net/netfilter/nf_conntrack_core: Fix net_conntrack_lock()
[linux/fpc-iii.git] / net / xfrm / xfrm_input.c
blob923205e279f72dbab95cee827987feccc690c678
1 /*
2 * xfrm_input.c
4 * Changes:
5 * YOSHIFUJI Hideaki @USAGI
6 * Split up af-specific portion
8 */
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 #include <net/ip_tunnels.h>
17 #include <net/ip6_tunnel.h>
19 static struct kmem_cache *secpath_cachep __read_mostly;
21 static DEFINE_SPINLOCK(xfrm_input_afinfo_lock);
22 static struct xfrm_input_afinfo const __rcu *xfrm_input_afinfo[AF_INET6 + 1];
24 static struct gro_cells gro_cells;
25 static struct net_device xfrm_napi_dev;
27 int xfrm_input_register_afinfo(const struct xfrm_input_afinfo *afinfo)
29 int err = 0;
31 if (WARN_ON(afinfo->family >= ARRAY_SIZE(xfrm_input_afinfo)))
32 return -EAFNOSUPPORT;
34 spin_lock_bh(&xfrm_input_afinfo_lock);
35 if (unlikely(xfrm_input_afinfo[afinfo->family] != NULL))
36 err = -EEXIST;
37 else
38 rcu_assign_pointer(xfrm_input_afinfo[afinfo->family], afinfo);
39 spin_unlock_bh(&xfrm_input_afinfo_lock);
40 return err;
42 EXPORT_SYMBOL(xfrm_input_register_afinfo);
44 int xfrm_input_unregister_afinfo(const struct xfrm_input_afinfo *afinfo)
46 int err = 0;
48 spin_lock_bh(&xfrm_input_afinfo_lock);
49 if (likely(xfrm_input_afinfo[afinfo->family] != NULL)) {
50 if (unlikely(xfrm_input_afinfo[afinfo->family] != afinfo))
51 err = -EINVAL;
52 else
53 RCU_INIT_POINTER(xfrm_input_afinfo[afinfo->family], NULL);
55 spin_unlock_bh(&xfrm_input_afinfo_lock);
56 synchronize_rcu();
57 return err;
59 EXPORT_SYMBOL(xfrm_input_unregister_afinfo);
61 static const struct xfrm_input_afinfo *xfrm_input_get_afinfo(unsigned int family)
63 const struct xfrm_input_afinfo *afinfo;
65 if (WARN_ON_ONCE(family >= ARRAY_SIZE(xfrm_input_afinfo)))
66 return NULL;
68 rcu_read_lock();
69 afinfo = rcu_dereference(xfrm_input_afinfo[family]);
70 if (unlikely(!afinfo))
71 rcu_read_unlock();
72 return afinfo;
75 static int xfrm_rcv_cb(struct sk_buff *skb, unsigned int family, u8 protocol,
76 int err)
78 int ret;
79 const struct xfrm_input_afinfo *afinfo = xfrm_input_get_afinfo(family);
81 if (!afinfo)
82 return -EAFNOSUPPORT;
84 ret = afinfo->callback(skb, protocol, err);
85 rcu_read_unlock();
87 return ret;
90 void __secpath_destroy(struct sec_path *sp)
92 int i;
93 for (i = 0; i < sp->len; i++)
94 xfrm_state_put(sp->xvec[i]);
95 kmem_cache_free(secpath_cachep, sp);
97 EXPORT_SYMBOL(__secpath_destroy);
99 struct sec_path *secpath_dup(struct sec_path *src)
101 struct sec_path *sp;
103 sp = kmem_cache_alloc(secpath_cachep, GFP_ATOMIC);
104 if (!sp)
105 return NULL;
107 sp->len = 0;
108 sp->olen = 0;
110 memset(sp->ovec, 0, sizeof(sp->ovec[XFRM_MAX_OFFLOAD_DEPTH]));
112 if (src) {
113 int i;
115 memcpy(sp, src, sizeof(*sp));
116 for (i = 0; i < sp->len; i++)
