[SPARC64]: Fix 2 bugs in huge page support.
[linux-2.6/verdex.git] / net / ipv4 / xfrm4_policy.c
blobf285bbf296e28d79099887afa2882be75ddf32bc
1 /*
2 * xfrm4_policy.c
4 * Changes:
5 * Kazunori MIYAZAWA @USAGI
6 * YOSHIFUJI Hideaki @USAGI
7 * Split up af-specific portion
8 *
9 */
11 #include <linux/compiler.h>
12 #include <linux/config.h>
13 #include <linux/inetdevice.h>
14 #include <net/xfrm.h>
15 #include <net/ip.h>
17 static struct dst_ops xfrm4_dst_ops;
18 static struct xfrm_policy_afinfo xfrm4_policy_afinfo;
20 static struct xfrm_type_map xfrm4_type_map = { .lock = RW_LOCK_UNLOCKED };
22 static int xfrm4_dst_lookup(struct xfrm_dst **dst, struct flowi *fl)
24 return __ip_route_output_key((struct rtable**)dst, fl);
27 static struct dst_entry *
28 __xfrm4_find_bundle(struct flowi *fl, struct xfrm_policy *policy)
30 struct dst_entry *dst;
32 read_lock_bh(&policy->lock);
33 for (dst = policy->bundles; dst; dst = dst->next) {
34 struct xfrm_dst *xdst = (struct xfrm_dst*)dst;
35 if (xdst->u.rt.fl.oif == fl->oif && /*XXX*/
36 xdst->u.rt.fl.fl4_dst == fl->fl4_dst &&
37 xdst->u.rt.fl.fl4_src == fl->fl4_src &&
38 xdst->u.rt.fl.fl4_tos == fl->fl4_tos &&
39 xfrm_bundle_ok(xdst, fl, AF_INET)) {
40 dst_clone(dst);
41 break;
44 read_unlock_bh(&policy->lock);
45 return dst;
48 /* Allocate chain of dst_entry's, attach known xfrm's, calculate
49 * all the metrics... Shortly, bundle a bundle.
52 static int
53 __xfrm4_bundle_create(struct xfrm_policy *policy, struct xfrm_state **xfrm, int nx,
54 struct flowi *fl, struct dst_entry **dst_p)
56 struct dst_entry *dst, *dst_prev;
57 struct rtable *rt0 = (struct rtable*)(*dst_p);
58 struct rtable *rt = rt0;
59 u32 remote = fl->fl4_dst;
60 u32 local = fl->fl4_src;
61 struct flowi fl_tunnel = {
62 .nl_u = {
63 .ip4_u = {
64 .saddr = local,
65 .daddr = remote,
66 .tos = fl->fl4_tos
70 int i;
71 int err;
72 int header_len = 0;
73 int trailer_len = 0;
75 dst = dst_prev = NULL;
76 dst_hold(&rt->u.dst);
78 for (i = 0; i < nx; i++) {
79 struct dst_entry *dst1 = dst_alloc(&xfrm4_dst_ops);
80 struct xfrm_dst *xdst;
81 int tunnel = 0;
83 if (unlikely(dst1 == NULL)) {
84 err = -ENOBUFS;
85 dst_release(&rt->u.dst);
86 goto error;
89 if (!dst)
90 dst = dst1;
91 else {
92 dst_prev->child = dst1;
93 dst1->flags |= DST_NOHASH;
94 dst_clone(dst1);
97 xdst = (struct xfrm_dst *)dst1;
98 xdst->route = &rt->u.dst;
100 dst1->next = dst_prev;
101 dst_prev = dst1;
102 if (xfrm[i]->props.mode) {
103 remote = xfrm[i]->id.daddr.a4;
104 local = xfrm[i]->props.saddr.a4;
105 tunnel = 1;
107 header_len += xfrm[i]->props.header_len;
108 trailer_len += xfrm[i]->props.trailer_len;
110 if (tunnel) {
