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
xfrm6_rcv_spi(struct sk_buff * skb,int nexthdr,__be32 spi,struct ip6_tnl * t)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
xfrm6_transport_finish2(struct net * net,struct sock * sk,struct sk_buff * skb)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
xfrm6_transport_finish(struct sk_buff * skb,int async)43 int xfrm6_transport_finish(struct sk_buff *skb, int async)
44 {
45 struct xfrm_offload *xo = xfrm_offload(skb);
46 struct net_device *dev = skb->dev;
47 int nhlen = -skb_network_offset(skb);
48
49 skb_network_header(skb)[IP6CB(skb)->nhoff] =
50 XFRM_MODE_SKB_CB(skb)->protocol;
51
52 #ifndef CONFIG_NETFILTER
53 if (!async)
54 return 1;
55 #endif
56
57 __skb_push(skb, nhlen);
58 ipv6_hdr(skb)->payload_len = htons(skb->len - sizeof(struct ipv6hdr));
59 skb_postpush_rcsum(skb, skb_network_header(skb), nhlen);
60
61 if (xo && (xo->flags & XFRM_GRO)) {
62 /* The full l2 header needs to be preserved so that re-injecting the packet at l2
63 * works correctly in the presence of vlan tags.
64 */
65 skb_mac_header_rebuild_full(skb, xo->orig_mac_len);
66 skb_reset_network_header(skb);
67 skb_reset_transport_header(skb);
68 return 0;
69 }
70
71 NF_HOOK(NFPROTO_IPV6, NF_INET_PRE_ROUTING,
72 dev_net(dev), NULL, skb, dev, NULL,
73 xfrm6_transport_finish2);
74 if (async)
75 dev_put(dev);
76 return 0;
77 }
78
__xfrm6_udp_encap_rcv(struct sock * sk,struct sk_buff * skb,bool pull)79 static int __xfrm6_udp_encap_rcv(struct sock *sk, struct sk_buff *skb, bool pull)
80 {
81 struct udp_sock *up = udp_sk(sk);
82 struct udphdr *uh;
83 struct ipv6hdr *ip6h;
84 int len;
85 int ip6hlen = sizeof(struct ipv6hdr);
86 __u8 *udpdata;
87 __be32 *udpdata32;
88 u16 encap_type;
89
90 encap_type = READ_ONCE(up->encap_type);
91 /* if this is not encapsulated socket, then just return now */
92 if (!encap_type)
93 return 1;
94
95 /* If this is a paged skb, make sure we pull up
96 * whatever data we need to look at. */
97 len = skb->len - sizeof(struct udphdr);
98 if (!pskb_may_pull(skb, sizeof(struct udphdr) + min(len, 8)))
99 return 1;
100
101 /* Now we can get the pointers */
102 uh = udp_hdr(skb);
103 udpdata = (__u8 *)uh + sizeof(struct udphdr);
104 udpdata32 = (__be32 *)udpdata;
105
106 switch (encap_type) {
107 default:
108 case UDP_ENCAP_ESPINUDP:
109 /* Check if this is a keepalive packet. If so, eat it. */
110 if (len == 1 && udpdata[0] == 0xff) {
111 return -EINVAL;
112 } else if (len > sizeof(struct ip_esp_hdr) && udpdata32[0] != 0) {
113 /* ESP Packet without Non-ESP header */
114 len = sizeof(struct udphdr);
115 } else
116 /* Must be an IKE packet.. pass it through */
117 return 1;
118 break;
119 }
120
121 /* At this point we are sure that this is an ESPinUDP packet,
122 * so we need to remove 'len' bytes from the packet (the UDP
123 * header and optional ESP marker bytes) and then modify the
124 * protocol to ESP, and then call into the transform receiver.
125 */
126 if (skb_unclone(skb, GFP_ATOMIC))
127 return -EINVAL;
128
129 /* Now we can update and verify the packet length... */
130 ip6h = ipv6_hdr(skb);
131 ip6h->payload_len = htons(ntohs(ip6h->payload_len) - len);
132 if (skb->len < ip6hlen + len) {
133 /* packet is too small!?! */
134 return -EINVAL;
135 }
136
137 /* pull the data buffer up to the ESP header and set the
138 * transport header to point to ESP. Keep UDP on the stack
139 * for later.
140 */
141 if (pull) {
142 __skb_pull(skb, len);
143 skb_reset_transport_header(skb);
144 } else {
145 skb_set_transport_header(skb, len);
146 }
147
148 /* process ESP */
149 return 0;
150 }
151
152 /* If it's a keepalive packet, then just eat it.
153 * If it's an encapsulated packet, then pass it to the
154 * IPsec xfrm input.
155 * Returns 0 if skb passed to xfrm or was dropped.
156 * Returns >0 if skb should be passed to UDP.
