Merge master.kernel.org:/pub/scm/linux/kernel/git/davem/net-2.6
[linux-2.6] / net / ipv4 / xfrm4_policy.c
1 /*
2  * xfrm4_policy.c
3  *
4  * Changes:
5  *      Kazunori MIYAZAWA @USAGI
6  *      YOSHIFUJI Hideaki @USAGI
7  *              Split up af-specific portion
8  *
9  */
10
11 #include <linux/compiler.h>
12 #include <linux/inetdevice.h>
13 #include <net/xfrm.h>
14 #include <net/ip.h>
15
16 static struct dst_ops xfrm4_dst_ops;
17 static struct xfrm_policy_afinfo xfrm4_policy_afinfo;
18
19 static int xfrm4_dst_lookup(struct xfrm_dst **dst, struct flowi *fl)
20 {
21         return __ip_route_output_key((struct rtable**)dst, fl);
22 }
23
24 static int xfrm4_get_saddr(xfrm_address_t *saddr, xfrm_address_t *daddr)
25 {
26         struct rtable *rt;
27         struct flowi fl_tunnel = {
28                 .nl_u = {
29                         .ip4_u = {
30                                 .daddr = daddr->a4,
31                         },
32                 },
33         };
34
35         if (!xfrm4_dst_lookup((struct xfrm_dst **)&rt, &fl_tunnel)) {
36                 saddr->a4 = rt->rt_src;
37                 dst_release(&rt->u.dst);
38                 return 0;
39         }
40         return -EHOSTUNREACH;
41 }
42
43 static struct dst_entry *
44 __xfrm4_find_bundle(struct flowi *fl, struct xfrm_policy *policy)
45 {
46         struct dst_entry *dst;
47
48         read_lock_bh(&policy->lock);
49         for (dst = policy->bundles; dst; dst = dst->next) {
50                 struct xfrm_dst *xdst = (struct xfrm_dst*)dst;
51                 if (xdst->u.rt.fl.oif == fl->oif &&     /*XXX*/
52                     xdst->u.rt.fl.fl4_dst == fl->fl4_dst &&
53                     xdst->u.rt.fl.fl4_src == fl->fl4_src &&
54                     xdst->u.rt.fl.fl4_tos == fl->fl4_tos &&
55                     xfrm_bundle_ok(policy, xdst, fl, AF_INET, 0)) {
56                         dst_clone(dst);
57                         break;
58                 }
59         }
60         read_unlock_bh(&policy->lock);
61         return dst;
62 }
63
64 /* Allocate chain of dst_entry's, attach known xfrm's, calculate
65  * all the metrics... Shortly, bundle a bundle.
66  */
67
68 static int
69 __xfrm4_bundle_create(struct xfrm_policy *policy, struct xfrm_state **xfrm, int nx,
70                       struct flowi *fl, struct dst_entry **dst_p)
71 {
72         struct dst_entry *dst, *dst_prev;
73         struct rtable *rt0 = (struct rtable*)(*dst_p);
74         struct rtable *rt = rt0;
75         struct flowi fl_tunnel = {
76                 .nl_u = {
77                         .ip4_u = {
78                                 .saddr = fl->fl4_src,
79                                 .daddr = fl->fl4_dst,
80                                 .tos = fl->fl4_tos
81                         }
82                 }
83         };
84         int i;
85         int err;
86         int header_len = 0;
87         int trailer_len = 0;
88
89         dst = dst_prev = NULL;
90         dst_hold(&rt->u.dst);
91
92         for (i = 0; i < nx; i++) {
93                 struct dst_entry *dst1 = dst_alloc(&xfrm4_dst_ops);
94                 struct xfrm_dst *xdst;
95
96                 if (unlikely(dst1 == NULL)) {
97                         err = -ENOBUFS;
98                         dst_release(&rt->u.dst);
99                         goto error;
100                 }
101
102                 if (!dst)
103                         dst = dst1;
104                 else {
105                         dst_prev->child = dst1;
106                         dst1->flags |= DST_NOHASH;
107                         dst_clone(dst1);
108                 }
109
110                 xdst = (struct xfrm_dst *)dst1;
111                 xdst->route = &rt->u.dst;
112                 xdst->genid = xfrm[i]->genid;
113
114                 dst1->next = dst_prev;
115                 dst_prev = dst1;
116
117                 header_len += xfrm[i]->props.header_len;
118                 trailer_len += xfrm[i]->props.trailer_len;
119
120                 if (xfrm[i]->props.mode == XFRM_MODE_TUNNEL) {
121                         unsigned short encap_family = xfrm[i]->props.family;
122                         switch (encap_family) {
123                         case AF_INET:
124                                 fl_tunnel.fl4_dst = xfrm[i]->id.daddr.a4;
125                                 fl_tunnel.fl4_src = xfrm[i]->props.saddr.a4;
126                                 break;
127 #if defined(CONFIG_IPV6) || defined (CONFIG_IPV6_MODULE)
128                         case AF_INET6:
129                                 ipv6_addr_copy(&fl_tunnel.fl6_dst, (struct in6_addr*)&xfrm[i]->id.daddr.a6);
130                                 ipv6_addr_copy(&fl_tunnel.fl6_src, (struct in6_addr*)&xfrm[i]->props.saddr.a6);
131                                 break;
132 #endif
133                         default:
134                                 BUG_ON(1);
135                         }
136                         err = xfrm_dst_lookup((struct xfrm_dst **)&rt,
137                                               &fl_tunnel, encap_family);
138                         if (err)
139                                 goto error;
140                 } else
141                         dst_hold(&rt->u.dst);
142         }
143
144         dst_prev->child = &rt->u.dst;
145         dst->path = &rt->u.dst;
146
147         *dst_p = dst;
148         dst = dst_prev;
149
150         dst_prev = *dst_p;
151         i = 0;
152         for (; dst_prev != &rt->u.dst; dst_prev = dst_prev->child) {
153                 struct xfrm_dst *x = (struct xfrm_dst*)dst_prev;
154                 struct xfrm_state_afinfo *afinfo;
155                 x->u.rt.fl = *fl;
156
157                 dst_prev->xfrm = xfrm[i++];
158                 dst_prev->dev = rt->u.dst.dev;
159                 if (rt->u.dst.dev)
160                         dev_hold(rt->u.dst.dev);
161                 dst_prev->obsolete      = -1;
162                 dst_prev->flags        |= DST_HOST;
163                 dst_prev->lastuse       = jiffies;
164                 dst_prev->header_len    = header_len;
165                 dst_prev->nfheader_len  = 0;
166                 dst_prev->trailer_len   = trailer_len;
167                 memcpy(&dst_prev->metrics, &x->route->metrics, sizeof(dst_prev->metrics));
168
169                 /* Copy neighbout for reachability confirmation */
170                 dst_prev->neighbour     = neigh_clone(rt->u.dst.neighbour);
171                 dst_prev->input         = rt->u.dst.input;
172                 /* XXX: When IPv6 module can be unloaded, we should manage reference
173                  * to xfrm6_output in afinfo->output. Miyazawa
174                  * */
175                 afinfo = xfrm_state_get_afinfo(dst_prev->xfrm->props.family);
176                 if (!afinfo) {
177                         dst = *dst_p;
178                         err = -EAFNOSUPPORT;
179                         goto error;
180                 }
181                 dst_prev->output = afinfo->output;
182                 xfrm_state_put_afinfo(afinfo);
183                 if (dst_prev->xfrm->props.family == AF_INET && rt->peer)
184                         atomic_inc(&rt->peer->refcnt);
185                 x->u.rt.peer = rt->peer;
186                 /* Sheit... I remember I did this right. Apparently,
187                  * it was magically lost, so this code needs audit */
188                 x->u.rt.rt_flags = rt0->rt_flags&(RTCF_BROADCAST|RTCF_MULTICAST|RTCF_LOCAL);
189                 x->u.rt.rt_type = rt->rt_type;
190                 x->u.rt.rt_src = rt0->rt_src;
191                 x->u.rt.rt_dst = rt0->rt_dst;
192                 x->u.rt.rt_gateway = rt->rt_gateway;
193                 x->u.rt.rt_spec_dst = rt0->rt_spec_dst;
194                 x->u.rt.idev = rt0->idev;
195                 in_dev_hold(rt0->idev);
196                 header_len -= x->u.dst.xfrm->props.header_len;
197                 trailer_len -= x->u.dst.xfrm->props.trailer_len;
198         }
199
200         xfrm_init_pmtu(dst);
201         return 0;
202
203 error:
204         if (dst)
205                 dst_free(dst);
206         return err;
207 }
208
209 static void
210 _decode_session4(struct sk_buff *skb, struct flowi *fl)
211 {
212         struct iphdr *iph = ip_hdr(skb);
213         u8 *xprth = skb_network_header(skb) + iph->ihl * 4;
214
215         memset(fl, 0, sizeof(struct flowi));
216         if (!(iph->frag_off & htons(IP_MF | IP_OFFSET))) {
217                 switch (iph->protocol) {
218                 case IPPROTO_UDP:
219                 case IPPROTO_UDPLITE:
220                 case IPPROTO_TCP:
221                 case IPPROTO_SCTP:
222                 case IPPROTO_DCCP:
223                         if (pskb_may_pull(skb, xprth + 4 - skb->data)) {
224                                 __be16 *ports = (__be16 *)xprth;
225
226                                 fl->fl_ip_sport = ports[0];
227                                 fl->fl_ip_dport = ports[1];
228                         }
229                         break;
230
231                 case IPPROTO_ICMP:
232                         if (pskb_may_pull(skb, xprth + 2 - skb->data)) {