111 fl_tunnel.fl4_src = local;
112 fl_tunnel.fl4_dst = remote;
113 err = xfrm_dst_lookup((struct xfrm_dst **)&rt,
114 &fl_tunnel, AF_INET);
115 if (err)
116 goto error;
117 } else
118 dst_hold(&rt->u.dst);
121 dst_prev->child = &rt->u.dst;
122 dst->path = &rt->u.dst;
124 *dst_p = dst;
125 dst = dst_prev;
127 dst_prev = *dst_p;
128 i = 0;
129 for (; dst_prev != &rt->u.dst; dst_prev = dst_prev->child) {
130 struct xfrm_dst *x = (struct xfrm_dst*)dst_prev;
131 x->u.rt.fl = *fl;
133 dst_prev->xfrm = xfrm[i++];
134 dst_prev->dev = rt->u.dst.dev;
135 if (rt->u.dst.dev)
136 dev_hold(rt->u.dst.dev);
137 dst_prev->obsolete = -1;
138 dst_prev->flags |= DST_HOST;
139 dst_prev->lastuse = jiffies;
140 dst_prev->header_len = header_len;
141 dst_prev->trailer_len = trailer_len;
142 memcpy(&dst_prev->metrics, &x->route->metrics, sizeof(dst_prev->metrics));
144 /* Copy neighbout for reachability confirmation */
145 dst_prev->neighbour = neigh_clone(rt->u.dst.neighbour);
146 dst_prev->input = rt->u.dst.input;
147 dst_prev->output = xfrm4_output;
148 if (rt->peer)
149 atomic_inc(&rt->peer->refcnt);
150 x->u.rt.peer = rt->peer;
151 /* Sheit... I remember I did this right. Apparently,
152 * it was magically lost, so this code needs audit */
153 x->u.rt.rt_flags = rt0->rt_flags&(RTCF_BROADCAST|RTCF_MULTICAST|RTCF_LOCAL);
154 x->u.rt.rt_type = rt->rt_type;
155 x->u.rt.rt_src = rt0->rt_src;
156 x->u.rt.rt_dst = rt0->rt_dst;
157 x->u.rt.rt_gateway = rt->rt_gateway;
158 x->u.rt.rt_spec_dst = rt0->rt_spec_dst;
159 x->u.rt.idev = rt0->idev;
160 in_dev_hold(rt0->idev);
161 header_len -= x->u.dst.xfrm->props.header_len;
162 trailer_len -= x->u.dst.xfrm->props.trailer_len;
165 xfrm_init_pmtu(dst);
166 return 0;
168 error:
169 if (dst)
170 dst_free(dst);
171 return err;
174 static void
175 _decode_session4(struct sk_buff *skb, struct flowi *fl)
177 struct iphdr *iph = skb->nh.iph;
178 u8 *xprth = skb->nh.raw + iph->ihl*4;
180 memset(fl, 0, sizeof(struct flowi));
181 if (!(iph->frag_off & htons(IP_MF | IP_OFFSET))) {
182 switch (iph->protocol) {
183 case IPPROTO_UDP:
184 case IPPROTO_TCP:
185 case IPPROTO_SCTP:
186 case IPPROTO_DCCP:
187 if (pskb_may_pull(skb, xprth + 4 - skb->data)) {
188 u16 *ports = (u16 *)xprth;
190 fl->fl_ip_sport = ports[0];
191 fl->fl_ip_dport = ports[1];
193 break;
195 case IPPROTO_ICMP:
196 if (pskb_may_pull(skb, xprth + 2 - skb->data)) {
197 u8 *icmp = xprth;
199 fl->fl_icmp_type = icmp[0];
200 fl->fl_icmp_code = icmp[1];
202 break;
204 case IPPROTO_ESP:
205 if (pskb_may_pull(skb, xprth + 4 - skb->data)) {
206 u32 *ehdr = (u32 *)xprth;
208 fl->fl_ipsec_spi = ehdr[0];
210 break;
212 case IPPROTO_AH:
213 if (pskb_may_pull(skb, xprth + 8 - skb->data)) {
214 u32 *ah_hdr = (u32*)xprth;
216 fl->fl_ipsec_spi = ah_hdr[1];
218 break;
220 case IPPROTO_COMP:
221 if (pskb_may_pull(skb, xprth + 4 - skb->data)) {
222 u16 *ipcomp_hdr = (u16 *)xprth;
224 fl->fl_ipsec_spi = ntohl(ntohs(ipcomp_hdr[1]));
226 break;
227 default:
228 fl->fl_ipsec_spi = 0;
229 break;
232 fl->proto = iph->protocol;
233 fl->fl4_dst = iph->daddr;
234 fl->fl4_src = iph->saddr;
235 fl->fl4_tos = iph->tos;
238 static inline int xfrm4_garbage_collect(void)
240 read_lock(&xfrm4_policy_afinfo.lock);
241 xfrm4_policy_afinfo.garbage_collect();
242 read_unlock(&xfrm4_policy_afinfo.lock);
243 return (atomic_read(&xfrm4_dst_ops.entries) > xfrm4_dst_ops.gc_thresh*2);
246 static void xfrm4_update_pmtu(struct dst_entry *dst, u32 mtu)
248 struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
249 struct dst_entry *path = xdst->route;
251 path->ops->update_pmtu(path, mtu);
254 static void xfrm4_dst_destroy(struct dst_entry *dst)
256 struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
258 if (likely(xdst->u.rt.idev))
259 in_dev_put(xdst->u.rt.idev);
260 xfrm_dst_destroy(xdst);
263 static void xfrm4_dst_ifdown(struct dst_entry *dst, struct net_device *dev,
264 int unregister)
266 struct xfrm_dst *xdst;
268 if (!unregister)
269 return;
271 xdst = (struct xfrm_dst *)dst;
272 if (xdst->u.rt.idev->dev == dev) {
273 struct in_device *loopback_idev = in_dev_get(&loopback_dev);
274 BUG_ON(!loopback_idev);
276 do {
277 in_dev_put(xdst->u.rt.idev);
278 xdst->u.rt.idev = loopback_idev;
279 in_dev_hold(loopback_idev);
280 xdst = (struct xfrm_dst *)xdst->u.dst.child;
281 } while (xdst->u.dst.xfrm);
283 __in_dev_put(loopback_idev);
286 xfrm_dst_ifdown(dst, dev);
289 static struct dst_ops xfrm4_dst_ops = {
290 .family = AF_INET,
291 .protocol = __constant_htons(ETH_P_IP),
292 .gc = xfrm4_garbage_collect,
293 .update_pmtu = xfrm4_update_pmtu,
294 .destroy = xfrm4_dst_destroy,
295 .ifdown = xfrm4_dst_ifdown,
296 .gc_thresh = 1024,
297 .entry_size = sizeof(struct xfrm_dst),
300 static struct xfrm_policy_afinfo xfrm4_policy_afinfo = {
301 .family = AF_INET,
302 .lock = RW_LOCK_UNLOCKED,
303 .type_map = &xfrm4_type_map,
304 .dst_ops = &xfrm4_dst_ops,
305 .dst_lookup = xfrm4_dst_lookup,
306 .find_bundle = __xfrm4_find_bundle,
307 .bundle_create = __xfrm4_bundle_create,
308 .decode_session = _decode_session4,
311 static void __init xfrm4_policy_init(void)
313 xfrm_policy_register_afinfo(&xfrm4_policy_afinfo);
316 static void __exit xfrm4_policy_fini(void)
318 xfrm_policy_unregister_afinfo(&xfrm4_policy_afinfo);
321 void __init xfrm4_init(void)
323 xfrm4_state_init();
324 xfrm4_policy_init();