retries = 0;
367	struct udp_port_cfg port6 = {
368		.family = AF_INET6,
369		.local_ip6 = IN6ADDR_ANY_INIT,
370		.use_udp6_tx_checksums = true,
371		.use_udp6_rx_checksums = true,
372		.ipv6_v6only = true
373	};
374#endif
375
376	rcu_read_lock();
377	net = rcu_dereference(wg->creating_net);
378	net = net ? maybe_get_net(net) : NULL;
379	rcu_read_unlock();
380	if (unlikely(!net))
381		return -ENONET;
382
383#if IS_ENABLED(CONFIG_IPV6)
384retry:
385#endif
386
387	ret = udp_sock_create(net, &port4, &new4);
388	if (ret < 0) {
389		pr_err("%s: Could not create IPv4 socket\n", wg->dev->name);
390		goto out;
391	}
392	set_sock_opts(new4);
393	setup_udp_tunnel_sock(net, new4, &cfg);
394
395#if IS_ENABLED(CONFIG_IPV6)
396	if (ipv6_mod_enabled()) {
397		port6.local_udp_port = inet_sk(new4->sk)->inet_sport;
398		ret = udp_sock_create(net, &port6, &new6);
399		if (ret < 0) {
400			udp_tunnel_sock_release(new4);
401			if (ret == -EADDRINUSE && !port && retries++ < 100)
402				goto retry;
403			pr_err("%s: Could not create IPv6 socket\n",
404			       wg->dev->name);
405			goto out;
406		}
407		set_sock_opts(new6);
408		setup_udp_tunnel_sock(net, new6, &cfg);
409	}
410#endif
411
412	wg_socket_reinit(wg, new4->sk, new6 ? new6->sk : NULL);
413	ret = 0;
414out:
415	put_net(net);
416	return ret;
417}
418
419void wg_socket_reinit(struct wg_device *wg, struct sock *new4,
420		      struct sock *new6)
421{
422	struct sock *old4, *old6;
423
424	mutex_lock(&wg->socket_update_lock);
425	old4 = rcu_dereference_protected(wg->sock4,
426				lockdep_is_held(&wg->socket_update_lock));
427	old6 = rcu_dereference_protected(wg->sock6,
428				lockdep_is_held(&wg->socket_update_lock));
429	rcu_assign_pointer(wg->sock4, new4);
430	rcu_assign_pointer(wg->sock6, new6);
431	if (new4)
432		wg->incoming_port = ntohs(inet_sk(new4)->inet_sport);
433	mutex_unlock(&wg->socket_update_lock);
434	synchronize_net();
435	sock_free(old4);
436	sock_free(old6);
437}
438