[NET]: Supporting UDP-Lite (RFC 3828) in Linux
[linux-2.6.git] / 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 **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 in6_addr *remote = &fl->fl6_dst;
135         struct in6_addr *local  = &fl->fl6_src;
136         struct flowi fl_tunnel = {
137                 .nl_u = {
138                         .ip6_u = {
139                                 .saddr = *local,
140                                 .daddr = *remote
141                         }
142                 }
143         };
144         int i;
145         int err = 0;
146         int header_len = 0;
147         int nfheader_len = 0;
148         int trailer_len = 0;
149
150         dst = dst_prev = NULL;
151         dst_hold(&rt->u.dst);
152
153         for (i = 0; i < nx; i++) {
154                 struct dst_entry *dst1 = dst_alloc(&xfrm6_dst_ops);
155                 struct xfrm_dst *xdst;
156                 int tunnel = 0;
157
158                 if (unlikely(dst1 == NULL)) {
159                         err = -ENOBUFS;
160                         dst_release(&rt->u.dst);
161                         goto error;
162                 }
163
164                 if (!dst)
165                         dst = dst1;
166                 else {
167                         dst_prev->child = dst1;
168                         dst1->flags |= DST_NOHASH;
169                         dst_clone(dst1);
170                 }
171
172                 xdst = (struct xfrm_dst *)dst1;
173                 xdst->route = &rt->u.dst;
174                 xdst->genid = xfrm[i]->genid;
175                 if (rt->rt6i_node)
176                         xdst->route_cookie = rt->rt6i_node->fn_sernum;
177
178                 dst1->next = dst_prev;
179                 dst_prev = dst1;
180                 if (xfrm[i]->props.mode != XFRM_MODE_TRANSPORT) {
181                         remote = __xfrm6_bundle_addr_remote(xfrm[i], remote);
182                         local  = __xfrm6_bundle_addr_local(xfrm[i], local);
183                         tunnel = 1;
184                 }
185                 __xfrm6_bundle_len_inc(&header_len, &nfheader_len, xfrm[i]);
186                 trailer_len += xfrm[i]->props.trailer_len;
187
188                 if (tunnel) {
189                         ipv6_addr_copy(&fl_tunnel.fl6_dst, remote);
190                         ipv6_addr_copy(&fl_tunnel.fl6_src, local);
191                         err = xfrm_dst_lookup((struct xfrm_dst **) &rt,
192                                               &fl_tunnel, AF_INET6);
193                         if (err)
194                                 goto error;
195                 } else
196                         dst_hold(&rt->u.dst);
197         }
198
199         dst_prev->child = &rt->u.dst;
200         dst->path = &rt->u.dst;
201         if (rt->rt6i_node)
202                 ((struct xfrm_dst *)dst)->path_cookie = rt->rt6i_node->fn_sernum;
203
204         *dst_p = dst;
205         dst = dst_prev;
206
207         dst_prev = *dst_p;
208         i = 0;
209         for (; dst_prev != &rt->u.dst; dst_prev = dst_prev->child) {
210                 struct xfrm_dst *x = (struct xfrm_dst*)dst_prev;
211
212                 dst_prev->xfrm = xfrm[i++];
213                 dst_prev->dev = rt->u.dst.dev;
214                 if (rt->u.dst.dev)
215                         dev_hold(rt->u.dst.dev);
216                 dst_prev->obsolete      = -1;
217                 dst_prev->flags        |= DST_HOST;
218                 dst_prev->lastuse       = jiffies;
219                 dst_prev->header_len    = header_len;
220                 dst_prev->nfheader_len  = nfheader_len;
221                 dst_prev->trailer_len   = trailer_len;
