Loading...
1// SPDX-License-Identifier: GPL-2.0
2/*
3 * xfrm6_policy.c: based on xfrm4_policy.c
4 *
5 * Authors:
6 * Mitsuru KANDA @USAGI
7 * Kazunori MIYAZAWA @USAGI
8 * Kunihiro Ishiguro <kunihiro@ipinfusion.com>
9 * IPv6 support
10 * YOSHIFUJI Hideaki
11 * Split up af-specific portion
12 *
13 */
14
15#include <linux/err.h>
16#include <linux/kernel.h>
17#include <linux/netdevice.h>
18#include <net/addrconf.h>
19#include <net/dst.h>
20#include <net/xfrm.h>
21#include <net/ip.h>
22#include <net/ipv6.h>
23#include <net/ip6_route.h>
24#include <net/l3mdev.h>
25
26static struct dst_entry *xfrm6_dst_lookup(const struct xfrm_dst_lookup_params *params)
27{
28 struct flowi6 fl6;
29 struct dst_entry *dst;
30 int err;
31
32 memset(&fl6, 0, sizeof(fl6));
33 fl6.flowi6_l3mdev = l3mdev_master_ifindex_by_index(params->net,
34 params->oif);
35 fl6.flowi6_mark = params->mark;
36 memcpy(&fl6.daddr, params->daddr, sizeof(fl6.daddr));
37 if (params->saddr)
38 memcpy(&fl6.saddr, params->saddr, sizeof(fl6.saddr));
39
40 fl6.flowi4_proto = params->ipproto;
41 fl6.uli = params->uli;
42
43 dst = ip6_route_output(params->net, NULL, &fl6);
44
45 err = dst->error;
46 if (dst->error) {
47 dst_release(dst);
48 dst = ERR_PTR(err);
49 }
50
51 return dst;
52}
53
54static int xfrm6_get_saddr(xfrm_address_t *saddr,
55 const struct xfrm_dst_lookup_params *params)
56{
57 struct dst_entry *dst;
58 struct net_device *dev;
59 struct inet6_dev *idev;
60
61 dst = xfrm6_dst_lookup(params);
62 if (IS_ERR(dst))
63 return -EHOSTUNREACH;
64
65 idev = ip6_dst_idev(dst);
66 if (!idev) {
67 dst_release(dst);
68 return -EHOSTUNREACH;
69 }
70 dev = idev->dev;
71 ipv6_dev_get_saddr(dev_net(dev), dev, ¶ms->daddr->in6, 0,
72 &saddr->in6);
73 dst_release(dst);
74 return 0;
75}
76
77static int xfrm6_fill_dst(struct xfrm_dst *xdst, struct net_device *dev,
78 const struct flowi *fl)
79{
80 struct rt6_info *rt = dst_rt6_info(xdst->route);
81
82 xdst->u.dst.dev = dev;
83 netdev_hold(dev, &xdst->u.dst.dev_tracker, GFP_ATOMIC);
84
85 xdst->u.rt6.rt6i_idev = in6_dev_get(dev);
86 if (!xdst->u.rt6.rt6i_idev) {
87 netdev_put(dev, &xdst->u.dst.dev_tracker);
88 return -ENODEV;
89 }
90
91 /* Sheit... I remember I did this right. Apparently,
92 * it was magically lost, so this code needs audit */
93 xdst->u.rt6.rt6i_flags = rt->rt6i_flags & (RTF_ANYCAST |
94 RTF_LOCAL);
95 xdst->route_cookie = rt6_get_cookie(rt);
96 xdst->u.rt6.rt6i_gateway = rt->rt6i_gateway;
97 xdst->u.rt6.rt6i_dst = rt->rt6i_dst;
98 xdst->u.rt6.rt6i_src = rt->rt6i_src;
99 rt6_uncached_list_add(&xdst->u.rt6);
100
101 return 0;
102}
103
104static void xfrm6_update_pmtu(struct dst_entry *dst, struct sock *sk,
105 struct sk_buff *skb, u32 mtu,
106 bool confirm_neigh)
107{
108 struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
109 struct dst_entry *path = xdst->route;
110
