[NET]: is it Andy or Andi ??
[linux-2.6] / net / ipv6 / xfrm6_policy.c
1 /*
2  * xfrm6_policy.c: based on xfrm4_policy.c
3  *
4  * Authors:
5  *      Mitsuru KANDA @USAGI
6  *      Kazunori MIYAZAWA @USAGI
7  *      Kunihiro Ishiguro <kunihiro@ipinfusion.com>
8  *              IPv6 support
9  *      YOSHIFUJI Hideaki
10  *              Split up af-specific portion
11  * 
12  */
13
14 #include <linux/compiler.h>
15 #include <linux/netdevice.h>
16 #include <net/addrconf.h>
17 #include <net/xfrm.h>
18 #include <net/ip.h>
19 #include <net/ipv6.h>
20 #include <net/ip6_route.h>
21 #ifdef CONFIG_IPV6_MIP6
22 #include <net/mip6.h>
23 #endif
24
25 static struct dst_ops xfrm6_dst_ops;
26 static struct xfrm_policy_afinfo xfrm6_policy_afinfo;
27
28 static int xfrm6_dst_lookup(struct xfrm_dst **dst, struct flowi *fl)
29 {
30         int err = 0;
31         *dst = (struct xfrm_dst*)ip6_route_output(NULL, fl);
32         if (!*dst)
33                 err = -ENETUNREACH;
34         return err;
35 }
36
37 static int xfrm6_get_saddr(xfrm_address_t *saddr, xfrm_address_t *daddr)
38 {
39         struct rt6_info *rt;
40         struct flowi fl_tunnel = {
41                 .nl_u = {
42                         .ip6_u = {
43                                 .daddr = *(struct in6_addr *)&daddr->a6,
44                         },
45                 },
46         };
47
48         if (!xfrm6_dst_lookup((struct xfrm_dst **)&rt, &fl_tunnel)) {
49                 ipv6_get_saddr(&rt->u.dst, (struct in6_addr *)&daddr->a6,
50                                (struct in6_addr *)&saddr->a6);
51                 dst_release(&rt->u.dst);
52                 return 0;
53         }
54         return -EHOSTUNREACH;
55 }
56
57 static struct dst_entry *
58 __xfrm6_find_bundle(struct flowi *fl, struct xfrm_policy *policy)
59 {
60         struct dst_entry *dst;
61
62         /* Still not clear if we should set fl->fl6_{src,dst}... */
63         read_lock_bh(&policy->lock);
64         for (dst = policy->bundles; dst; dst = dst->next) {
65                 struct xfrm_dst *xdst = (struct xfrm_dst*)dst;
66                 struct in6_addr fl_dst_prefix, fl_src_prefix;
67
68                 ipv6_addr_prefix(&fl_dst_prefix,
69                                  &fl->fl6_dst,
70                                  xdst->u.rt6.rt6i_dst.plen);
71                 ipv6_addr_prefix(&fl_src_prefix,
72                                  &fl->fl6_src,
73                                  xdst->u.rt6.rt6i_src.plen);
74                 if (ipv6_addr_equal(&xdst->u.rt6.rt6i_dst.addr, &fl_dst_prefix) &&
75                     ipv6_addr_equal(&xdst->u.rt6.rt6i_src.addr, &fl_src_prefix) &&
76                     xfrm_bundle_ok(xdst, fl, AF_INET6,
77                                    (xdst->u.rt6.rt6i_dst.plen != 128 ||
78                                     xdst->u.rt6.rt6i_src.plen != 128))) {
79                         dst_clone(dst);
80                         break;
81                 }
82         }
83         read_unlock_bh(&policy->lock);
84         return dst;
85 }
86
87 static inline struct in6_addr*
88 __xfrm6_bundle_addr_remote(struct xfrm_state *x, struct in6_addr *addr)
89 {
90         return (x->type->remote_addr) ?
91                 (struct in6_addr*)x->type->remote_addr(x, (xfrm_address_t *)addr) :
92                 (struct in6_addr*)&x->id.daddr;
93 }
94
95 static inline struct in6_addr*
96 __xfrm6_bundle_addr_local(struct xfrm_state *x, struct in6_addr *addr)
97 {
98         return (x->type->local_addr) ?
