Merge branch 'master' of master.kernel.org:/pub/scm/linux/kernel/git/davem/net-2.6
[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 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
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 **xdst, struct flowi *fl)
29 {
30         struct dst_entry *dst = ip6_route_output(NULL, fl);
31         int err = dst->error;
32         if (!err)
33                 *xdst = (struct xfrm_dst *) dst;
34         else
35                 dst_release(dst);
36         return err;
37 }
38
39 static int xfrm6_get_saddr(xfrm_address_t *saddr, xfrm_address_t *daddr)
40 {
41         struct rt6_info *rt;
42         struct flowi fl_tunnel = {
43                 .nl_u = {
44                         .ip6_u = {
45                                 .daddr = *(struct in6_addr *)&daddr->a6,
46                         },
47                 },
48         };
49
50         if (!xfrm6_dst_lookup((struct xfrm_dst **)&rt, &fl_tunnel)) {
51                 ipv6_get_saddr(&rt->u.dst, (struct in6_addr *)&daddr->a6,
52                                (struct in6_addr *)&saddr->a6);
53                 dst_release(&rt->u.dst);
54                 return 0;
55         }
56         return -EHOSTUNREACH;
57 }
58
59 static struct dst_entry *
60 __xfrm6_find_bundle(struct flowi *fl, struct xfrm_policy *policy)
61 {
62         struct dst_entry *dst;
63
64         /* Still not clear if we should set fl->fl6_{src,dst}... */
65         read_lock_bh(&policy->lock);
66         for (dst = policy->bundles; dst; dst = dst->next) {
67                 struct xfrm_dst *xdst = (struct xfrm_dst*)dst;
68                 struct in6_addr fl_dst_prefix, fl_src_prefix;
69
70                 ipv6_addr_prefix(&fl_dst_prefix,
71                                  &fl->fl6_dst,
72                                  xdst->u.rt6.rt6i_dst.plen);
73                 ipv6_addr_prefix(&fl_src_prefix,
74                                  &fl->fl6_src,
75                                  xdst->u.rt6.rt6i_src.plen);
76                 if (ipv6_addr_equal(&xdst->u.rt6.rt6i_dst.addr, &fl_dst_prefix) &&
77                     ipv6_addr_equal(&xdst->u.rt6.rt6i_src.addr, &fl_src_prefix) &&
78                     xfrm_bundle_ok(policy, xdst, fl, AF_INET6,
79                                    (xdst->u.rt6.rt6i_dst.plen != 128 ||
80                                     xdst->u.rt6.rt6i_src.plen != 128))) {
81                         dst_clone(dst);
82                         break;
83                 }
84         }
85         read_unlock_bh(&policy->lock);
86         return dst;
87 }
88
89 static inline struct in6_addr*
90 __xfrm6_bundle_addr_remote(struct xfrm_state *x, struct in6_addr *addr)
91 {
92         return (x->type->remote_addr) ?
93                 (struct in6_addr*)x->type->remote_addr(x, (xfrm_address_t *)addr) :
94                 (struct in6_addr*)&x->id.daddr;
95 }
96
97 static inline struct in6_addr*
98 __xfrm6_bundle_addr_local(struct xfrm_state *x, struct in6_addr *addr)
99 {
100         return (x->type->local_addr) ?
101                 (struct in6_addr*)x->type->local_addr(x, (xfrm_address_t *)addr) :
102                 (struct in6_addr*)&x->props.saddr;
103 }
104
105 static inline void
106 __xfrm6_bundle_len_inc(int *len, int *nflen, struct xfrm_state *x)
107 {
108         if (x->type->flags & XFRM_TYPE_NON_FRAGMENT)
109                 *nflen += x->props.header_len;
110         else
111                 *len += x->props.header_len;
112 }
113
114 static inline void
115 __xfrm6_bundle_len_dec(int *len, int *nflen, struct xfrm_state *x)
116 {
117         if (x->type->flags & XFRM_TYPE_NON_FRAGMENT)
118                 *nflen -= x->props.header_len;
119         else
120                 *len -= x->props.header_len;
121 }
122
123 /* Allocate chain of dst_entry's, attach known xfrm's, calculate
124  * all the metrics... Shortly, bundle a bundle.
