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, encap_type: 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, finish: 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->data - skb_network_header(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, len: nhlen); |
57 | ipv6_hdr(skb)->payload_len = htons(skb->len - sizeof(struct ipv6hdr)); |
58 | skb_postpush_rcsum(skb, start: skb_network_header(skb), len: nhlen); |
59 | |
60 | if (xo && (xo->flags & XFRM_GRO)) { |
61 | skb_mac_header_rebuild(skb); |
62 | skb_reset_transport_header(skb); |
63 | return 0; |
64 | } |
65 | |
66 | NF_HOOK(pf: NFPROTO_IPV6, hook: NF_INET_PRE_ROUTING, |
67 | net: dev_net(dev: skb->dev), NULL, skb, in: skb->dev, NULL, |
68 | okfn: xfrm6_transport_finish2); |
69 | return 0; |
70 | } |
71 | |
72 | static int __xfrm6_udp_encap_rcv(struct sock *sk, struct sk_buff *skb, bool pull) |
73 | { |
74 | struct udp_sock *up = udp_sk(sk); |
75 | struct udphdr *uh; |
76 | struct ipv6hdr *ip6h; |
77 | int len; |
78 | int ip6hlen = sizeof(struct ipv6hdr); |
79 | __u8 *udpdata; |
80 | __be32 *udpdata32; |
81 | u16 encap_type; |
82 | |
83 | encap_type = READ_ONCE(up->encap_type); |
84 | /* if this is not encapsulated socket, then just return now */ |
85 | if (!encap_type) |
86 | return 1; |
87 | |
88 | /* If this is a paged skb, make sure we pull up |
89 | * whatever data we need to look at. */ |
90 | len = skb->len - sizeof(struct udphdr); |
91 | if (!pskb_may_pull(skb, len: sizeof(struct udphdr) + min(len, 8))) |
92 | return 1; |
93 | |
94 | /* Now we can get the pointers */ |
95 | uh = udp_hdr(skb); |
96 | udpdata = (__u8 *)uh + sizeof(struct udphdr); |
97 | udpdata32 = (__be32 *)udpdata; |
98 | |
99 | switch (encap_type) { |
100 | default: |
101 | case UDP_ENCAP_ESPINUDP: |
102 | /* Check if this is a keepalive packet. If so, eat it. */ |
103 | if (len == 1 && udpdata[0] == 0xff) { |
104 | return -EINVAL; |
105 | } else if (len > sizeof(struct ip_esp_hdr) && udpdata32[0] != 0) { |
106 | /* ESP Packet without Non-ESP header */ |
107 | len = sizeof(struct udphdr); |
108 | } else |
109 | /* Must be an IKE packet.. pass it through */ |
110 | return 1; |
111 | break; |
112 | case UDP_ENCAP_ESPINUDP_NON_IKE: |
113 | /* Check if this is a keepalive packet. If so, eat it. */ |
114 | if (len == 1 && udpdata[0] == 0xff) { |
115 | return -EINVAL; |
116 | } else if (len > 2 * sizeof(u32) + sizeof(struct ip_esp_hdr) && |
117 | udpdata32[0] == 0 && udpdata32[1] == 0) { |
118 | |
119 | /* ESP Packet with Non-IKE marker */ |
120 | len = sizeof(struct udphdr) + 2 * sizeof(u32); |
121 | } else |
122 | /* Must be an IKE packet.. pass it through */ |
123 | return 1; |
124 | break; |
125 | } |
126 | |
127 | /* At this point we are sure that this is an ESPinUDP packet, |
128 | * so we need to remove 'len' bytes from the packet (the UDP |
129 | * header and optional ESP marker bytes) and then modify the |
130 | * protocol to ESP, and then call into the transform receiver. |
131 | */ |
132 | if (skb_unclone(skb, GFP_ATOMIC)) |
133 | return -EINVAL; |
134 | |
135 | /* Now we can update and verify the packet length... */ |
136 | ip6h = ipv6_hdr(skb); |
137 | ip6h->payload_len = htons(ntohs(ip6h->payload_len) - len); |
138 | if (skb->len < ip6hlen + len) { |
139 | /* packet is too small!?! */ |
140 | return -EINVAL; |
141 | } |
142 | |
143 | /* pull the data buffer up to the ESP header and set the |
144 | * transport header to point to ESP. Keep UDP on the stack |
145 | * for later. |
146 | */ |
147 | if (pull) { |
148 | __skb_pull(skb, len); |
149 | skb_reset_transport_header(skb); |
150 | } else { |
151 | skb_set_transport_header(skb, offset: len); |
152 | } |
153 | |
154 | /* process ESP */ |
155 | return 0; |
156 | } |
157 | |
158 | /* If it's a keepalive packet, then just eat it. |
159 | * If it's an encapsulated packet, then pass it to the |
160 | * IPsec xfrm input. |
161 | * Returns 0 if skb passed to xfrm or was dropped. |
162 | * Returns >0 if skb should be passed to UDP. |
163 | * Returns <0 if skb should be resubmitted (-ret is protocol) |
164 | */ |
165 | int xfrm6_udp_encap_rcv(struct sock *sk, struct sk_buff *skb) |
