Add the rt linux 4.1.3-rt3 as base
[kvmfornfv.git] / kernel / 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 #include <net/ip_tunnels.h>
17 #include <net/ip6_tunnel.h>
18
19 static struct kmem_cache *secpath_cachep __read_mostly;
20
21 static DEFINE_SPINLOCK(xfrm_input_afinfo_lock);
22 static struct xfrm_input_afinfo __rcu *xfrm_input_afinfo[NPROTO];
23
24 int xfrm_input_register_afinfo(struct xfrm_input_afinfo *afinfo)
25 {
26         int err = 0;
27
28         if (unlikely(afinfo == NULL))
29                 return -EINVAL;
30         if (unlikely(afinfo->family >= NPROTO))
31                 return -EAFNOSUPPORT;
32         spin_lock_bh(&xfrm_input_afinfo_lock);
33         if (unlikely(xfrm_input_afinfo[afinfo->family] != NULL))
34                 err = -ENOBUFS;
35         else
36                 rcu_assign_pointer(xfrm_input_afinfo[afinfo->family], afinfo);
37         spin_unlock_bh(&xfrm_input_afinfo_lock);
38         return err;
39 }
40 EXPORT_SYMBOL(xfrm_input_register_afinfo);
41
42 int xfrm_input_unregister_afinfo(struct xfrm_input_afinfo *afinfo)
43 {
44         int err = 0;
45
46         if (unlikely(afinfo == NULL))
47                 return -EINVAL;
48         if (unlikely(afinfo->family >= NPROTO))
49                 return -EAFNOSUPPORT;
50         spin_lock_bh(&xfrm_input_afinfo_lock);
51         if (likely(xfrm_input_afinfo[afinfo->family] != NULL)) {
52                 if (unlikely(xfrm_input_afinfo[afinfo->family] != afinfo))
53                         err = -EINVAL;
54                 else
55                         RCU_INIT_POINTER(xfrm_input_afinfo[afinfo->family], NULL);
56         }
57         spin_unlock_bh(&xfrm_input_afinfo_lock);
58         synchronize_rcu();
59         return err;
60 }
61 EXPORT_SYMBOL(xfrm_input_unregister_afinfo);
62
63 static struct xfrm_input_afinfo *xfrm_input_get_afinfo(unsigned int family)
64 {
65         struct xfrm_input_afinfo *afinfo;
66
67         if (unlikely(family >= NPROTO))
68                 return NULL;
69         rcu_read_lock();
70         afinfo = rcu_dereference(xfrm_input_afinfo[family]);
71         if (unlikely(!afinfo))
72                 rcu_read_unlock();
73         return afinfo;
74 }
75
76 static void xfrm_input_put_afinfo(struct xfrm_input_afinfo *afinfo)
77 {
78         rcu_read_unlock();
79 }
80
81 static int xfrm_rcv_cb(struct sk_buff *skb, unsigned int family, u8 protocol,
82                        int err)
83 {
84         int ret;
85         struct xfrm_input_afinfo *afinfo = xfrm_input_get_afinfo(family);
86
87         if (!afinfo)
88                 return -EAFNOSUPPORT;
89
90         ret = afinfo->callback(skb, protocol, err);
91         xfrm_input_put_afinfo(afinfo);
92
93         return ret;
94 }
95
96 void __secpath_destroy(struct sec_path *sp)
97 {
98         int i;
99         for (i = 0; i < sp->len; i++)
100                 xfrm_state_put(sp->xvec[i]);
101         kmem_cache_free(secpath_cachep, sp);
102 }
103 EXPORT_SYMBOL(__secpath_destroy);
104
105 struct sec_path *secpath_dup(struct sec_path *src)
106 {
107         struct sec_path *sp;
108
109         sp = kmem_cache_alloc(secpath_cachep, GFP_ATOMIC);
110         if (!sp)
111                 return NULL;
112
113         sp->len = 0;
114         if (src) {
115                 int i;
116
117                 memcpy(sp, src, sizeof(*sp));
118                 for (i = 0; i < sp->len; i++)
119                         xfrm_state_hold(sp->xvec[i]);
120         }
121         atomic_set(&sp->refcnt, 1);
122         return sp;
123 }
124 EXPORT_SYMBOL(secpath_dup);
125
126 /* Fetch spi and seq from ipsec header */
127
128 int xfrm_parse_spi(struct sk_buff *skb, u8 nexthdr, __be32 *spi, __be32 *seq)
129 {
130         int offset, offset_seq;
131         int hlen;
132
133         switch (nexthdr) {
134         case IPPROTO_AH:
135                 hlen = sizeof(struct ip_auth_hdr);
136                 offset = offsetof(struct ip_auth_hdr, spi);
137                 offset_seq = offsetof(struct ip_auth_hdr, seq_no);
138                 break;
139         case IPPROTO_ESP:
140                 hlen = sizeof(struct ip_esp_hdr);
141                 offset = offsetof(struct ip_esp_hdr, spi);
142                 offset_seq = offsetof(struct ip_esp_hdr, seq_no);
143                 break;
144         case IPPROTO_COMP:
145                 if (!pskb_may_pull(skb, sizeof(struct ip_comp_hdr)))
146                         return -EINVAL;
147                 *spi = htonl(ntohs(*(__be16 *)(skb_transport_header(skb) + 2)));
148                 *seq = 0;
149                 return 0;