99                 (struct in6_addr*)x->type->local_addr(x, (xfrm_address_t *)addr) :
100                 (struct in6_addr*)&x->props.saddr;
101 }
102
103 static inline void
104 __xfrm6_bundle_len_inc(int *len, int *nflen, struct xfrm_state *x)
105 {
106         if (x->type->flags & XFRM_TYPE_NON_FRAGMENT)
107                 *nflen += x->props.header_len;
108         else
109                 *len += x->props.header_len;
110 }
111
112 static inline void
113 __xfrm6_bundle_len_dec(int *len, int *nflen, struct xfrm_state *x)
114 {
115         if (x->type->flags & XFRM_TYPE_NON_FRAGMENT)
116                 *nflen -= x->props.header_len;
117         else
118                 *len -= x->props.header_len;
119 }
120
121 /* Allocate chain of dst_entry's, attach known xfrm's, calculate
122  * all the metrics... Shortly, bundle a bundle.
123  */
124
125 static int
126 __xfrm6_bundle_create(struct xfrm_policy *policy, struct xfrm_state **xfrm, int nx,
127                       struct flowi *fl, struct dst_entry **dst_p)
128 {
129         struct dst_entry *dst, *dst_prev;
130         struct rt6_info *rt0 = (struct rt6_info*)(*dst_p);
131         struct rt6_info *rt  = rt0;
132         struct in6_addr *remote = &fl->fl6_dst;
133         struct in6_addr *local  = &fl->fl6_src;
134         struct flowi fl_tunnel = {
135                 .nl_u = {
136                         .ip6_u = {
137                                 .saddr = *local,
138                                 .daddr = *remote
139                         }
140                 }
141         };
142         int i;
143         int err = 0;
144         int header_len = 0;
145         int nfheader_len = 0;
146         int trailer_len = 0;
147
148         dst = dst_prev = NULL;
149         dst_hold(&rt->u.dst);
150
151         for (i = 0; i < nx; i++) {
152                 struct dst_entry *dst1 = dst_alloc(&xfrm6_dst_ops);
153                 struct xfrm_dst *xdst;
154                 int tunnel = 0;
155
156                 if (unlikely(dst1 == NULL)) {
157                         err = -ENOBUFS;
158                         dst_release(&rt->u.dst);
159                         goto error;
160                 }
161
162                 if (!dst)
163                         dst = dst1;
164                 else {
165                         dst_prev->child = dst1;
166                         dst1->flags |= DST_NOHASH;
167                         dst_clone(dst1);
168                 }
169
170                 xdst = (struct xfrm_dst *)dst1;
171                 xdst->route = &rt->u.dst;
172                 xdst->genid = xfrm[i]->genid;
173                 if (rt->rt6i_node)
174                         xdst->route_cookie = rt->rt6i_node->fn_sernum;
175
176                 dst1->next = dst_prev;
177                 dst_prev = dst1;
178                 if (xfrm[i]->props.mode != XFRM_MODE_TRANSPORT) {
179                         remote = __xfrm6_bundle_addr_remote(xfrm[i], remote);
180                         local  = __xfrm6_bundle_addr_local(xfrm[i], local);
181                         tunnel = 1;
182                 }
183                 __xfrm6_bundle_len_inc(&header_len, &nfheader_len, xfrm[i]);
184                 trailer_len += xfrm[i]->props.trailer_len;
185
186                 if (tunnel) {
187                         ipv6_addr_copy(&fl_tunnel.fl6_dst, remote);
188                         ipv6_addr_copy(&fl_tunnel.fl6_src, local);
189                         err = xfrm_dst_lookup((struct xfrm_dst **) &rt,
190                                               &fl_tunnel, AF_INET6);
191                         if (err)
192                                 goto error;
193                 } else
194                         dst_hold(&rt->u.dst);
195         }
196
197         dst_prev->child = &rt->u.dst;
198         dst->path = &rt->u.dst;
199         if (rt->rt6i_node)
200                 ((struct xfrm_dst *)dst)->path_cookie = rt->rt6i_node->fn_sernum;
201
202         *dst_p = dst;
203         dst = dst_prev;
204
205         dst_prev = *dst_p;
206         i = 0;