166 | { |
167 | int ret; |
168 | |
169 | if (skb->protocol == htons(ETH_P_IP)) |
170 | return xfrm4_udp_encap_rcv(sk, skb); |
171 | |
172 | ret = __xfrm6_udp_encap_rcv(sk, skb, pull: true); |
173 | if (!ret) |
174 | return xfrm6_rcv_encap(skb, IPPROTO_ESP, spi: 0, |
175 | udp_sk(sk)->encap_type); |
176 | |
177 | if (ret < 0) { |
178 | kfree_skb(skb); |
179 | return 0; |
180 | } |
181 | |
182 | return ret; |
183 | } |
184 | |
185 | struct sk_buff *xfrm6_gro_udp_encap_rcv(struct sock *sk, struct list_head *head, |
186 | struct sk_buff *skb) |
187 | { |
188 | int offset = skb_gro_offset(skb); |
189 | const struct net_offload *ops; |
190 | struct sk_buff *pp = NULL; |
191 | int ret; |
192 | |
193 | if (skb->protocol == htons(ETH_P_IP)) |
194 | return xfrm4_gro_udp_encap_rcv(sk, head, skb); |
195 | |
196 | offset = offset - sizeof(struct udphdr); |
197 | |
198 | if (!pskb_pull(skb, len: offset)) |
199 | return NULL; |
200 | |
201 | rcu_read_lock(); |
202 | ops = rcu_dereference(inet6_offloads[IPPROTO_ESP]); |
203 | if (!ops || !ops->callbacks.gro_receive) |
204 | goto out; |
205 | |
206 | ret = __xfrm6_udp_encap_rcv(sk, skb, pull: false); |
207 | if (ret) |
208 | goto out; |
209 | |
210 | skb_push(skb, len: offset); |
211 | NAPI_GRO_CB(skb)->proto = IPPROTO_UDP; |
212 | |
213 | pp = call_gro_receive(cb: ops->callbacks.gro_receive, head, skb); |
214 | rcu_read_unlock(); |
215 | |
216 | return pp; |
217 | |
218 | out: |
219 | rcu_read_unlock(); |
220 | skb_push(skb, len: offset); |
221 | NAPI_GRO_CB(skb)->same_flow = 0; |
222 | NAPI_GRO_CB(skb)->flush = 1; |
223 | |
224 | return NULL; |
225 | } |
226 | |
227 | int xfrm6_rcv_tnl(struct sk_buff *skb, struct ip6_tnl *t) |
228 | { |
229 | return xfrm6_rcv_spi(skb, skb_network_header(skb)[IP6CB(skb)->nhoff], |
230 | 0, t); |
231 | } |
232 | EXPORT_SYMBOL(xfrm6_rcv_tnl); |
233 | |
234 | int xfrm6_rcv(struct sk_buff *skb) |
235 | { |
236 | return xfrm6_rcv_tnl(skb, NULL); |
237 | } |
238 | EXPORT_SYMBOL(xfrm6_rcv); |
239 | int xfrm6_input_addr(struct sk_buff *skb, xfrm_address_t *daddr, |
240 | xfrm_address_t *saddr, u8 proto) |
241 | { |
242 | struct net *net = dev_net(dev: skb->dev); |
243 | struct xfrm_state *x = NULL; |
244 | struct sec_path *sp; |
245 | int i = 0; |
246 | |
247 | sp = secpath_set(skb); |
248 | if (!sp) { |
249 | XFRM_INC_STATS(net, LINUX_MIB_XFRMINERROR); |
250 | goto drop; |
251 | } |
252 | |
253 | if (1 + sp->len == XFRM_MAX_DEPTH) { |
254 | XFRM_INC_STATS(net, LINUX_MIB_XFRMINBUFFERERROR); |
255 | goto drop; |
256 | } |
257 | |
258 | for (i = 0; i < 3; i++) { |
259 | xfrm_address_t *dst, *src; |
260 | |
261 | switch (i) { |
262 | case 0: |
263 | dst = daddr; |
264 | src = saddr; |
265 | break; |
266 | case 1: |
267 | /* lookup state with wild-card source address */ |
268 | dst = daddr; |
269 | src = (xfrm_address_t *)&in6addr_any; |
270 | break; |
271 | default: |
272 | /* lookup state with wild-card addresses */ |
273 | dst = (xfrm_address_t *)&in6addr_any; |
274 | src = (xfrm_address_t *)&in6addr_any; |
275 | break; |
276 | } |
277 | |
278 | x = xfrm_state_lookup_byaddr(net, mark: skb->mark, daddr: dst, saddr: src, proto, AF_INET6); |
279 | if (!x) |
280 | continue; |
281 | |
282 | spin_lock(lock: &x->lock); |
283 | |
284 | if ((!i || (x->props.flags & XFRM_STATE_WILDRECV)) && |
285 | likely(x->km.state == XFRM_STATE_VALID) && |
286 | !xfrm_state_check_expire(x)) { |
287 | spin_unlock(lock: &x->lock); |
288 | if (x->type->input(x, skb) > 0) { |
289 | /* found a valid state */ |
290 | break; |
291 | } |
292 | } else |
293 | spin_unlock(lock: &x->lock); |
294 | |
295 | xfrm_state_put(x); |
296 | x = NULL; |
297 | } |
298 | |
299 | if (!x) { |
300 | XFRM_INC_STATS(net, LINUX_MIB_XFRMINNOSTATES); |
301 | xfrm_audit_state_notfound_simple(skb, AF_INET6); |
302 | goto drop; |
303 | } |
304 | |
305 | sp->xvec[sp->len++] = x; |
306 | |
307 | spin_lock(lock: &x->lock); |
308 | |
309 | x->curlft.bytes += skb->len; |
310 | x->curlft.packets++; |
311 | |
312 | spin_unlock(lock: &x->lock); |
313 | |
314 | return 1; |
315 | |
316 | drop: |
317 | return -1; |
318 | } |
319 | EXPORT_SYMBOL(xfrm6_input_addr); |
320 | |