150         default:
151                 return 1;
152         }
153
154         if (!pskb_may_pull(skb, hlen))
155                 return -EINVAL;
156
157         *spi = *(__be32 *)(skb_transport_header(skb) + offset);
158         *seq = *(__be32 *)(skb_transport_header(skb) + offset_seq);
159         return 0;
160 }
161
162 int xfrm_prepare_input(struct xfrm_state *x, struct sk_buff *skb)
163 {
164         struct xfrm_mode *inner_mode = x->inner_mode;
165         int err;
166
167         err = x->outer_mode->afinfo->extract_input(x, skb);
168         if (err)
169                 return err;
170
171         if (x->sel.family == AF_UNSPEC) {
172                 inner_mode = xfrm_ip2inner_mode(x, XFRM_MODE_SKB_CB(skb)->protocol);
173                 if (inner_mode == NULL)
174                         return -EAFNOSUPPORT;
175         }
176
177         skb->protocol = inner_mode->afinfo->eth_proto;
178         return inner_mode->input2(x, skb);
179 }
180 EXPORT_SYMBOL(xfrm_prepare_input);
181
182 int xfrm_input(struct sk_buff *skb, int nexthdr, __be32 spi, int encap_type)
183 {
184         struct net *net = dev_net(skb->dev);
185         int err;
186         __be32 seq;
187         __be32 seq_hi;
188         struct xfrm_state *x = NULL;
189         xfrm_address_t *daddr;
190         struct xfrm_mode *inner_mode;
191         u32 mark = skb->mark;
192         unsigned int family;
193         int decaps = 0;
194         int async = 0;
195
196         /* A negative encap_type indicates async resumption. */
197         if (encap_type < 0) {
198                 async = 1;
199                 x = xfrm_input_state(skb);
200                 seq = XFRM_SKB_CB(skb)->seq.input.low;
201                 family = x->outer_mode->afinfo->family;
202                 goto resume;
203         }
204
205         daddr = (xfrm_address_t *)(skb_network_header(skb) +
206                                    XFRM_SPI_SKB_CB(skb)->daddroff);
207         family = XFRM_SPI_SKB_CB(skb)->family;
208
209         /* if tunnel is present override skb->mark value with tunnel i_key */
210         if (XFRM_TUNNEL_SKB_CB(skb)->tunnel.ip4) {
211                 switch (family) {
212                 case AF_INET:
213                         mark = be32_to_cpu(XFRM_TUNNEL_SKB_CB(skb)->tunnel.ip4->parms.i_key);
214                         break;
215                 case AF_INET6:
216                         mark = be32_to_cpu(XFRM_TUNNEL_SKB_CB(skb)->tunnel.ip6->parms.i_key);
217                         break;
218                 }
219         }
220
221         /* Allocate new secpath or COW existing one. */
222         if (!skb->sp || atomic_read(&skb->sp->refcnt) != 1) {
223                 struct sec_path *sp;
224
225                 sp = secpath_dup(skb->sp);
226                 if (!sp) {
227                         XFRM_INC_STATS(net, LINUX_MIB_XFRMINERROR);
228                         goto drop;
229                 }
230                 if (skb->sp)
231                         secpath_put(skb->sp);
232                 skb->sp = sp;
233         }
234
235         seq = 0;
236         if (!spi && (err = xfrm_parse_spi(skb, nexthdr, &spi, &seq)) != 0) {
237                 XFRM_INC_STATS(net, LINUX_MIB_XFRMINHDRERROR);
238                 goto drop;
239         }
240
241         do {
242                 if (skb->sp->len == XFRM_MAX_DEPTH) {
243                         XFRM_INC_STATS(net, LINUX_MIB_XFRMINBUFFERERROR);
244                         goto drop;
245                 }
246
247                 x = xfrm_state_lookup(net, mark, daddr, spi, nexthdr, family);
248                 if (x == NULL) {
249                         XFRM_INC_STATS(net, LINUX_MIB_XFRMINNOSTATES);
250                         xfrm_audit_state_notfound(skb, family, spi, seq);
251                         goto drop;
252                 }
253
254                 skb->sp->xvec[skb->sp->len++] = x;
255
256                 spin_lock(&x->lock);
257                 if (unlikely(x->km.state == XFRM_STATE_ACQ)) {
258                         XFRM_INC_STATS(net, LINUX_MIB_XFRMACQUIREERROR);
259                         goto drop_unlock;
260                 }
261
262                 if (unlikely(x->km.state != XFRM_STATE_VALID)) {
263                         XFRM_INC_STATS(net, LINUX_MIB_XFRMINSTATEINVALID);
264                         goto drop_unlock;
265                 }
266
267                 if ((x->encap ? x->encap->encap_type : 0) != encap_type) {
268                         XFRM_INC_STATS(net, LINUX_MIB_XFRMINSTATEMISMATCH);
269                         goto drop_unlock;