111 path->ops->update_pmtu(path, sk, skb, mtu, confirm_neigh);
112}
113
114static void xfrm6_redirect(struct dst_entry *dst, struct sock *sk,
115 struct sk_buff *skb)
116{
117 struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
118 struct dst_entry *path = xdst->route;
119
120 path->ops->redirect(path, sk, skb);
121}
122
123static void xfrm6_dst_destroy(struct dst_entry *dst)
124{
125 struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
126
127 dst_destroy_metrics_generic(dst);
128 rt6_uncached_list_del(&xdst->u.rt6);
129 if (likely(xdst->u.rt6.rt6i_idev))
130 in6_dev_put(xdst->u.rt6.rt6i_idev);
131 xfrm_dst_destroy(xdst);
132}
133
134static void xfrm6_dst_ifdown(struct dst_entry *dst, struct net_device *dev)
135{
136 struct xfrm_dst *xdst;
137
138 xdst = (struct xfrm_dst *)dst;
139 if (xdst->u.rt6.rt6i_idev->dev == dev) {
140 struct inet6_dev *loopback_idev =
141 in6_dev_get(dev_net(dev)->loopback_dev);
142
143 do {
144 in6_dev_put(xdst->u.rt6.rt6i_idev);
145 xdst->u.rt6.rt6i_idev = loopback_idev;
146 in6_dev_hold(loopback_idev);
147 xdst = (struct xfrm_dst *)xfrm_dst_child(&xdst->u.dst);
148 } while (xdst->u.dst.xfrm);
149
150 __in6_dev_put(loopback_idev);
151 }
152
153 xfrm_dst_ifdown(dst, dev);
154}
155
156static struct dst_ops xfrm6_dst_ops_template = {
157 .family = AF_INET6,
158 .update_pmtu = xfrm6_update_pmtu,
159 .redirect = xfrm6_redirect,
160 .cow_metrics = dst_cow_metrics_generic,
161 .destroy = xfrm6_dst_destroy,
162 .ifdown = xfrm6_dst_ifdown,
163 .local_out = __ip6_local_out,
164 .gc_thresh = 32768,
165};
166
167static const struct xfrm_policy_afinfo xfrm6_policy_afinfo = {
168 .dst_ops = &xfrm6_dst_ops_template,
169 .dst_lookup = xfrm6_dst_lookup,
170 .get_saddr = xfrm6_get_saddr,
171 .fill_dst = xfrm6_fill_dst,
172 .blackhole_route = ip6_blackhole_route,
173};
174
175static int __init xfrm6_policy_init(void)
176{
177 return xfrm_policy_register_afinfo(&xfrm6_policy_afinfo, AF_INET6);
178}
179
180static void xfrm6_policy_fini(void)
181{
182 xfrm_policy_unregister_afinfo(&xfrm6_policy_afinfo);
183}
184
185#ifdef CONFIG_SYSCTL
186static struct ctl_table xfrm6_policy_table[] = {
187 {
188 .procname = "xfrm6_gc_thresh",
189 .data = &init_net.xfrm.xfrm6_dst_ops.gc_thresh,
190 .maxlen = sizeof(int),
191 .mode = 0644,
192 .proc_handler = proc_dointvec,
193 },
194};
195
196static int __net_init xfrm6_net_sysctl_init(struct net *net)
197{
198 struct ctl_table *table;
199 struct ctl_table_header *hdr;
200
201 table = xfrm6_policy_table;
202 if (!net_eq(net, &init_net)) {
203 table = kmemdup(table, sizeof(xfrm6_policy_table), GFP_KERNEL);
204 if (!table)
205 goto err_alloc;
206
207 table[0].data = &net->xfrm.xfrm6_dst_ops.gc_thresh;
208 }
209
210 hdr = register_net_sysctl_sz(net, "net/ipv6", table,
211 ARRAY_SIZE(xfrm6_policy_table));
212 if (!hdr)
213 goto err_reg;
214
215 net->ipv6.sysctl.xfrm6_hdr = hdr;
216 return 0;