222                 memcpy(&dst_prev->metrics, &x->route->metrics, sizeof(dst_prev->metrics));
223
224                 /* Copy neighbour for reachability confirmation */
225                 dst_prev->neighbour     = neigh_clone(rt->u.dst.neighbour);
226                 dst_prev->input         = rt->u.dst.input;
227                 dst_prev->output        = xfrm6_output;
228                 /* Sheit... I remember I did this right. Apparently,
229                  * it was magically lost, so this code needs audit */
230                 x->u.rt6.rt6i_flags    = rt0->rt6i_flags&(RTCF_BROADCAST|RTCF_MULTICAST|RTCF_LOCAL);
231                 x->u.rt6.rt6i_metric   = rt0->rt6i_metric;
232                 x->u.rt6.rt6i_node     = rt0->rt6i_node;
233                 x->u.rt6.rt6i_gateway  = rt0->rt6i_gateway;
234                 memcpy(&x->u.rt6.rt6i_gateway, &rt0->rt6i_gateway, sizeof(x->u.rt6.rt6i_gateway)); 
235                 x->u.rt6.rt6i_dst      = rt0->rt6i_dst;
236                 x->u.rt6.rt6i_src      = rt0->rt6i_src; 
237                 x->u.rt6.rt6i_idev     = rt0->rt6i_idev;
238                 in6_dev_hold(rt0->rt6i_idev);
239                 __xfrm6_bundle_len_dec(&header_len, &nfheader_len, x->u.dst.xfrm);
240                 trailer_len -= x->u.dst.xfrm->props.trailer_len;
241         }
242
243         xfrm_init_pmtu(dst);
244         return 0;
245
246 error:
247         if (dst)
248                 dst_free(dst);
249         return err;
250 }
251
252 static inline void
253 _decode_session6(struct sk_buff *skb, struct flowi *fl)
254 {
255         u16 offset = skb->h.raw - skb->nh.raw;
256         struct ipv6hdr *hdr = skb->nh.ipv6h;
257         struct ipv6_opt_hdr *exthdr;
258         u8 nexthdr = skb->nh.raw[IP6CB(skb)->nhoff];
259
260         memset(fl, 0, sizeof(struct flowi));
261         ipv6_addr_copy(&fl->fl6_dst, &hdr->daddr);
262         ipv6_addr_copy(&fl->fl6_src, &hdr->saddr);
263
264         while (pskb_may_pull(skb, skb->nh.raw + offset + 1 - skb->data)) {
265                 exthdr = (struct ipv6_opt_hdr*)(skb->nh.raw + offset);
266
267                 switch (nexthdr) {
268                 case NEXTHDR_ROUTING:
269                 case NEXTHDR_HOP:
270                 case NEXTHDR_DEST:
271                         offset += ipv6_optlen(exthdr);
272                         nexthdr = exthdr->nexthdr;
273                         exthdr = (struct ipv6_opt_hdr*)(skb->nh.raw + offset);
274                         break;
275
276                 case IPPROTO_UDP:
277                 case IPPROTO_UDPLITE:
278                 case IPPROTO_TCP:
279                 case IPPROTO_SCTP:
280                 case IPPROTO_DCCP:
281                         if (pskb_may_pull(skb, skb->nh.raw + offset + 4 - skb->data)) {
282                                 __be16 *ports = (__be16 *)exthdr;
283
284                                 fl->fl_ip_sport = ports[0];
285                                 fl->fl_ip_dport = ports[1];
286                         }
287                         fl->proto = nexthdr;
288                         return;
289
290                 case IPPROTO_ICMPV6:
291                         if (pskb_may_pull(skb, skb->nh.raw + offset + 2 - skb->data)) {
292                                 u8 *icmp = (u8 *)exthdr;
293
294                                 fl->fl_icmp_type = icmp[0];
295                                 fl->fl_icmp_code = icmp[1];
296                         }
297                         fl->proto = nexthdr;
298                         return;
299
300 #ifdef CONFIG_IPV6_MIP6
301                 case IPPROTO_MH:
302                         if (pskb_may_pull(skb, skb->nh.raw + offset + 3 - skb->data)) {
303                                 struct ip6_mh *mh;