207         for (; dst_prev != &rt->u.dst; dst_prev = dst_prev->child) {
208                 struct xfrm_dst *x = (struct xfrm_dst*)dst_prev;
209
210                 dst_prev->xfrm = xfrm[i++];
211                 dst_prev->dev = rt->u.dst.dev;
212                 if (rt->u.dst.dev)
213                         dev_hold(rt->u.dst.dev);
214                 dst_prev->obsolete      = -1;
215                 dst_prev->flags        |= DST_HOST;
216                 dst_prev->lastuse       = jiffies;
217                 dst_prev->header_len    = header_len;
218                 dst_prev->nfheader_len  = nfheader_len;
219                 dst_prev->trailer_len   = trailer_len;
220                 memcpy(&dst_prev->metrics, &x->route->metrics, sizeof(dst_prev->metrics));
221
222                 /* Copy neighbour for reachability confirmation */
223                 dst_prev->neighbour     = neigh_clone(rt->u.dst.neighbour);
224                 dst_prev->input         = rt->u.dst.input;
225                 dst_prev->output        = xfrm6_output;
226                 /* Sheit... I remember I did this right. Apparently,
227                  * it was magically lost, so this code needs audit */
228                 x->u.rt6.rt6i_flags    = rt0->rt6i_flags&(RTCF_BROADCAST|RTCF_MULTICAST|RTCF_LOCAL);
229                 x->u.rt6.rt6i_metric   = rt0->rt6i_metric;
230                 x->u.rt6.rt6i_node     = rt0->rt6i_node;
231                 x->u.rt6.rt6i_gateway  = rt0->rt6i_gateway;
232                 memcpy(&x->u.rt6.rt6i_gateway, &rt0->rt6i_gateway, sizeof(x->u.rt6.rt6i_gateway)); 
233                 x->u.rt6.rt6i_dst      = rt0->rt6i_dst;
234                 x->u.rt6.rt6i_src      = rt0->rt6i_src; 
235                 x->u.rt6.rt6i_idev     = rt0->rt6i_idev;
236                 in6_dev_hold(rt0->rt6i_idev);
237                 __xfrm6_bundle_len_dec(&header_len, &nfheader_len, x->u.dst.xfrm);
238                 trailer_len -= x->u.dst.xfrm->props.trailer_len;
239         }
240
241         xfrm_init_pmtu(dst);
242         return 0;
243
244 error:
245         if (dst)
246                 dst_free(dst);
247         return err;
248 }
249
250 static inline void
251 _decode_session6(struct sk_buff *skb, struct flowi *fl)
252 {
253         u16 offset = skb->h.raw - skb->nh.raw;
254         struct ipv6hdr *hdr = skb->nh.ipv6h;
255         struct ipv6_opt_hdr *exthdr;
256         u8 nexthdr = skb->nh.raw[IP6CB(skb)->nhoff];
257
258         memset(fl, 0, sizeof(struct flowi));
259         ipv6_addr_copy(&fl->fl6_dst, &hdr->daddr);
260         ipv6_addr_copy(&fl->fl6_src, &hdr->saddr);
261
262         while (pskb_may_pull(skb, skb->nh.raw + offset + 1 - skb->data)) {
263                 exthdr = (struct ipv6_opt_hdr*)(skb->nh.raw + offset);
264
265                 switch (nexthdr) {
266                 case NEXTHDR_ROUTING:
267                 case NEXTHDR_HOP:
268                 case NEXTHDR_DEST:
269                         offset += ipv6_optlen(exthdr);
270                         nexthdr = exthdr->nexthdr;
271                         exthdr = (struct ipv6_opt_hdr*)(skb->nh.raw + offset);
272                         break;
273
274                 case IPPROTO_UDP:
275                 case IPPROTO_TCP:
276                 case IPPROTO_SCTP:
277                 case IPPROTO_DCCP:
278                         if (pskb_may_pull(skb, skb->nh.raw + offset + 4 - skb->data)) {
279                                 u16 *ports = (u16 *)exthdr;
280
281                                 fl->fl_ip_sport = ports[0];
282                                 fl->fl_ip_dport = ports[1];
283                         }
284                         fl->proto = nexthdr;
285                         return;
286
287                 case IPPROTO_ICMPV6:
288                         if (pskb_may_pull(skb, skb->nh.raw + offset + 2 - skb->data)) {
289                                 u8 *icmp = (u8 *)exthdr;
290
291                                 fl->fl_icmp_type = icmp[0];
292                                 fl->fl_icmp_code = icmp[1];