125  */
126
127 static int
128 __xfrm6_bundle_create(struct xfrm_policy *policy, struct xfrm_state **xfrm, int nx,
129                       struct flowi *fl, struct dst_entry **dst_p)
130 {
131         struct dst_entry *dst, *dst_prev;
132         struct rt6_info *rt0 = (struct rt6_info*)(*dst_p);
133         struct rt6_info *rt  = rt0;
134         struct flowi fl_tunnel = {
135                 .nl_u = {
136                         .ip6_u = {
137                                 .saddr = fl->fl6_src,
138                                 .daddr = fl->fl6_dst,
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
155                 if (unlikely(dst1 == NULL)) {
156                         err = -ENOBUFS;
157                         dst_release(&rt->u.dst);
158                         goto error;
159                 }
160
161                 if (!dst)
162                         dst = dst1;
163                 else {
164                         dst_prev->child = dst1;
165                         dst1->flags |= DST_NOHASH;
166                         dst_clone(dst1);
167                 }
168
169                 xdst = (struct xfrm_dst *)dst1;
170                 xdst->route = &rt->u.dst;
171                 xdst->genid = xfrm[i]->genid;
172                 if (rt->rt6i_node)
173                         xdst->route_cookie = rt->rt6i_node->fn_sernum;
174
175                 dst1->next = dst_prev;
176                 dst_prev = dst1;
177
178                 __xfrm6_bundle_len_inc(&header_len, &nfheader_len, xfrm[i]);
179                 trailer_len += xfrm[i]->props.trailer_len;
180
181                 if (xfrm[i]->props.mode == XFRM_MODE_TUNNEL ||
182                     xfrm[i]->props.mode == XFRM_MODE_ROUTEOPTIMIZATION) {
183                         unsigned short encap_family = xfrm[i]->props.family;
184                         switch(encap_family) {
185                         case AF_INET:
186                                 fl_tunnel.fl4_dst = xfrm[i]->id.daddr.a4;
187                                 fl_tunnel.fl4_src = xfrm[i]->props.saddr.a4;
188                                 break;
189                         case AF_INET6:
190                                 ipv6_addr_copy(&fl_tunnel.fl6_dst, __xfrm6_bundle_addr_remote(xfrm[i], &fl->fl6_dst));
191
192                                 ipv6_addr_copy(&fl_tunnel.fl6_src, __xfrm6_bundle_addr_local(xfrm[i], &fl->fl6_src));
193                                 break;
194                         default:
195                                 BUG_ON(1);
196                         }
197
198                         err = xfrm_dst_lookup((struct xfrm_dst **) &rt,
199                                               &fl_tunnel, encap_family);
200                         if (err)
201                                 goto error;
202                 } else
203                         dst_hold(&rt->u.dst);
204         }
205
206         dst_prev->child = &rt->u.dst;
207         dst->path = &rt->u.dst;
208         if (rt->rt6i_node)
209                 ((struct xfrm_dst *)dst)->path_cookie = rt->rt6i_node->fn_sernum;
210
211         *dst_p = dst;
212         dst = dst_prev;
213
214         dst_prev = *dst_p;
215         i = 0;
216         for (; dst_prev != &rt->u.dst; dst_prev = dst_prev->child) {
217                 struct xfrm_dst *x = (struct xfrm_dst*)dst_prev;
218                 struct xfrm_state_afinfo *afinfo;
219
220                 dst_prev->xfrm = xfrm[i++];
221                 dst_prev->dev = rt->u.dst.dev;
222                 if (rt->u.dst.dev)
223                         dev_hold(rt->u.dst.dev);
224                 dst_prev->obsolete      = -1;
225                 dst_prev->flags        |= DST_HOST;
226                 dst_prev->lastuse       = jiffies;
227                 dst_prev->header_len    = header_len;
228                 dst_prev->nfheader_len  = nfheader_len;
229                 dst_prev->trailer_len   = trailer_len;
230                 memcpy(&dst_prev->metrics, &x->route->metrics, sizeof(dst_prev->metrics));
231
232                 /* Copy neighbour for reachability confirmation */
233                 dst_prev->neighbour     = neigh_clone(rt->u.dst.neighbour);
234                 dst_prev->input         = rt->u.dst.input;
235                 /* XXX: When IPv4 is implemented as module and can be unloaded,
236                  * we should manage reference to xfrm4_output in afinfo->output.
