xref: /linux/net/ipv6/xfrm6_input.c (revision 8f5b5f78113e881cb8570c961b0dc42b218a1b9e)
1 // SPDX-License-Identifier: GPL-2.0
2 /*
3  * xfrm6_input.c: based on net/ipv4/xfrm4_input.c
4  *
5  * Authors:
6  *	Mitsuru KANDA @USAGI
7  *	Kazunori MIYAZAWA @USAGI
8  *	Kunihiro Ishiguro <kunihiro@ipinfusion.com>
9  *	YOSHIFUJI Hideaki @USAGI
10  *		IPv6 support
11  */
12 
13 #include <linux/module.h>
14 #include <linux/string.h>
15 #include <linux/netfilter.h>
16 #include <linux/netfilter_ipv6.h>
17 #include <net/ipv6.h>
18 #include <net/xfrm.h>
19 #include <net/protocol.h>
20 #include <net/gro.h>
21 
22 int xfrm6_rcv_spi(struct sk_buff *skb, int nexthdr, __be32 spi,
23 		  struct ip6_tnl *t)
24 {
25 	XFRM_TUNNEL_SKB_CB(skb)->tunnel.ip6 = t;
26 	XFRM_SPI_SKB_CB(skb)->family = AF_INET6;
27 	XFRM_SPI_SKB_CB(skb)->daddroff = offsetof(struct ipv6hdr, daddr);
28 	return xfrm_input(skb, nexthdr, spi, 0);
29 }
30 EXPORT_SYMBOL(xfrm6_rcv_spi);
31 
32 static int xfrm6_transport_finish2(struct net *net, struct sock *sk,
33 				   struct sk_buff *skb)
34 {
35 	if (xfrm_trans_queue(skb, ip6_rcv_finish)) {
36 		kfree_skb(skb);
37 		return NET_RX_DROP;
38 	}
39 
40 	return 0;
41 }
42 
43 int xfrm6_transport_finish(struct sk_buff *skb, int async)
44 {
45 	struct xfrm_offload *xo = xfrm_offload(skb);
46 	int nhlen = -skb_network_offset(skb);
47 
48 	skb_network_header(skb)[IP6CB(skb)->nhoff] =
49 		XFRM_MODE_SKB_CB(skb)->protocol;
50 
51 #ifndef CONFIG_NETFILTER
52 	if (!async)
53 		return 1;
54 #endif
55 
56 	__skb_push(skb, nhlen);
57 	ipv6_hdr(skb)->payload_len = htons(skb->len - sizeof(struct ipv6hdr));
58 	skb_postpush_rcsum(skb, skb_network_header(skb), nhlen);
59 
60 	if (xo && (xo->flags & XFRM_GRO)) {
61 		/* The full l2 header needs to be preserved so that re-injecting the packet at l2
62 		 * works correctly in the presence of vlan tags.
63 		 */
64 		skb_mac_header_rebuild_full(skb, xo->orig_mac_len);
65 		skb_reset_network_header(skb);
66 		skb_reset_transport_header(skb);
67 		return 0;
68 	}
69 
70 	NF_HOOK(NFPROTO_IPV6, NF_INET_PRE_ROUTING,
71 		dev_net(skb->dev), NULL, skb, skb->dev, NULL,
72 		xfrm6_transport_finish2);
73 	return 0;
74 }
75 
76 static int __xfrm6_udp_encap_rcv(struct sock *sk, struct sk_buff *skb, bool pull)
77 {
78 	struct udp_sock *up = udp_sk(sk);
79 	struct udphdr *uh;
80 	struct ipv6hdr *ip6h;
81 	int len;
82 	int ip6hlen = sizeof(struct ipv6hdr);
83 	__u8 *udpdata;
84 	__be32 *udpdata32;
85 	u16 encap_type;
86 
87 	encap_type = READ_ONCE(up->encap_type);
88 	/* if this is not encapsulated socket, then just return now */
89 	if (!encap_type)
90 		return 1;
91 
92 	/* If this is a paged skb, make sure we pull up
93 	 * whatever data we need to look at. */
94 	len = skb->len - sizeof(struct udphdr);
95 	if (!pskb_may_pull(skb, sizeof(struct udphdr) + min(len, 8)))
96 		return 1;
97 
98 	/* Now we can get the pointers */
99 	uh = udp_hdr(skb);
100 	udpdata = (__u8 *)uh + sizeof(struct udphdr);
101 	udpdata32 = (__be32 *)udpdata;