117 xfrm_state_hold(sp->xvec[i]);
119 refcount_set(&sp->refcnt, 1);
120 return sp;
122 EXPORT_SYMBOL(secpath_dup);
124 int secpath_set(struct sk_buff *skb)
126 struct sec_path *sp;
128 /* Allocate new secpath or COW existing one. */
129 if (!skb->sp || refcount_read(&skb->sp->refcnt) != 1) {
130 sp = secpath_dup(skb->sp);
131 if (!sp)
132 return -ENOMEM;
134 if (skb->sp)
135 secpath_put(skb->sp);
136 skb->sp = sp;
138 return 0;
140 EXPORT_SYMBOL(secpath_set);
142 /* Fetch spi and seq from ipsec header */
144 int xfrm_parse_spi(struct sk_buff *skb, u8 nexthdr, __be32 *spi, __be32 *seq)
146 int offset, offset_seq;
147 int hlen;
149 switch (nexthdr) {
150 case IPPROTO_AH:
151 hlen = sizeof(struct ip_auth_hdr);
152 offset = offsetof(struct ip_auth_hdr, spi);
153 offset_seq = offsetof(struct ip_auth_hdr, seq_no);
154 break;
155 case IPPROTO_ESP:
156 hlen = sizeof(struct ip_esp_hdr);
157 offset = offsetof(struct ip_esp_hdr, spi);
158 offset_seq = offsetof(struct ip_esp_hdr, seq_no);
159 break;
160 case IPPROTO_COMP:
161 if (!pskb_may_pull(skb, sizeof(struct ip_comp_hdr)))
162 return -EINVAL;
163 *spi = htonl(ntohs(*(__be16 *)(skb_transport_header(skb) + 2)));
164 *seq = 0;
165 return 0;
166 default:
167 return 1;
170 if (!pskb_may_pull(skb, hlen))
171 return -EINVAL;
173 *spi = *(__be32 *)(skb_transport_header(skb) + offset);
174 *seq = *(__be32 *)(skb_transport_header(skb) + offset_seq);
175 return 0;
177 EXPORT_SYMBOL(xfrm_parse_spi);
179 int xfrm_prepare_input(struct xfrm_state *x, struct sk_buff *skb)
181 struct xfrm_mode *inner_mode = x->inner_mode;
182 int err;
184 err = x->outer_mode->afinfo->extract_input(x, skb);
185 if (err)
186 return err;
188 if (x->sel.family == AF_UNSPEC) {
189 inner_mode = xfrm_ip2inner_mode(x, XFRM_MODE_SKB_CB(skb)->protocol);
190 if (inner_mode == NULL)
191 return -EAFNOSUPPORT;
194 skb->protocol = inner_mode->afinfo->eth_proto;
195 return inner_mode->input2(x, skb);
197 EXPORT_SYMBOL(xfrm_prepare_input);
199 int xfrm_input(struct sk_buff *skb, int nexthdr, __be32 spi, int encap_type)
201 struct net *net = dev_net(skb->dev);
202 int err;
203 __be32 seq;
204 __be32 seq_hi;
205 struct xfrm_state *x = NULL;
206 xfrm_address_t *daddr;
207 struct xfrm_mode *inner_mode;
208 u32 mark = skb->mark;
209 unsigned int family;
210 int decaps = 0;
211 int async = 0;
212 bool xfrm_gro = false;
213 bool crypto_done = false;
214 struct xfrm_offload *xo = xfrm_offload(skb);
216 if (encap_type < 0) {
217 x = xfrm_input_state(skb);
218 family = x->outer_mode->afinfo->family;
220 /* An encap_type of -1 indicates async resumption. */
221 if (encap_type == -1) {
222 async = 1;
223 seq = XFRM_SKB_CB(skb)->seq.input.low;
224 goto resume;
227 /* encap_type < -1 indicates a GRO call. */
228 encap_type = 0;
229 seq = XFRM_SPI_SKB_CB(skb)->seq;