217
218err_reg:
219 if (!net_eq(net, &init_net))
220 kfree(table);
221err_alloc:
222 return -ENOMEM;
223}
224
225static void __net_exit xfrm6_net_sysctl_exit(struct net *net)
226{
227 const struct ctl_table *table;
228
229 if (!net->ipv6.sysctl.xfrm6_hdr)
230 return;
231
232 table = net->ipv6.sysctl.xfrm6_hdr->ctl_table_arg;
233 unregister_net_sysctl_table(net->ipv6.sysctl.xfrm6_hdr);
234 if (!net_eq(net, &init_net))
235 kfree(table);
236}
237#else /* CONFIG_SYSCTL */
238static inline int xfrm6_net_sysctl_init(struct net *net)
239{
240 return 0;
241}
242
243static inline void xfrm6_net_sysctl_exit(struct net *net)
244{
245}
246#endif
247
248static int __net_init xfrm6_net_init(struct net *net)
249{
250 int ret;
251
252 memcpy(&net->xfrm.xfrm6_dst_ops, &xfrm6_dst_ops_template,
253 sizeof(xfrm6_dst_ops_template));
254 ret = dst_entries_init(&net->xfrm.xfrm6_dst_ops);
255 if (ret)
256 return ret;
257
258 ret = xfrm6_net_sysctl_init(net);
259 if (ret)
260 dst_entries_destroy(&net->xfrm.xfrm6_dst_ops);
261
262 return ret;
263}
264
265static void __net_exit xfrm6_net_exit(struct net *net)
266{
267 xfrm6_net_sysctl_exit(net);
268 dst_entries_destroy(&net->xfrm.xfrm6_dst_ops);
269}
270
271static struct pernet_operations xfrm6_net_ops = {
272 .init = xfrm6_net_init,
273 .exit = xfrm6_net_exit,
274};
275
276int __init xfrm6_init(void)
277{
278 int ret;
279
280 ret = xfrm6_policy_init();
281 if (ret)
282 goto out;
283 ret = xfrm6_state_init();
284 if (ret)
285 goto out_policy;
286
287 ret = xfrm6_protocol_init();
288 if (ret)
289 goto out_state;
290
291 ret = register_pernet_subsys(&xfrm6_net_ops);
292 if (ret)
293 goto out_protocol;
294
295 ret = xfrm_nat_keepalive_init(AF_INET6);
296 if (ret)
297 goto out_nat_keepalive;
298out:
299 return ret;
300out_nat_keepalive:
301 unregister_pernet_subsys(&xfrm6_net_ops);
302out_protocol:
303 xfrm6_protocol_fini();
304out_state:
305 xfrm6_state_fini();
306out_policy:
307 xfrm6_policy_fini();
308 goto out;
309}
310
311void xfrm6_fini(void)
312{
313 xfrm_nat_keepalive_fini(AF_INET6);
314 unregister_pernet_subsys(&xfrm6_net_ops);
315 xfrm6_protocol_fini();
316 xfrm6_policy_fini();
317 xfrm6_state_fini();
318}
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/err.h>
15#include <linux/kernel.h>
16#include <linux/netdevice.h>
17#include <net/addrconf.h>
18#include <net/dst.h>
19#include <net/xfrm.h>
20#include <net/ip.h>
21#include <net/ipv6.h>
22#include <net/ip6_route.h>
23#if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
24#include <net/mip6.h>
25#endif
26
27static struct xfrm_policy_afinfo xfrm6_policy_afinfo;
28
29static struct dst_entry *xfrm6_dst_lookup(struct net *net, int tos,
30 const xfrm_address_t *saddr,
31 const xfrm_address_t *daddr)
32{
33 struct flowi6 fl6;
34 struct dst_entry *dst;
35 int err;
36
37 memset(&fl6, 0, sizeof(fl6));