102 
103 	switch (encap_type) {
104 	default:
105 	case UDP_ENCAP_ESPINUDP:
106 		/* Check if this is a keepalive packet.  If so, eat it. */
107 		if (len == 1 && udpdata[0] == 0xff) {
108 			return -EINVAL;
109 		} else if (len > sizeof(struct ip_esp_hdr) && udpdata32[0] != 0) {
110 			/* ESP Packet without Non-ESP header */
111 			len = sizeof(struct udphdr);
112 		} else
113 			/* Must be an IKE packet.. pass it through */
114 			return 1;
115 		break;
116 	case UDP_ENCAP_ESPINUDP_NON_IKE:
117 		/* Check if this is a keepalive packet.  If so, eat it. */
118 		if (len == 1 && udpdata[0] == 0xff) {
119 			return -EINVAL;
120 		} else if (len > 2 * sizeof(u32) + sizeof(struct ip_esp_hdr) &&
121 			   udpdata32[0] == 0 && udpdata32[1] == 0) {
122 
123 			/* ESP Packet with Non-IKE marker */
124 			len = sizeof(struct udphdr) + 2 * sizeof(u32);
125 		} else
126 			/* Must be an IKE packet.. pass it through */
127 			return 1;
128 		break;
129 	}
130 
131 	/* At this point we are sure that this is an ESPinUDP packet,
132 	 * so we need to remove 'len' bytes from the packet (the UDP
133 	 * header and optional ESP marker bytes) and then modify the
134 	 * protocol to ESP, and then call into the transform receiver.
135 	 */
136 	if (skb_unclone(skb, GFP_ATOMIC))
137 		return -EINVAL;
138 
139 	/* Now we can update and verify the packet length... */
140 	ip6h = ipv6_hdr(skb);
141 	ip6h->payload_len = htons(ntohs(ip6h->payload_len) - len);
142 	if (skb->len < ip6hlen + len) {
143 		/* packet is too small!?! */
144 		return -EINVAL;
145 	}
146 
147 	/* pull the data buffer up to the ESP header and set the
148 	 * transport header to point to ESP.  Keep UDP on the stack
149 	 * for later.
150 	 */
151 	if (pull) {
152 		__skb_pull(skb, len);
153 		skb_reset_transport_header(skb);
154 	} else {
155 		skb_set_transport_header(skb, len);
156 	}
157 
158 	/* process ESP */
159 	return 0;
160 }
161 
162 /* If it's a keepalive packet, then just eat it.
163  * If it's an encapsulated packet, then pass it to the
164  * IPsec xfrm input.
165  * Returns 0 if skb passed to xfrm or was dropped.
166  * Returns >0 if skb should be passed to UDP.
167  * Returns <0 if skb should be resubmitted (-ret is protocol)
168  */
169 int xfrm6_udp_encap_rcv(struct sock *sk, struct sk_buff *skb)
170 {
171 	int ret;
172 
173 	if (skb->protocol == htons(ETH_P_IP))
174 		return xfrm4_udp_encap_rcv(sk, skb);
175 
176 	ret = __xfrm6_udp_encap_rcv(sk, skb, true);
177 	if (!ret)
178 		return xfrm6_rcv_encap(skb, IPPROTO_ESP, 0,
179 				       udp_sk(sk)->encap_type);
180 
181 	if (ret < 0) {
182 		kfree_skb(skb);
183 		return 0;
184 	}
185 
186 	return ret;
187 }
188 
189 struct sk_buff *xfrm6_gro_udp_encap_rcv(struct sock *sk, struct list_head *head,
190 					struct sk_buff *skb)
191 {
192 	int offset = skb_gro_offset(skb);
193 	const struct net_offload *ops;
194 	struct sk_buff *pp = NULL;
195 	int ret;
196 
197 	if (skb->protocol == htons(ETH_P_IP))