237                  * Miyazawa
238                  */
239                 afinfo = xfrm_state_get_afinfo(dst_prev->xfrm->props.family);
240                 if (!afinfo) {
241                         dst = *dst_p;
242                         goto error;
243                 }
244
245                 dst_prev->output = afinfo->output;
246                 xfrm_state_put_afinfo(afinfo);
247                 /* Sheit... I remember I did this right. Apparently,
248                  * it was magically lost, so this code needs audit */
249                 x->u.rt6.rt6i_flags    = rt0->rt6i_flags&(RTCF_BROADCAST|RTCF_MULTICAST|RTCF_LOCAL);
250                 x->u.rt6.rt6i_metric   = rt0->rt6i_metric;
251                 x->u.rt6.rt6i_node     = rt0->rt6i_node;
252                 x->u.rt6.rt6i_gateway  = rt0->rt6i_gateway;
253                 memcpy(&x->u.rt6.rt6i_gateway, &rt0->rt6i_gateway, sizeof(x->u.rt6.rt6i_gateway));
254                 x->u.rt6.rt6i_dst      = rt0->rt6i_dst;
255                 x->u.rt6.rt6i_src      = rt0->rt6i_src;
256                 x->u.rt6.rt6i_idev     = rt0->rt6i_idev;
257                 in6_dev_hold(rt0->rt6i_idev);
258                 __xfrm6_bundle_len_dec(&header_len, &nfheader_len, x->u.dst.xfrm);
259                 trailer_len -= x->u.dst.xfrm->props.trailer_len;
260         }
261
262         xfrm_init_pmtu(dst);
263         return 0;
264
265 error:
266         if (dst)
267                 dst_free(dst);
268         return err;
269 }
270
271 static inline void
272 _decode_session6(struct sk_buff *skb, struct flowi *fl)
273 {
274         u16 offset = skb_network_header_len(skb);
275         struct ipv6hdr *hdr = ipv6_hdr(skb);
276         struct ipv6_opt_hdr *exthdr;
277         const unsigned char *nh = skb_network_header(skb);
278         u8 nexthdr = nh[IP6CB(skb)->nhoff];
279
280         memset(fl, 0, sizeof(struct flowi));
281         ipv6_addr_copy(&fl->fl6_dst, &hdr->daddr);
282         ipv6_addr_copy(&fl->fl6_src, &hdr->saddr);
283
284         while (pskb_may_pull(skb, nh + offset + 1 - skb->data)) {
285                 nh = skb_network_header(skb);
286                 exthdr = (struct ipv6_opt_hdr *)(nh + offset);
287
288                 switch (nexthdr) {
289                 case NEXTHDR_ROUTING:
290                 case NEXTHDR_HOP:
291                 case NEXTHDR_DEST:
292                         offset += ipv6_optlen(exthdr);
293                         nexthdr = exthdr->nexthdr;
294                         exthdr = (struct ipv6_opt_hdr *)(nh + offset);
295                         break;
296
297                 case IPPROTO_UDP:
298                 case IPPROTO_UDPLITE:
299                 case IPPROTO_TCP:
300                 case IPPROTO_SCTP:
301                 case IPPROTO_DCCP:
302                         if (pskb_may_pull(skb, nh + offset + 4 - skb->data)) {
303                                 __be16 *ports = (__be16 *)exthdr;
304
305                                 fl->fl_ip_sport = ports[0];
306                                 fl->fl_ip_dport = ports[1];
307                         }
308                         fl->proto = nexthdr;
309                         return;
310
311                 case IPPROTO_ICMPV6:
312                         if (pskb_may_pull(skb, nh + offset + 2 - skb->data)) {
313                                 u8 *icmp = (u8 *)exthdr;
314
315                                 fl->fl_icmp_type = icmp[0];
316                                 fl->fl_icmp_code = icmp[1];
317                         }
318                         fl->proto = nexthdr;
319                         return;
320
321 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
322                 case IPPROTO_MH:
323                         if (pskb_may_pull(skb, nh + offset + 3 - skb->data)) {