38 memcpy(&fl6.daddr, daddr, sizeof(fl6.daddr));
39 if (saddr)
40 memcpy(&fl6.saddr, saddr, sizeof(fl6.saddr));
41
42 dst = ip6_route_output(net, NULL, &fl6);
43
44 err = dst->error;
45 if (dst->error) {
46 dst_release(dst);
47 dst = ERR_PTR(err);
48 }
49
50 return dst;
51}
52
53static int xfrm6_get_saddr(struct net *net,
54 xfrm_address_t *saddr, xfrm_address_t *daddr)
55{
56 struct dst_entry *dst;
57 struct net_device *dev;
58
59 dst = xfrm6_dst_lookup(net, 0, NULL, daddr);
60 if (IS_ERR(dst))
61 return -EHOSTUNREACH;
62
63 dev = ip6_dst_idev(dst)->dev;
64 ipv6_dev_get_saddr(dev_net(dev), dev,
65 (struct in6_addr *)&daddr->a6, 0,
66 (struct in6_addr *)&saddr->a6);
67 dst_release(dst);
68 return 0;
69}
70
71static int xfrm6_get_tos(const struct flowi *fl)
72{
73 return 0;
74}
75
76static int xfrm6_init_path(struct xfrm_dst *path, struct dst_entry *dst,
77 int nfheader_len)
78{
79 if (dst->ops->family == AF_INET6) {
80 struct rt6_info *rt = (struct rt6_info*)dst;
81 if (rt->rt6i_node)
82 path->path_cookie = rt->rt6i_node->fn_sernum;
83 }
84
85 path->u.rt6.rt6i_nfheader_len = nfheader_len;
86
87 return 0;
88}
89
90static int xfrm6_fill_dst(struct xfrm_dst *xdst, struct net_device *dev,
91 const struct flowi *fl)
92{
93 struct rt6_info *rt = (struct rt6_info*)xdst->route;
94
95 xdst->u.dst.dev = dev;
96 dev_hold(dev);
97
98 xdst->u.rt6.rt6i_idev = in6_dev_get(dev);
99 if (!xdst->u.rt6.rt6i_idev)
100 return -ENODEV;
101
102 xdst->u.rt6.rt6i_peer = rt->rt6i_peer;
103 if (rt->rt6i_peer)
104 atomic_inc(&rt->rt6i_peer->refcnt);
105
106 /* Sheit... I remember I did this right. Apparently,
107 * it was magically lost, so this code needs audit */
108 xdst->u.rt6.rt6i_flags = rt->rt6i_flags & (RTF_ANYCAST |
109 RTF_LOCAL);
110 xdst->u.rt6.rt6i_metric = rt->rt6i_metric;
111 xdst->u.rt6.rt6i_node = rt->rt6i_node;
112 if (rt->rt6i_node)
113 xdst->route_cookie = rt->rt6i_node->fn_sernum;
114 xdst->u.rt6.rt6i_gateway = rt->rt6i_gateway;
115 xdst->u.rt6.rt6i_dst = rt->rt6i_dst;
116 xdst->u.rt6.rt6i_src = rt->rt6i_src;
117
118 return 0;
119}
120
121static inline void
122_decode_session6(struct sk_buff *skb, struct flowi *fl, int reverse)
123{
124 struct flowi6 *fl6 = &fl->u.ip6;
125 int onlyproto = 0;
126 u16 offset = skb_network_header_len(skb);
127 const struct ipv6hdr *hdr = ipv6_hdr(skb);
128 struct ipv6_opt_hdr *exthdr;
129 const unsigned char *nh = skb_network_header(skb);
130 u8 nexthdr = nh[IP6CB(skb)->nhoff];
131
132 memset(fl6, 0, sizeof(struct flowi6));
133 fl6->flowi6_mark = skb->mark;
134
135 ipv6_addr_copy(&fl6->daddr, reverse ? &hdr->saddr : &hdr->daddr);
136 ipv6_addr_copy(&fl6->saddr, reverse ? &hdr->daddr : &hdr->saddr);
137
138 while (nh + offset + 1 < skb->data ||
139 pskb_may_pull(skb, nh + offset + 1 - skb->data)) {
140 nh = skb_network_header(skb);
141 exthdr = (struct ipv6_opt_hdr *)(nh + offset);
142
143 switch (nexthdr) {