304                                 mh = (struct ip6_mh *)exthdr;
305
306                                 fl->fl_mh_type = mh->ip6mh_type;
307                         }
308                         fl->proto = nexthdr;
309                         return;
310 #endif
311
312                 /* XXX Why are there these headers? */
313                 case IPPROTO_AH:
314                 case IPPROTO_ESP:
315                 case IPPROTO_COMP:
316                 default:
317                         fl->fl_ipsec_spi = 0;
318                         fl->proto = nexthdr;
319                         return;
320                 };
321         }
322 }
323
324 static inline int xfrm6_garbage_collect(void)
325 {
326         xfrm6_policy_afinfo.garbage_collect();
327         return (atomic_read(&xfrm6_dst_ops.entries) > xfrm6_dst_ops.gc_thresh*2);
328 }
329
330 static void xfrm6_update_pmtu(struct dst_entry *dst, u32 mtu)
331 {
332         struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
333         struct dst_entry *path = xdst->route;
334
335         path->ops->update_pmtu(path, mtu);
336 }
337
338 static void xfrm6_dst_destroy(struct dst_entry *dst)
339 {
340         struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
341
342         if (likely(xdst->u.rt6.rt6i_idev))
343                 in6_dev_put(xdst->u.rt6.rt6i_idev);
344         xfrm_dst_destroy(xdst);
345 }
346
347 static void xfrm6_dst_ifdown(struct dst_entry *dst, struct net_device *dev,
348                              int unregister)
349 {
350         struct xfrm_dst *xdst;
351
352         if (!unregister)
353                 return;
354
355         xdst = (struct xfrm_dst *)dst;
356         if (xdst->u.rt6.rt6i_idev->dev == dev) {
357                 struct inet6_dev *loopback_idev = in6_dev_get(&loopback_dev);
358                 BUG_ON(!loopback_idev);
359
360                 do {
361                         in6_dev_put(xdst->u.rt6.rt6i_idev);
362                         xdst->u.rt6.rt6i_idev = loopback_idev;
363                         in6_dev_hold(loopback_idev);
364                         xdst = (struct xfrm_dst *)xdst->u.dst.child;
365                 } while (xdst->u.dst.xfrm);
366
367                 __in6_dev_put(loopback_idev);
368         }
369
370         xfrm_dst_ifdown(dst, dev);
371 }
372
373 static struct dst_ops xfrm6_dst_ops = {
374         .family =               AF_INET6,
375         .protocol =             __constant_htons(ETH_P_IPV6),
376         .gc =                   xfrm6_garbage_collect,
377         .update_pmtu =          xfrm6_update_pmtu,
378         .destroy =              xfrm6_dst_destroy,
379         .ifdown =               xfrm6_dst_ifdown,
380         .gc_thresh =            1024,
381         .entry_size =           sizeof(struct xfrm_dst),
382 };
383
384 static struct xfrm_policy_afinfo xfrm6_policy_afinfo = {
385         .family =               AF_INET6,
386         .dst_ops =              &xfrm6_dst_ops,
387         .dst_lookup =           xfrm6_dst_lookup,
388         .get_saddr =            xfrm6_get_saddr,
389         .find_bundle =          __xfrm6_find_bundle,
390         .bundle_create =        __xfrm6_bundle_create,
391         .decode_session =       _decode_session6,
392 };
393
394 static void __init xfrm6_policy_init(void)
395 {
396         xfrm_policy_register_afinfo(&xfrm6_policy_afinfo);
397 }
398
399 static void xfrm6_policy_fini(void)
400 {
401         xfrm_policy_unregister_afinfo(&xfrm6_policy_afinfo);
402 }
403
404 void __init xfrm6_init(void)
405 {
406         xfrm6_policy_init();
407         xfrm6_state_init();
408 }
409
410 void xfrm6_fini(void)
411 {
412         //xfrm6_input_fini();
413         xfrm6_policy_fini();
414         xfrm6_state_fini();
415 }