231 if (xo && (xo->flags & CRYPTO_DONE)) {
232 crypto_done = true;
233 x = xfrm_input_state(skb);
234 family = XFRM_SPI_SKB_CB(skb)->family;
236 if (!(xo->status & CRYPTO_SUCCESS)) {
237 if (xo->status &
238 (CRYPTO_TRANSPORT_AH_AUTH_FAILED |
239 CRYPTO_TRANSPORT_ESP_AUTH_FAILED |
240 CRYPTO_TUNNEL_AH_AUTH_FAILED |
241 CRYPTO_TUNNEL_ESP_AUTH_FAILED)) {
243 xfrm_audit_state_icvfail(x, skb,
244 x->type->proto);
245 x->stats.integrity_failed++;
246 XFRM_INC_STATS(net, LINUX_MIB_XFRMINSTATEPROTOERROR);
247 goto drop;
250 XFRM_INC_STATS(net, LINUX_MIB_XFRMINBUFFERERROR);
251 goto drop;
254 if ((err = xfrm_parse_spi(skb, nexthdr, &spi, &seq)) != 0) {
255 XFRM_INC_STATS(net, LINUX_MIB_XFRMINHDRERROR);
256 goto drop;
260 goto lock;
263 daddr = (xfrm_address_t *)(skb_network_header(skb) +
264 XFRM_SPI_SKB_CB(skb)->daddroff);
265 family = XFRM_SPI_SKB_CB(skb)->family;
267 /* if tunnel is present override skb->mark value with tunnel i_key */
268 switch (family) {
269 case AF_INET:
270 if (XFRM_TUNNEL_SKB_CB(skb)->tunnel.ip4)
271 mark = be32_to_cpu(XFRM_TUNNEL_SKB_CB(skb)->tunnel.ip4->parms.i_key);
272 break;
273 case AF_INET6:
274 if (XFRM_TUNNEL_SKB_CB(skb)->tunnel.ip6)
275 mark = be32_to_cpu(XFRM_TUNNEL_SKB_CB(skb)->tunnel.ip6->parms.i_key);
276 break;
279 err = secpath_set(skb);
280 if (err) {
281 XFRM_INC_STATS(net, LINUX_MIB_XFRMINERROR);
282 goto drop;
285 seq = 0;
286 if (!spi && (err = xfrm_parse_spi(skb, nexthdr, &spi, &seq)) != 0) {
287 XFRM_INC_STATS(net, LINUX_MIB_XFRMINHDRERROR);
288 goto drop;
291 do {
292 if (skb->sp->len == XFRM_MAX_DEPTH) {
293 XFRM_INC_STATS(net, LINUX_MIB_XFRMINBUFFERERROR);
294 goto drop;
297 x = xfrm_state_lookup(net, mark, daddr, spi, nexthdr, family);
298 if (x == NULL) {
299 XFRM_INC_STATS(net, LINUX_MIB_XFRMINNOSTATES);
300 xfrm_audit_state_notfound(skb, family, spi, seq);
301 goto drop;
304 skb->sp->xvec[skb->sp->len++] = x;
306 lock:
307 spin_lock(&x->lock);
309 if (unlikely(x->km.state != XFRM_STATE_VALID)) {
310 if (x->km.state == XFRM_STATE_ACQ)
311 XFRM_INC_STATS(net, LINUX_MIB_XFRMACQUIREERROR);
312 else
313 XFRM_INC_STATS(net,
314 LINUX_MIB_XFRMINSTATEINVALID);
315 goto drop_unlock;
318 if ((x->encap ? x->encap->encap_type : 0) != encap_type) {
319 XFRM_INC_STATS(net, LINUX_MIB_XFRMINSTATEMISMATCH);
320 goto drop_unlock;
323 if (x->repl->check(x, skb, seq)) {
324 XFRM_INC_STATS(net, LINUX_MIB_XFRMINSTATESEQERROR);
325 goto drop_unlock;
328 if (xfrm_state_check_expire(x)) {
329 XFRM_INC_STATS(net, LINUX_MIB_XFRMINSTATEEXPIRED);
330 goto drop_unlock;
333 spin_unlock(&x->lock);
335 if (xfrm_tunnel_check(skb, x, family)) {
336 XFRM_INC_STATS(net, LINUX_MIB_XFRMINSTATEMODEERROR);
337 goto drop;
340 seq_hi = htonl(xfrm_replay_seqhi(x, seq));
342 XFRM_SKB_CB(skb)->seq.input.low = seq;
343 XFRM_SKB_CB(skb)->seq.input.hi = seq_hi;
