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