xfrm: Add support for IPsec extended sequence numbers
[linux-2.6.git] / net / xfrm / xfrm_input.c
1 /*
2  * xfrm_input.c
3  *
4  * Changes:
5  *      YOSHIFUJI Hideaki @USAGI
6  *              Split up af-specific portion
7  *
8  */
9
10 #include <linux/slab.h>
11 #include <linux/module.h>
12 #include <linux/netdevice.h>
13 #include <net/dst.h>
14 #include <net/ip.h>
15 #include <net/xfrm.h>
16
17 static struct kmem_cache *secpath_cachep __read_mostly;
18
19 void __secpath_destroy(struct sec_path *sp)
20 {
21         int i;
22         for (i = 0; i < sp->len; i++)
23                 xfrm_state_put(sp->xvec[i]);
24         kmem_cache_free(secpath_cachep, sp);
25 }
26 EXPORT_SYMBOL(__secpath_destroy);
27
28 struct sec_path *secpath_dup(struct sec_path *src)
29 {
30         struct sec_path *sp;
31
32         sp = kmem_cache_alloc(secpath_cachep, GFP_ATOMIC);
33         if (!sp)
34                 return NULL;
35
36         sp->len = 0;
37         if (src) {
38                 int i;
39
40                 memcpy(sp, src, sizeof(*sp));
41                 for (i = 0; i < sp->len; i++)
42                         xfrm_state_hold(sp->xvec[i]);
43         }
44         atomic_set(&sp->refcnt, 1);
45         return sp;
46 }
47 EXPORT_SYMBOL(secpath_dup);
48
49 /* Fetch spi and seq from ipsec header */
50
51 int xfrm_parse_spi(struct sk_buff *skb, u8 nexthdr, __be32 *spi, __be32 *seq)
52 {
53         int offset, offset_seq;
54         int hlen;
55
56         switch (nexthdr) {
57         case IPPROTO_AH:
58                 hlen = sizeof(struct ip_auth_hdr);
59                 offset = offsetof(struct ip_auth_hdr, spi);
60                 offset_seq = offsetof(struct ip_auth_hdr, seq_no);
61                 break;
62         case IPPROTO_ESP:
63                 hlen = sizeof(struct ip_esp_hdr);
64                 offset = offsetof(struct ip_esp_hdr, spi);
65                 offset_seq = offsetof(struct ip_esp_hdr, seq_no);
66                 break;
67         case IPPROTO_COMP:
68                 if (!pskb_may_pull(skb, sizeof(struct ip_comp_hdr)))
69                         return -EINVAL;
70                 *spi = htonl(ntohs(*(__be16*)(skb_transport_header(skb) + 2)));
71                 *seq = 0;
72                 return 0;
73         default:
74                 return 1;
75         }
76
77         if (!pskb_may_pull(skb, hlen))
78                 return -EINVAL;
79
80         *spi = *(__be32*)(skb_transport_header(skb) + offset);
81         *seq = *(__be32*)(skb_transport_header(skb) + offset_seq);
82         return 0;
83 }
84
85 int xfrm_prepare_input(struct xfrm_state *x, struct sk_buff *skb)
86 {
87         struct xfrm_mode *inner_mode = x->inner_mode;
88         int err;
89
90         err = x->outer_mode->afinfo->extract_input(x, skb);
91         if (err)
92                 return err;
93
94         if (x->sel.family == AF_UNSPEC) {
95                 inner_mode = xfrm_ip2inner_mode(x, XFRM_MODE_SKB_CB(skb)->protocol);
96                 if (inner_mode == NULL)
97                         return -EAFNOSUPPORT;
98         }
99
100         skb->protocol = inner_mode->afinfo->eth_proto;
101         return inner_mode->input2(x, skb);
102 }
103 EXPORT_SYMBOL(xfrm_prepare_input);
104
105 int xfrm_input(struct sk_buff *skb, int nexthdr, __be32 spi, int encap_type)
106 {
107         struct net *net = dev_net(skb->dev);
108         int err;
109         __be32 seq;
110         __be32 seq_hi;
111         struct xfrm_state *x;
112         xfrm_address_t *daddr;
113         struct xfrm_mode *inner_mode;
114         unsigned int family;
115         int decaps = 0;
116         int async = 0;
117
118         /* A negative encap_type indicates async resumption. */
119         if (encap_type < 0) {
120                 async = 1;
121                 x = xfrm_input_state(skb);
122                 seq = XFRM_SKB_CB(skb)->seq.input.low;
123                 goto resume;
124         }
125
126         /* Allocate new secpath or COW existing one. */
127         if (!skb->sp || atomic_read(&skb->sp->refcnt) != 1) {
128                 struct sec_path *sp;
129
130                 sp = secpath_dup(skb->sp);
131                 if (!sp) {
132                         XFRM_INC_STATS(net, LINUX_MIB_XFRMINERROR);
133                         goto drop;
134                 }
135                 if (skb->sp)
136                         secpath_put(skb->sp);
137                 skb->sp = sp;
138         }
139
140         daddr = (xfrm_address_t *)(skb_network_header(skb) +
141                                    XFRM_SPI_SKB_CB(skb)->daddroff);
142         family = XFRM_SPI_SKB_CB(skb)->family;
143
144         seq = 0;
145         if (!spi && (err = xfrm_parse_spi(skb, nexthdr, &spi, &seq)) != 0) {
146                 XFRM_INC_STATS(net, LINUX_MIB_XFRMINHDRERROR);
147                 goto drop;
148         }
149
150         do {
151                 if (skb->sp->len == XFRM_MAX_DEPTH) {