144 case NEXTHDR_FRAGMENT:
145 onlyproto = 1;
146 case NEXTHDR_ROUTING:
147 case NEXTHDR_HOP:
148 case NEXTHDR_DEST:
149 offset += ipv6_optlen(exthdr);
150 nexthdr = exthdr->nexthdr;
151 exthdr = (struct ipv6_opt_hdr *)(nh + offset);
152 break;
153
154 case IPPROTO_UDP:
155 case IPPROTO_UDPLITE:
156 case IPPROTO_TCP:
157 case IPPROTO_SCTP:
158 case IPPROTO_DCCP:
159 if (!onlyproto && (nh + offset + 4 < skb->data ||
160 pskb_may_pull(skb, nh + offset + 4 - skb->data))) {
161 __be16 *ports = (__be16 *)exthdr;
162
163 fl6->fl6_sport = ports[!!reverse];
164 fl6->fl6_dport = ports[!reverse];
165 }
166 fl6->flowi6_proto = nexthdr;
167 return;
168
169 case IPPROTO_ICMPV6:
170 if (!onlyproto && pskb_may_pull(skb, nh + offset + 2 - skb->data)) {
171 u8 *icmp = (u8 *)exthdr;
172
173 fl6->fl6_icmp_type = icmp[0];
174 fl6->fl6_icmp_code = icmp[1];
175 }
176 fl6->flowi6_proto = nexthdr;
177 return;
178
179#if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
180 case IPPROTO_MH:
181 if (!onlyproto && pskb_may_pull(skb, nh + offset + 3 - skb->data)) {
182 struct ip6_mh *mh;
183 mh = (struct ip6_mh *)exthdr;
184
185 fl6->fl6_mh_type = mh->ip6mh_type;
186 }
187 fl6->flowi6_proto = nexthdr;
188 return;
189#endif
190
191 /* XXX Why are there these headers? */
192 case IPPROTO_AH:
193 case IPPROTO_ESP:
194 case IPPROTO_COMP:
195 default:
196 fl6->fl6_ipsec_spi = 0;
197 fl6->flowi6_proto = nexthdr;
198 return;
199 }
200 }
201}
202
203static inline int xfrm6_garbage_collect(struct dst_ops *ops)
204{
205 struct net *net = container_of(ops, struct net, xfrm.xfrm6_dst_ops);
206
207 xfrm6_policy_afinfo.garbage_collect(net);
208 return dst_entries_get_fast(ops) > ops->gc_thresh * 2;
209}
210
211static void xfrm6_update_pmtu(struct dst_entry *dst, u32 mtu)
212{
213 struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
214 struct dst_entry *path = xdst->route;
215
216 path->ops->update_pmtu(path, mtu);
217}
218
219static void xfrm6_dst_destroy(struct dst_entry *dst)
220{
221 struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
222
223 if (likely(xdst->u.rt6.rt6i_idev))
224 in6_dev_put(xdst->u.rt6.rt6i_idev);
225 dst_destroy_metrics_generic(dst);
226 if (likely(xdst->u.rt6.rt6i_peer))
227 inet_putpeer(xdst->u.rt6.rt6i_peer);
228 xfrm_dst_destroy(xdst);
229}
230
231static void xfrm6_dst_ifdown(struct dst_entry *dst, struct net_device *dev,
232 int unregister)
233{
234 struct xfrm_dst *xdst;
235
236 if (!unregister)
237 return;
238
239 xdst = (struct xfrm_dst *)dst;
240 if (xdst->u.rt6.rt6i_idev->dev == dev) {
241 struct inet6_dev *loopback_idev =
242 in6_dev_get(dev_net(dev)->loopback_dev);
243 BUG_ON(!loopback_idev);
244
245 do {
246 in6_dev_put(xdst->u.rt6.rt6i_idev);
247 xdst->u.rt6.rt6i_idev = loopback_idev;
248 in6_dev_hold(loopback_idev);
249 xdst = (struct xfrm_dst *)xdst->u.dst.child;
250 } while (xdst->u.dst.xfrm);