293                         }
294                         fl->proto = nexthdr;
295                         return;
296
297 #ifdef CONFIG_IPV6_MIP6
298                 case IPPROTO_MH:
299                         if (pskb_may_pull(skb, skb->nh.raw + offset + 3 - skb->data)) {
300                                 struct ip6_mh *mh;
301                                 mh = (struct ip6_mh *)exthdr;
302
303                                 fl->fl_mh_type = mh->ip6mh_type;
304                         }
305                         fl->proto = nexthdr;
306                         return;
307 #endif
308
309                 /* XXX Why are there these headers? */
310                 case IPPROTO_AH:
311                 case IPPROTO_ESP:
312                 case IPPROTO_COMP:
313                 default:
314                         fl->fl_ipsec_spi = 0;
315                         fl->proto = nexthdr;
316                         return;
317                 };
318         }
319 }
320
321 static inline int xfrm6_garbage_collect(void)
322 {
323         xfrm6_policy_afinfo.garbage_collect();
324         return (atomic_read(&xfrm6_dst_ops.entries) > xfrm6_dst_ops.gc_thresh*2);
325 }
326
327 static void xfrm6_update_pmtu(struct dst_entry *dst, u32 mtu)
328 {
329         struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
330         struct dst_entry *path = xdst->route;
331
332         path->ops->update_pmtu(path, mtu);
333 }
334
335 static void xfrm6_dst_destroy(struct dst_entry *dst)
336 {
337         struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
338
339         if (likely(xdst->u.rt6.rt6i_idev))
340                 in6_dev_put(xdst->u.rt6.rt6i_idev);
341         xfrm_dst_destroy(xdst);
342 }
343
344 static void xfrm6_dst_ifdown(struct dst_entry *dst, struct net_device *dev,
345                              int unregister)
346 {
347         struct xfrm_dst *xdst;
348
349         if (!unregister)
350                 return;
351
352         xdst = (struct xfrm_dst *)dst;
353         if (xdst->u.rt6.rt6i_idev->dev == dev) {
354                 struct inet6_dev *loopback_idev = in6_dev_get(&loopback_dev);
355                 BUG_ON(!loopback_idev);
356
357                 do {
358                         in6_dev_put(xdst->u.rt6.rt6i_idev);
359                         xdst->u.rt6.rt6i_idev = loopback_idev;
360                         in6_dev_hold(loopback_idev);
361                         xdst = (struct xfrm_dst *)xdst->u.dst.child;
362                 } while (xdst->u.dst.xfrm);
363
364                 __in6_dev_put(loopback_idev);
365         }
366
367         xfrm_dst_ifdown(dst, dev);
368 }
369
370 static struct dst_ops xfrm6_dst_ops = {
371         .family =               AF_INET6,
372         .protocol =             __constant_htons(ETH_P_IPV6),
373         .gc =                   xfrm6_garbage_collect,
374         .update_pmtu =          xfrm6_update_pmtu,
375         .destroy =              xfrm6_dst_destroy,
376         .ifdown =               xfrm6_dst_ifdown,
377         .gc_thresh =            1024,
378         .entry_size =           sizeof(struct xfrm_dst),
379 };
380
381 static struct xfrm_policy_afinfo xfrm6_policy_afinfo = {
382         .family =               AF_INET6,
383         .dst_ops =              &xfrm6_dst_ops,
384         .dst_lookup =           xfrm6_dst_lookup,
385         .get_saddr =            xfrm6_get_saddr,
386         .find_bundle =          __xfrm6_find_bundle,
387         .bundle_create =        __xfrm6_bundle_create,
388         .decode_session =       _decode_session6,
389 };
390
391 static void __init xfrm6_policy_init(void)
392 {
393         xfrm_policy_register_afinfo(&xfrm6_policy_afinfo);
394 }
395
396 static void xfrm6_policy_fini(void)
397 {
398         xfrm_policy_unregister_afinfo(&xfrm6_policy_afinfo);
399 }
400
401 void __init xfrm6_init(void)
402 {
403         xfrm6_policy_init();
404         xfrm6_state_init();
405 }
406
407 void xfrm6_fini(void)
408 {
409         //xfrm6_input_fini();
410         xfrm6_policy_fini();
411         xfrm6_state_fini();
412 }