198 		return xfrm4_gro_udp_encap_rcv(sk, head, skb);
199 
200 	offset = offset - sizeof(struct udphdr);
201 
202 	if (!pskb_pull(skb, offset))
203 		return NULL;
204 
205 	rcu_read_lock();
206 	ops = rcu_dereference(inet6_offloads[IPPROTO_ESP]);
207 	if (!ops || !ops->callbacks.gro_receive)
208 		goto out;
209 
210 	ret = __xfrm6_udp_encap_rcv(sk, skb, false);
211 	if (ret)
212 		goto out;
213 
214 	skb_push(skb, offset);
215 	NAPI_GRO_CB(skb)->proto = IPPROTO_UDP;
216 
217 	pp = call_gro_receive(ops->callbacks.gro_receive, head, skb);
218 	rcu_read_unlock();
219 
220 	return pp;
221 
222 out:
223 	rcu_read_unlock();
224 	skb_push(skb, offset);
225 	NAPI_GRO_CB(skb)->same_flow = 0;
226 	NAPI_GRO_CB(skb)->flush = 1;
227 
228 	return NULL;
229 }
230 
231 int xfrm6_rcv_tnl(struct sk_buff *skb, struct ip6_tnl *t)
232 {
233 	return xfrm6_rcv_spi(skb, skb_network_header(skb)[IP6CB(skb)->nhoff],
234 			     0, t);
235 }
236 EXPORT_SYMBOL(xfrm6_rcv_tnl);
237 
238 int xfrm6_rcv(struct sk_buff *skb)
239 {
240 	return xfrm6_rcv_tnl(skb, NULL);
241 }
242 EXPORT_SYMBOL(xfrm6_rcv);
243 int xfrm6_input_addr(struct sk_buff *skb, xfrm_address_t *daddr,
244 		     xfrm_address_t *saddr, u8 proto)
245 {
246 	struct net *net = dev_net(skb->dev);
247 	struct xfrm_state *x = NULL;
248 	struct sec_path *sp;
249 	int i = 0;
250 
251 	sp = secpath_set(skb);
252 	if (!sp) {
253 		XFRM_INC_STATS(net, LINUX_MIB_XFRMINERROR);
254 		goto drop;
255 	}
256 
257 	if (1 + sp->len == XFRM_MAX_DEPTH) {
258 		XFRM_INC_STATS(net, LINUX_MIB_XFRMINBUFFERERROR);
259 		goto drop;
260 	}
261 
262 	for (i = 0; i < 3; i++) {
263 		xfrm_address_t *dst, *src;
264 
265 		switch (i) {
266 		case 0:
267 			dst = daddr;
268 			src = saddr;
269 			break;
270 		case 1:
271 			/* lookup state with wild-card source address */
272 			dst = daddr;
273 			src = (xfrm_address_t *)&in6addr_any;
274 			break;
275 		default:
276 			/* lookup state with wild-card addresses */
277 			dst = (xfrm_address_t *)&in6addr_any;
278 			src = (xfrm_address_t *)&in6addr_any;
279 			break;
280 		}
281 
282 		x = xfrm_state_lookup_byaddr(net, skb->mark, dst, src, proto, AF_INET6);
283 		if (!x)
284 			continue;
285 
286 		spin_lock(&x->lock);
287 
288 		if ((!i || (x->props.flags & XFRM_STATE_WILDRECV)) &&
289 		    likely(x->km.state == XFRM_STATE_VALID) &&
290 		    !xfrm_state_check_expire(x)) {
291 			spin_unlock(&x->lock);
292 			if (x->type->input(x, skb) > 0) {
293 				/* found a valid state */
294 				break;
295 			}
296 		} else
297 			spin_unlock(&x->lock);
298 
299 		xfrm_state_put(x);
300 		x = NULL;
301 	}
302 
303 	if (!x) {
304 		XFRM_INC_STATS(net, LINUX_MIB_XFRMINNOSTATES);
305 		xfrm_audit_state_notfound_simple(skb, AF_INET6);
306 		goto drop;
307 	}
308 
309 	sp->xvec[sp->len++] = x;
310 
311 	spin_lock(&x->lock);
312 
313 	x->curlft.bytes += skb->len;
314 	x->curlft.packets++;
315 
316 	spin_unlock(&x->lock);
317 
318 	return 1;
319 
320 drop:
321 	return -1;
322 }
323 EXPORT_SYMBOL(xfrm6_input_addr);
324