Merge git://git.kernel.org/pub/scm/linux/kernel/git/bart/ide-2.6
[linux-2.6] / 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         int err;
108         __be32 seq;
109         struct xfrm_state *x;
110         xfrm_address_t *daddr;
111         struct xfrm_mode *inner_mode;
112         unsigned int family;
113         int decaps = 0;
114         int async = 0;
115
116         /* A negative encap_type indicates async resumption. */
117         if (encap_type < 0) {
118                 async = 1;
119                 x = xfrm_input_state(skb);
120                 seq = XFRM_SKB_CB(skb)->seq.input;
121                 goto resume;
122         }
123
124         /* Allocate new secpath or COW existing one. */
125         if (!skb->sp || atomic_read(&skb->sp->refcnt) != 1) {
126                 struct sec_path *sp;
127
128                 sp = secpath_dup(skb->sp);
129                 if (!sp) {
130                         XFRM_INC_STATS(LINUX_MIB_XFRMINERROR);
131                         goto drop;
132                 }
133                 if (skb->sp)
134                         secpath_put(skb->sp);
135                 skb->sp = sp;
136         }
137
138         daddr = (xfrm_address_t *)(skb_network_header(skb) +
139                                    XFRM_SPI_SKB_CB(skb)->daddroff);
140         family = XFRM_SPI_SKB_CB(skb)->family;
141
142         seq = 0;
143         if (!spi && (err = xfrm_parse_spi(skb, nexthdr, &spi, &seq)) != 0) {
144                 XFRM_INC_STATS(LINUX_MIB_XFRMINHDRERROR);
145                 goto drop;
146         }
147
148         do {
149                 if (skb->sp->len == XFRM_MAX_DEPTH) {
150                         XFRM_INC_STATS(LINUX_MIB_XFRMINBUFFERERROR);
151                         goto drop;
152                 }
153
154                 x = xfrm_state_lookup(daddr, spi, nexthdr, family);
155                 if (x == NULL) {
156                         XFRM_INC_STATS(LINUX_MIB_XFRMINNOSTATES);
157                         xfrm_audit_state_notfound(skb, family, spi, seq);
158                         goto drop;
159                 }
160
161                 skb->sp->xvec[skb->sp->len++] = x;
162
163                 spin_lock(&x->lock);
164                 if (unlikely(x->km.state != XFRM_STATE_VALID)) {
165                         XFRM_INC_STATS(LINUX_MIB_XFRMINSTATEINVALID);
166                         goto drop_unlock;
167                 }
168
169                 if ((x->encap ? x->encap->encap_type : 0) != encap_type) {
170                         XFRM_INC_STATS(LINUX_MIB_XFRMINSTATEMISMATCH);
171                         goto drop_unlock;
172                 }
173
174                 if (x->props.replay_window && xfrm_replay_check(x, skb, seq)) {
175                         XFRM_INC_STATS(LINUX_MIB_XFRMINSTATESEQERROR);
176                         goto drop_unlock;
177                 }
178
179                 if (xfrm_state_check_expire(x)) {
180                         XFRM_INC_STATS(LINUX_MIB_XFRMINSTATEEXPIRED);
181                         goto drop_unlock;
182                 }
183
184                 spin_unlock(&x->lock);
185
186                 XFRM_SKB_CB(skb)->seq.input = seq;
187
188                 nexthdr = x->type->input(x, skb);
189
190                 if (nexthdr == -EINPROGRESS)
191                         return 0;
192
193 resume:
194                 spin_lock(&x->lock);
195                 if (nexthdr <= 0) {
196                         if (nexthdr == -EBADMSG) {
197                                 xfrm_audit_state_icvfail(x, skb,
198                                                          x->type->proto);
199                                 x->stats.integrity_failed++;
200                         }
201                         XFRM_INC_STATS(LINUX_MIB_XFRMINSTATEPROTOERROR);
202                         goto drop_unlock;
203                 }
204
205                 /* only the first xfrm gets the encap type */
206                 encap_type = 0;
207
208                 if (x->props.replay_window)
209                         xfrm_replay_advance(x, seq);
210
211                 x->curlft.bytes += skb->len;
212                 x->curlft.packets++;
213
214                 spin_unlock(&x->lock);
215
216                 XFRM_MODE_SKB_CB(skb)->protocol = nexthdr;
217
218                 inner_mode = x->inner_mode;
219
220                 if (x->sel.family == AF_UNSPEC) {
221                         inner_mode = xfrm_ip2inner_mode(x, XFRM_MODE_SKB_CB(skb)->protocol);
222                         if (inner_mode == NULL)
223                                 goto drop;
224                 }
225
226                 if (inner_mode->input(x, skb)) {
227                         XFRM_INC_STATS(LINUX_MIB_XFRMINSTATEMODEERROR);
228                         goto drop;
229                 }
230
231                 if (x->outer_mode->flags & XFRM_MODE_FLAG_TUNNEL) {
232                         decaps = 1;
233                         break;
234                 }
235
236                 /*
237                  * We need the inner address.  However, we only get here for
238                  * transport mode so the outer address is identical.
239                  */
240                 daddr = &x->id.daddr;
241                 family = x->outer_mode->afinfo->family;
242
243                 err = xfrm_parse_spi(skb, nexthdr, &spi, &seq);
244                 if (err < 0) {
245                         XFRM_INC_STATS(LINUX_MIB_XFRMINHDRERROR);
246                         goto drop;
247                 }
248         } while (!err);
249
250         nf_reset(skb);
251
252         if (decaps) {
253                 dst_release(skb->dst);
254                 skb->dst = NULL;
255                 netif_rx(skb);
256                 return 0;
257         } else {
258                 return x->inner_mode->afinfo->transport_finish(skb, async);
259         }
260
261 drop_unlock:
262         spin_unlock(&x->lock);
263 drop:
264         kfree_skb(skb);
265         return 0;
266 }
267 EXPORT_SYMBOL(xfrm_input);
268
269 int xfrm_input_resume(struct sk_buff *skb, int nexthdr)
270 {
271         return xfrm_input(skb, nexthdr, 0, -1);
272 }
273 EXPORT_SYMBOL(xfrm_input_resume);
274
275 void __init xfrm_input_init(void)
276 {
277         secpath_cachep = kmem_cache_create("secpath_cache",
278                                            sizeof(struct sec_path),
279                                            0, SLAB_HWCACHE_ALIGN|SLAB_PANIC,
280                                            NULL);
281 }