345 skb_dst_force(skb);
346 dev_hold(skb->dev);
348 if (crypto_done)
349 nexthdr = x->type_offload->input_tail(x, skb);
350 else
351 nexthdr = x->type->input(x, skb);
353 if (nexthdr == -EINPROGRESS)
354 return 0;
355 resume:
356 dev_put(skb->dev);
358 spin_lock(&x->lock);
359 if (nexthdr <= 0) {
360 if (nexthdr == -EBADMSG) {
361 xfrm_audit_state_icvfail(x, skb,
362 x->type->proto);
363 x->stats.integrity_failed++;
365 XFRM_INC_STATS(net, LINUX_MIB_XFRMINSTATEPROTOERROR);
366 goto drop_unlock;
369 /* only the first xfrm gets the encap type */
370 encap_type = 0;
372 if (async && x->repl->recheck(x, skb, seq)) {
373 XFRM_INC_STATS(net, LINUX_MIB_XFRMINSTATESEQERROR);
374 goto drop_unlock;
377 x->repl->advance(x, seq);
379 x->curlft.bytes += skb->len;
380 x->curlft.packets++;
382 spin_unlock(&x->lock);
384 XFRM_MODE_SKB_CB(skb)->protocol = nexthdr;
386 inner_mode = x->inner_mode;
388 if (x->sel.family == AF_UNSPEC) {
389 inner_mode = xfrm_ip2inner_mode(x, XFRM_MODE_SKB_CB(skb)->protocol);
390 if (inner_mode == NULL) {
391 XFRM_INC_STATS(net, LINUX_MIB_XFRMINSTATEMODEERROR);
392 goto drop;
396 if (inner_mode->input(x, skb)) {
397 XFRM_INC_STATS(net, LINUX_MIB_XFRMINSTATEMODEERROR);
398 goto drop;
401 if (x->outer_mode->flags & XFRM_MODE_FLAG_TUNNEL) {
402 decaps = 1;
403 break;
407 * We need the inner address. However, we only get here for
408 * transport mode so the outer address is identical.
410 daddr = &x->id.daddr;
411 family = x->outer_mode->afinfo->family;
413 err = xfrm_parse_spi(skb, nexthdr, &spi, &seq);
414 if (err < 0) {
415 XFRM_INC_STATS(net, LINUX_MIB_XFRMINHDRERROR);
416 goto drop;
418 } while (!err);
420 err = xfrm_rcv_cb(skb, family, x->type->proto, 0);
421 if (err)
422 goto drop;
424 nf_reset(skb);
426 if (decaps) {
427 skb_dst_drop(skb);
428 gro_cells_receive(&gro_cells, skb);
429 return 0;
430 } else {
431 xo = xfrm_offload(skb);
432 if (xo)
433 xfrm_gro = xo->flags & XFRM_GRO;
435 err = x->inner_mode->afinfo->transport_finish(skb, xfrm_gro || async);
436 if (xfrm_gro) {
437 skb_dst_drop(skb);
438 gro_cells_receive(&gro_cells, skb);
439 return err;
442 return err;
445 drop_unlock:
446 spin_unlock(&x->lock);
447 drop:
448 xfrm_rcv_cb(skb, family, x && x->type ? x->type->proto : nexthdr, -1);
449 kfree_skb(skb);
450 return 0;
452 EXPORT_SYMBOL(xfrm_input);
454 int xfrm_input_resume(struct sk_buff *skb, int nexthdr)
456 return xfrm_input(skb, nexthdr, 0, -1);
458 EXPORT_SYMBOL(xfrm_input_resume);
460 void __init xfrm_input_init(void)
462 int err;
464 init_dummy_netdev(&xfrm_napi_dev);
465 err = gro_cells_init(&gro_cells, &xfrm_napi_dev);
466 if (err)
467 gro_cells.cells = NULL;
469 secpath_cachep = kmem_cache_create("secpath_cache",
470 sizeof(struct sec_path),
471 0, SLAB_HWCACHE_ALIGN|SLAB_PANIC,
472 NULL);