157 * Returns <0 if skb should be resubmitted (-ret is protocol)
158 */
xfrm6_udp_encap_rcv(struct sock * sk,struct sk_buff * skb)159 int xfrm6_udp_encap_rcv(struct sock *sk, struct sk_buff *skb)
160 {
161 int ret;
162
163 if (skb->protocol == htons(ETH_P_IP))
164 return xfrm4_udp_encap_rcv(sk, skb);
165
166 ret = __xfrm6_udp_encap_rcv(sk, skb, true);
167 if (!ret)
168 return xfrm6_rcv_encap(skb, IPPROTO_ESP, 0,
169 udp_sk(sk)->encap_type);
170
171 if (ret < 0) {
172 kfree_skb(skb);
173 return 0;
174 }
175
176 return ret;
177 }
178
xfrm6_gro_udp_encap_rcv(struct sock * sk,struct list_head * head,struct sk_buff * skb)179 struct sk_buff *xfrm6_gro_udp_encap_rcv(struct sock *sk, struct list_head *head,
180 struct sk_buff *skb)
181 {
182 int offset = skb_gro_offset(skb);
183 const struct net_offload *ops;
184 struct sk_buff *pp = NULL;
185 int len, dlen;
186 __u8 *udpdata;
187 __be32 *udpdata32;
188
189 if (skb->protocol == htons(ETH_P_IP))
190 return xfrm4_gro_udp_encap_rcv(sk, head, skb);
191
192 len = skb->len - offset;
193 dlen = offset + min(len, 8);
194 udpdata = skb_gro_header(skb, dlen, offset);
195 udpdata32 = (__be32 *)udpdata;
196 if (unlikely(!udpdata))
197 return NULL;
198
199 rcu_read_lock();
200 ops = rcu_dereference(inet6_offloads[IPPROTO_ESP]);
201 if (!ops || !ops->callbacks.gro_receive)
202 goto out;
203
204 /* check if it is a keepalive or IKE packet */
205 if (len <= sizeof(struct ip_esp_hdr) || udpdata32[0] == 0)
206 goto out;
207
208 /* set the transport header to ESP */
209 skb_set_transport_header(skb, offset);
210
211 NAPI_GRO_CB(skb)->proto = IPPROTO_UDP;
212
213 pp = call_gro_receive(ops->callbacks.gro_receive, head, skb);
214 rcu_read_unlock();
215
216 return pp;
217
218 out:
219 rcu_read_unlock();
220 NAPI_GRO_CB(skb)->same_flow = 0;
221 NAPI_GRO_CB(skb)->flush = 1;
222
223 return NULL;
224 }
225
xfrm6_rcv_tnl(struct sk_buff * skb,struct ip6_tnl * t)226 int xfrm6_rcv_tnl(struct sk_buff *skb, struct ip6_tnl *t)
227 {
228 return xfrm6_rcv_spi(skb, skb_network_header(skb)[IP6CB(skb)->nhoff],
229 0, t);
230 }
231 EXPORT_SYMBOL(xfrm6_rcv_tnl);
232
xfrm6_rcv(struct sk_buff * skb)233 int xfrm6_rcv(struct sk_buff *skb)
234 {
235 return xfrm6_rcv_tnl(skb, NULL);
236 }
237 EXPORT_SYMBOL(xfrm6_rcv);
xfrm6_input_addr(struct sk_buff * skb,xfrm_address_t * daddr,xfrm_address_t * saddr,u8 proto)238 int xfrm6_input_addr(struct sk_buff *skb, xfrm_address_t *daddr,
239 xfrm_address_t *saddr, u8 proto)
240 {
241 struct net *net = dev_net(skb->dev);
242 struct xfrm_state *x = NULL;
243 struct sec_path *sp;
244 int i = 0;
245
246 sp = secpath_set(skb);
247 if (!sp) {
248 XFRM_INC_STATS(net, LINUX_MIB_XFRMINERROR);
249 goto drop;
250 }
251
252 if (1 + sp->len == XFRM_MAX_DEPTH) {
253 XFRM_INC_STATS(net, LINUX_MIB_XFRMINBUFFERERROR);
254 goto drop;
255 }
256
257 for (i = 0; i < 3; i++) {
258 xfrm_address_t *dst, *src;
259
260 switch (i) {
261 case 0:
262 dst = daddr;
263 src = saddr;
264 break;
265 case 1:
266 /* lookup state with wild-card source address */
267 dst = daddr;
268 src = (xfrm_address_t *)&in6addr_any;
269 break;
270 default:
271 /* lookup state with wild-card addresses */
272 dst = (xfrm_address_t *)&in6addr_any;
273 src = (xfrm_address_t *)&in6addr_any;
274 break;
275 }
276
277 x = xfrm_state_lookup_byaddr(net, skb->mark, dst, src, proto, AF_INET6);
278 if (!x)
279 continue;
280
281 if (unlikely(x->dir && x->dir != XFRM_SA_DIR_IN)) {
282 XFRM_INC_STATS(net, LINUX_MIB_XFRMINSTATEDIRERROR);
283 xfrm_state_put(x);
284 x = NULL;
285 continue;
286 }
287
288 spin_lock(&x->lock);
289
290 if ((!i || (x->props.flags & XFRM_STATE_WILDRECV)) &&
291 likely(x->km.state == XFRM_STATE_VALID) &&
292 !xfrm_state_check_expire(x)) {
293 spin_unlock(&x->lock);
294 if (x->type->input(x, skb) > 0) {
295 /* found a valid state */
296 break;
297 }
298 } else
299 spin_unlock(&x->lock);
300
301 xfrm_state_put(x);
302 x = NULL;
303 }
304
305 if (!x) {
306 XFRM_INC_STATS(net, LINUX_MIB_XFRMINNOSTATES);
307 xfrm_audit_state_notfound_simple(skb, AF_INET6);
308 goto drop;
309 }
310
311 sp->xvec[sp->len++] = x;
312
313 spin_lock(&x->lock);
314
315 x->curlft.bytes += skb->len;
316 x->curlft.packets++;
317
318 spin_unlock(&x->lock);
319
320 return 1;
321
322 drop:
323 return -1;
324 }
325 EXPORT_SYMBOL(xfrm6_input_addr);
326