270                 }
271
272                 if (x->repl->check(x, skb, seq)) {
273                         XFRM_INC_STATS(net, LINUX_MIB_XFRMINSTATESEQERROR);
274                         goto drop_unlock;
275                 }
276
277                 if (xfrm_state_check_expire(x)) {
278                         XFRM_INC_STATS(net, LINUX_MIB_XFRMINSTATEEXPIRED);
279                         goto drop_unlock;
280                 }
281
282                 spin_unlock(&x->lock);
283
284                 if (xfrm_tunnel_check(skb, x, family)) {
285                         XFRM_INC_STATS(net, LINUX_MIB_XFRMINSTATEMODEERROR);
286                         goto drop;
287                 }
288
289                 seq_hi = htonl(xfrm_replay_seqhi(x, seq));
290
291                 XFRM_SKB_CB(skb)->seq.input.low = seq;
292                 XFRM_SKB_CB(skb)->seq.input.hi = seq_hi;
293
294                 skb_dst_force(skb);
295
296                 nexthdr = x->type->input(x, skb);
297
298                 if (nexthdr == -EINPROGRESS)
299                         return 0;
300 resume:
301                 spin_lock(&x->lock);
302                 if (nexthdr <= 0) {
303                         if (nexthdr == -EBADMSG) {
304                                 xfrm_audit_state_icvfail(x, skb,
305                                                          x->type->proto);
306                                 x->stats.integrity_failed++;
307                         }
308                         XFRM_INC_STATS(net, LINUX_MIB_XFRMINSTATEPROTOERROR);
309                         goto drop_unlock;
310                 }
311
312                 /* only the first xfrm gets the encap type */
313                 encap_type = 0;
314
315                 if (async && x->repl->recheck(x, skb, seq)) {
316                         XFRM_INC_STATS(net, LINUX_MIB_XFRMINSTATESEQERROR);
317                         goto drop_unlock;
318                 }
319
320                 x->repl->advance(x, seq);
321
322                 x->curlft.bytes += skb->len;
323                 x->curlft.packets++;
324
325                 spin_unlock(&x->lock);
326
327                 XFRM_MODE_SKB_CB(skb)->protocol = nexthdr;
328
329                 inner_mode = x->inner_mode;
330
331                 if (x->sel.family == AF_UNSPEC) {
332                         inner_mode = xfrm_ip2inner_mode(x, XFRM_MODE_SKB_CB(skb)->protocol);
333                         if (inner_mode == NULL)
334                                 goto drop;
335                 }
336
337                 if (inner_mode->input(x, skb)) {
338                         XFRM_INC_STATS(net, LINUX_MIB_XFRMINSTATEMODEERROR);
339                         goto drop;
340                 }
341
342                 if (x->outer_mode->flags & XFRM_MODE_FLAG_TUNNEL) {
343                         decaps = 1;
344                         break;
345                 }
346
347                 /*
348                  * We need the inner address.  However, we only get here for
349                  * transport mode so the outer address is identical.
350                  */
351                 daddr = &x->id.daddr;
352                 family = x->outer_mode->afinfo->family;
353
354                 err = xfrm_parse_spi(skb, nexthdr, &spi, &seq);
355                 if (err < 0) {
356                         XFRM_INC_STATS(net, LINUX_MIB_XFRMINHDRERROR);
357                         goto drop;
358                 }
359         } while (!err);
360
361         err = xfrm_rcv_cb(skb, family, x->type->proto, 0);
362         if (err)
363                 goto drop;
364
365         nf_reset(skb);
366
367         if (decaps) {
368                 skb_dst_drop(skb);
369                 netif_rx(skb);
370                 return 0;
371         } else {
372                 return x->inner_mode->afinfo->transport_finish(skb, async);
373         }
374
375 drop_unlock:
376         spin_unlock(&x->lock);
377 drop:
378         xfrm_rcv_cb(skb, family, x && x->type ? x->type->proto : nexthdr, -1);
379         kfree_skb(skb);
380         return 0;
381 }
382 EXPORT_SYMBOL(xfrm_input);
383
384 int xfrm_input_resume(struct sk_buff *skb, int nexthdr)
385 {
386         return xfrm_input(skb, nexthdr, 0, -1);
387 }
388 EXPORT_SYMBOL(xfrm_input_resume);
389
390 void __init xfrm_input_init(void)
391 {
392         secpath_cachep = kmem_cache_create("secpath_cache",
393                                            sizeof(struct sec_path),
394                                            0, SLAB_HWCACHE_ALIGN|SLAB_PANIC,
395                                            NULL);
396 }