251
252 __in6_dev_put(loopback_idev);
253 }
254
255 xfrm_dst_ifdown(dst, dev);
256}
257
258static struct dst_ops xfrm6_dst_ops = {
259 .family = AF_INET6,
260 .protocol = cpu_to_be16(ETH_P_IPV6),
261 .gc = xfrm6_garbage_collect,
262 .update_pmtu = xfrm6_update_pmtu,
263 .cow_metrics = dst_cow_metrics_generic,
264 .destroy = xfrm6_dst_destroy,
265 .ifdown = xfrm6_dst_ifdown,
266 .local_out = __ip6_local_out,
267 .gc_thresh = 1024,
268};
269
270static struct xfrm_policy_afinfo xfrm6_policy_afinfo = {
271 .family = AF_INET6,
272 .dst_ops = &xfrm6_dst_ops,
273 .dst_lookup = xfrm6_dst_lookup,
274 .get_saddr = xfrm6_get_saddr,
275 .decode_session = _decode_session6,
276 .get_tos = xfrm6_get_tos,
277 .init_path = xfrm6_init_path,
278 .fill_dst = xfrm6_fill_dst,
279 .blackhole_route = ip6_blackhole_route,
280};
281
282static int __init xfrm6_policy_init(void)
283{
284 return xfrm_policy_register_afinfo(&xfrm6_policy_afinfo);
285}
286
287static void xfrm6_policy_fini(void)
288{
289 xfrm_policy_unregister_afinfo(&xfrm6_policy_afinfo);
290}
291
292#ifdef CONFIG_SYSCTL
293static struct ctl_table xfrm6_policy_table[] = {
294 {
295 .procname = "xfrm6_gc_thresh",
296 .data = &init_net.xfrm.xfrm6_dst_ops.gc_thresh,
297 .maxlen = sizeof(int),
298 .mode = 0644,
299 .proc_handler = proc_dointvec,
300 },
301 { }
302};
303
304static struct ctl_table_header *sysctl_hdr;
305#endif
306
307int __init xfrm6_init(void)
308{
309 int ret;
310 unsigned int gc_thresh;
311
312 /*
313 * We need a good default value for the xfrm6 gc threshold.
314 * In ipv4 we set it to the route hash table size * 8, which
315 * is half the size of the maximaum route cache for ipv4. It
316 * would be good to do the same thing for v6, except the table is
317 * constructed differently here. Here each table for a net namespace
318 * can have FIB_TABLE_HASHSZ entries, so lets go with the same
319 * computation that we used for ipv4 here. Also, lets keep the initial
320 * gc_thresh to a minimum of 1024, since, the ipv6 route cache defaults
321 * to that as a minimum as well
322 */
323 gc_thresh = FIB6_TABLE_HASHSZ * 8;
324 xfrm6_dst_ops.gc_thresh = (gc_thresh < 1024) ? 1024 : gc_thresh;
325 dst_entries_init(&xfrm6_dst_ops);
326
327 ret = xfrm6_policy_init();
328 if (ret) {
329 dst_entries_destroy(&xfrm6_dst_ops);
330 goto out;
331 }
332 ret = xfrm6_state_init();
333 if (ret)
334 goto out_policy;
335
336#ifdef CONFIG_SYSCTL
337 sysctl_hdr = register_net_sysctl_table(&init_net, net_ipv6_ctl_path,
338 xfrm6_policy_table);
339#endif
340out:
341 return ret;
342out_policy:
343 xfrm6_policy_fini();
344 goto out;
345}
346
347void xfrm6_fini(void)
348{
349#ifdef CONFIG_SYSCTL
350 if (sysctl_hdr)
351 unregister_net_sysctl_table(sysctl_hdr);
352#endif
353 //xfrm6_input_fini();
354 xfrm6_policy_fini();
355 xfrm6_state_fini();
356 dst_entries_destroy(&xfrm6_dst_ops);
357}