324                                 struct ip6_mh *mh;
325                                 mh = (struct ip6_mh *)exthdr;
326
327                                 fl->fl_mh_type = mh->ip6mh_type;
328                         }
329                         fl->proto = nexthdr;
330                         return;
331 #endif
332
333                 /* XXX Why are there these headers? */
334                 case IPPROTO_AH:
335                 case IPPROTO_ESP:
336                 case IPPROTO_COMP:
337                 default:
338                         fl->fl_ipsec_spi = 0;
339                         fl->proto = nexthdr;
340                         return;
341                 }
342         }
343 }
344
345 static inline int xfrm6_garbage_collect(void)
346 {
347         xfrm6_policy_afinfo.garbage_collect();
348         return (atomic_read(&xfrm6_dst_ops.entries) > xfrm6_dst_ops.gc_thresh*2);
349 }
350
351 static void xfrm6_update_pmtu(struct dst_entry *dst, u32 mtu)
352 {
353         struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
354         struct dst_entry *path = xdst->route;
355
356         path->ops->update_pmtu(path, mtu);
357 }
358
359 static void xfrm6_dst_destroy(struct dst_entry *dst)
360 {
361         struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
362
363         if (likely(xdst->u.rt6.rt6i_idev))
364                 in6_dev_put(xdst->u.rt6.rt6i_idev);
365         xfrm_dst_destroy(xdst);
366 }
367
368 static void xfrm6_dst_ifdown(struct dst_entry *dst, struct net_device *dev,
369                              int unregister)
370 {
371         struct xfrm_dst *xdst;
372
373         if (!unregister)
374                 return;
375
376         xdst = (struct xfrm_dst *)dst;
377         if (xdst->u.rt6.rt6i_idev->dev == dev) {
378                 struct inet6_dev *loopback_idev = in6_dev_get(&loopback_dev);
379                 BUG_ON(!loopback_idev);
380
381                 do {
382                         in6_dev_put(xdst->u.rt6.rt6i_idev);
383                         xdst->u.rt6.rt6i_idev = loopback_idev;
384                         in6_dev_hold(loopback_idev);
385                         xdst = (struct xfrm_dst *)xdst->u.dst.child;
386                 } while (xdst->u.dst.xfrm);
387
388                 __in6_dev_put(loopback_idev);
389         }
390
391         xfrm_dst_ifdown(dst, dev);
392 }
393
394 static struct dst_ops xfrm6_dst_ops = {
395         .family =               AF_INET6,
396         .protocol =             __constant_htons(ETH_P_IPV6),
397         .gc =                   xfrm6_garbage_collect,
398         .update_pmtu =          xfrm6_update_pmtu,
399         .destroy =              xfrm6_dst_destroy,
400         .ifdown =               xfrm6_dst_ifdown,
401         .gc_thresh =            1024,
402         .entry_size =           sizeof(struct xfrm_dst),
403 };
404
405 static struct xfrm_policy_afinfo xfrm6_policy_afinfo = {
406         .family =               AF_INET6,
407         .dst_ops =              &xfrm6_dst_ops,
408         .dst_lookup =           xfrm6_dst_lookup,
409         .get_saddr =            xfrm6_get_saddr,
410         .find_bundle =          __xfrm6_find_bundle,
411         .bundle_create =        __xfrm6_bundle_create,
412         .decode_session =       _decode_session6,
413 };
414
415 static void __init xfrm6_policy_init(void)
416 {
417         xfrm_policy_register_afinfo(&xfrm6_policy_afinfo);
418 }
419
420 static void xfrm6_policy_fini(void)
421 {
422         xfrm_policy_unregister_afinfo(&xfrm6_policy_afinfo);
423 }
424
425 void __init xfrm6_init(void)
426 {
427         xfrm6_policy_init();
428         xfrm6_state_init();
429 }
430
431 void xfrm6_fini(void)
432 {
433         //xfrm6_input_fini();
434         xfrm6_policy_fini();
435         xfrm6_state_fini();
436 }