233                                 u8 *icmp = xprth;
234
235                                 fl->fl_icmp_type = icmp[0];
236                                 fl->fl_icmp_code = icmp[1];
237                         }
238                         break;
239
240                 case IPPROTO_ESP:
241                         if (pskb_may_pull(skb, xprth + 4 - skb->data)) {
242                                 __be32 *ehdr = (__be32 *)xprth;
243
244                                 fl->fl_ipsec_spi = ehdr[0];
245                         }
246                         break;
247
248                 case IPPROTO_AH:
249                         if (pskb_may_pull(skb, xprth + 8 - skb->data)) {
250                                 __be32 *ah_hdr = (__be32*)xprth;
251
252                                 fl->fl_ipsec_spi = ah_hdr[1];
253                         }
254                         break;
255
256                 case IPPROTO_COMP:
257                         if (pskb_may_pull(skb, xprth + 4 - skb->data)) {
258                                 __be16 *ipcomp_hdr = (__be16 *)xprth;
259
260                                 fl->fl_ipsec_spi = htonl(ntohs(ipcomp_hdr[1]));
261                         }
262                         break;
263                 default:
264                         fl->fl_ipsec_spi = 0;
265                         break;
266                 }
267         }
268         fl->proto = iph->protocol;
269         fl->fl4_dst = iph->daddr;
270         fl->fl4_src = iph->saddr;
271         fl->fl4_tos = iph->tos;
272 }
273
274 static inline int xfrm4_garbage_collect(void)
275 {
276         xfrm4_policy_afinfo.garbage_collect();
277         return (atomic_read(&xfrm4_dst_ops.entries) > xfrm4_dst_ops.gc_thresh*2);
278 }
279
280 static void xfrm4_update_pmtu(struct dst_entry *dst, u32 mtu)
281 {
282         struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
283         struct dst_entry *path = xdst->route;
284
285         path->ops->update_pmtu(path, mtu);
286 }
287
288 static void xfrm4_dst_destroy(struct dst_entry *dst)
289 {
290         struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
291
292         if (likely(xdst->u.rt.idev))
293                 in_dev_put(xdst->u.rt.idev);
294         if (dst->xfrm && dst->xfrm->props.family == AF_INET && likely(xdst->u.rt.peer))
295                 inet_putpeer(xdst->u.rt.peer);
296         xfrm_dst_destroy(xdst);
297 }
298
299 static void xfrm4_dst_ifdown(struct dst_entry *dst, struct net_device *dev,
300                              int unregister)
301 {
302         struct xfrm_dst *xdst;
303
304         if (!unregister)
305                 return;
306
307         xdst = (struct xfrm_dst *)dst;
308         if (xdst->u.rt.idev->dev == dev) {
309                 struct in_device *loopback_idev = in_dev_get(&loopback_dev);
310                 BUG_ON(!loopback_idev);
311
312                 do {
313                         in_dev_put(xdst->u.rt.idev);
314                         xdst->u.rt.idev = loopback_idev;
315                         in_dev_hold(loopback_idev);
316                         xdst = (struct xfrm_dst *)xdst->u.dst.child;
317                 } while (xdst->u.dst.xfrm);
318
319                 __in_dev_put(loopback_idev);
320         }
321
322         xfrm_dst_ifdown(dst, dev);
323 }
324
325 static struct dst_ops xfrm4_dst_ops = {
326         .family =               AF_INET,
327         .protocol =             __constant_htons(ETH_P_IP),
328         .gc =                   xfrm4_garbage_collect,
329         .update_pmtu =          xfrm4_update_pmtu,
330         .destroy =              xfrm4_dst_destroy,
331         .ifdown =               xfrm4_dst_ifdown,
332         .gc_thresh =            1024,
333         .entry_size =           sizeof(struct xfrm_dst),
334 };
335
336 static struct xfrm_policy_afinfo xfrm4_policy_afinfo = {
337         .family =               AF_INET,
338         .dst_ops =              &xfrm4_dst_ops,
339         .dst_lookup =           xfrm4_dst_lookup,
340         .get_saddr =            xfrm4_get_saddr,
341         .find_bundle =          __xfrm4_find_bundle,
342         .bundle_create =        __xfrm4_bundle_create,
343         .decode_session =       _decode_session4,
344 };
345
346 static void __init xfrm4_policy_init(void)
347 {
348         xfrm_policy_register_afinfo(&xfrm4_policy_afinfo);
349 }
350
351 static void __exit xfrm4_policy_fini(void)
352 {
353         xfrm_policy_unregister_afinfo(&xfrm4_policy_afinfo);
354 }
355
356 void __init xfrm4_init(void)
357 {
358         xfrm4_state_init();
359         xfrm4_policy_init();
360 }
361