152                         XFRM_INC_STATS(net, LINUX_MIB_XFRMINBUFFERERROR);
153                         goto drop;
154                 }
155
156                 x = xfrm_state_lookup(net, skb->mark, daddr, spi, nexthdr, family);
157                 if (x == NULL) {
158                         XFRM_INC_STATS(net, LINUX_MIB_XFRMINNOSTATES);
159                         xfrm_audit_state_notfound(skb, family, spi, seq);
160                         goto drop;
161                 }
162
163                 skb->sp->xvec[skb->sp->len++] = x;
164
165                 spin_lock(&x->lock);
166                 if (unlikely(x->km.state != XFRM_STATE_VALID)) {
167                         XFRM_INC_STATS(net, LINUX_MIB_XFRMINSTATEINVALID);
168                         goto drop_unlock;
169                 }
170
171                 if ((x->encap ? x->encap->encap_type : 0) != encap_type) {
172                         XFRM_INC_STATS(net, LINUX_MIB_XFRMINSTATEMISMATCH);
173                         goto drop_unlock;
174                 }
175
176                 if (x->props.replay_window && x->repl->check(x, skb, seq)) {
177                         XFRM_INC_STATS(net, LINUX_MIB_XFRMINSTATESEQERROR);
178                         goto drop_unlock;
179                 }
180
181                 if (xfrm_state_check_expire(x)) {
182                         XFRM_INC_STATS(net, LINUX_MIB_XFRMINSTATEEXPIRED);
183                         goto drop_unlock;
184                 }
185
186                 spin_unlock(&x->lock);
187
188                 seq_hi = htonl(xfrm_replay_seqhi(x, seq));
189
190                 XFRM_SKB_CB(skb)->seq.input.low = seq;
191                 XFRM_SKB_CB(skb)->seq.input.hi = seq_hi;
192
193                 nexthdr = x->type->input(x, skb);
194
195                 if (nexthdr == -EINPROGRESS)
196                         return 0;
197
198 resume:
199                 spin_lock(&x->lock);
200                 if (nexthdr <= 0) {
201                         if (nexthdr == -EBADMSG) {
202                                 xfrm_audit_state_icvfail(x, skb,
203                                                          x->type->proto);
204                                 x->stats.integrity_failed++;
205                         }
206                         XFRM_INC_STATS(net, LINUX_MIB_XFRMINSTATEPROTOERROR);
207                         goto drop_unlock;
208                 }
209
210                 /* only the first xfrm gets the encap type */
211                 encap_type = 0;
212
213                 x->repl->advance(x, seq);
214
215                 x->curlft.bytes += skb->len;
216                 x->curlft.packets++;
217
218                 spin_unlock(&x->lock);
219
220                 XFRM_MODE_SKB_CB(skb)->protocol = nexthdr;
221
222                 inner_mode = x->inner_mode;
223
224                 if (x->sel.family == AF_UNSPEC) {
225                         inner_mode = xfrm_ip2inner_mode(x, XFRM_MODE_SKB_CB(skb)->protocol);
226                         if (inner_mode == NULL)
227                                 goto drop;
228                 }
229
230                 if (inner_mode->input(x, skb)) {
231                         XFRM_INC_STATS(net, LINUX_MIB_XFRMINSTATEMODEERROR);
232                         goto drop;
233                 }
234
235                 if (x->outer_mode->flags & XFRM_MODE_FLAG_TUNNEL) {
236                         decaps = 1;
237                         break;
238                 }
239
240                 /*
241                  * We need the inner address.  However, we only get here for
242                  * transport mode so the outer address is identical.
243                  */
244                 daddr = &x->id.daddr;
245                 family = x->outer_mode->afinfo->family;
246
247                 err = xfrm_parse_spi(skb, nexthdr, &spi, &seq);
248                 if (err < 0) {
249                         XFRM_INC_STATS(net, LINUX_MIB_XFRMINHDRERROR);
250                         goto drop;
251                 }
252         } while (!err);
253
254         nf_reset(skb);
255
256         if (decaps) {
257                 skb_dst_drop(skb);
258                 netif_rx(skb);
259                 return 0;
260         } else {
261                 return x->inner_mode->afinfo->transport_finish(skb, async);
262         }
263
264 drop_unlock:
265         spin_unlock(&x->lock);
266 drop:
267         kfree_skb(skb);
268         return 0;
269 }
270 EXPORT_SYMBOL(xfrm_input);
271
272 int xfrm_input_resume(struct sk_buff *skb, int nexthdr)
273 {
274         return xfrm_input(skb, nexthdr, 0, -1);
275 }
276 EXPORT_SYMBOL(xfrm_input_resume);
277
278 void __init xfrm_input_init(void)
279 {
280         secpath_cachep = kmem_cache_create("secpath_cache",
281                                            sizeof(struct sec_path),
282                                            0, SLAB_HWCACHE_ALIGN|SLAB_PANIC,
283                                            NULL);
284 }