Loading...
1// SPDX-License-Identifier: GPL-2.0
2/*
3 * xfrm4_policy.c
4 *
5 * Changes:
6 * Kazunori MIYAZAWA @USAGI
7 * YOSHIFUJI Hideaki @USAGI
8 * Split up af-specific portion
9 *
10 */
11
12#include <linux/err.h>
13#include <linux/kernel.h>
14#include <linux/inetdevice.h>
15#include <net/dst.h>
16#include <net/xfrm.h>
17#include <net/inet_dscp.h>
18#include <net/ip.h>
19#include <net/l3mdev.h>
20
21static struct dst_entry *__xfrm4_dst_lookup(struct flowi4 *fl4,
22 const struct xfrm_dst_lookup_params *params)
23{
24 struct rtable *rt;
25
26 memset(fl4, 0, sizeof(*fl4));
27 fl4->daddr = params->daddr->a4;
28 fl4->flowi4_tos = inet_dscp_to_dsfield(params->dscp);
29 fl4->flowi4_l3mdev = l3mdev_master_ifindex_by_index(params->net,
30 params->oif);
31 fl4->flowi4_mark = params->mark;
32 if (params->saddr)
33 fl4->saddr = params->saddr->a4;
34 fl4->flowi4_proto = params->ipproto;
35 fl4->uli = params->uli;
36
37 rt = __ip_route_output_key(params->net, fl4);
38 if (!IS_ERR(rt))
39 return &rt->dst;
40
41 return ERR_CAST(rt);
42}
43
44static struct dst_entry *xfrm4_dst_lookup(const struct xfrm_dst_lookup_params *params)
45{
46 struct flowi4 fl4;
47
48 return __xfrm4_dst_lookup(&fl4, params);
49}
50
51static int xfrm4_get_saddr(xfrm_address_t *saddr,
52 const struct xfrm_dst_lookup_params *params)
53{
54 struct dst_entry *dst;
55 struct flowi4 fl4;
56
57 dst = __xfrm4_dst_lookup(&fl4, params);
58 if (IS_ERR(dst))
59 return -EHOSTUNREACH;
60
61 saddr->a4 = fl4.saddr;
62 dst_release(dst);
63 return 0;
64}
65
66static int xfrm4_fill_dst(struct xfrm_dst *xdst, struct net_device *dev,
67 const struct flowi *fl)
68{
69 struct rtable *rt = dst_rtable(xdst->route);
70 const struct flowi4 *fl4 = &fl->u.ip4;
71
72 xdst->u.rt.rt_iif = fl4->flowi4_iif;
73
74 xdst->u.dst.dev = dev;
75 netdev_hold(dev, &xdst->u.dst.dev_tracker, GFP_ATOMIC);
76
77 /* Sheit... I remember I did this right. Apparently,
78 * it was magically lost, so this code needs audit */
79 xdst->u.rt.rt_is_input = rt->rt_is_input;
80 xdst->u.rt.rt_flags = rt->rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST |
81 RTCF_LOCAL);
82 xdst->u.rt.rt_type = rt->rt_type;
83 xdst->u.rt.rt_uses_gateway = rt->rt_uses_gateway;
84 xdst->u.rt.rt_gw_family = rt->rt_gw_family;
85 if (rt->rt_gw_family == AF_INET)
86 xdst->u.rt.rt_gw4 = rt->rt_gw4;
87 else if (rt->rt_gw_family == AF_INET6)
88 xdst->u.rt.rt_gw6 = rt->rt_gw6;
89 xdst->u.rt.rt_pmtu = rt->rt_pmtu;
90 xdst->u.rt.rt_mtu_locked = rt->rt_mtu_locked;
91 rt_add_uncached_list(&xdst->u.rt);
92
93 return 0;
94}
95
96static void xfrm4_update_pmtu(struct dst_entry *dst, struct sock *sk,
97 struct sk_buff *skb, u32 mtu,
98 bool confirm_neigh)
99{
100 struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
101 struct dst_entry *path = xdst->route;
102
103 path->ops->update_pmtu(path, sk, skb, mtu, confirm_neigh);
104}
105
106static void xfrm4_redirect(struct dst_entry *dst, struct sock *sk,
107 struct sk_buff *skb)
108{
109 struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
110 struct dst_entry *path = xdst->route;
111
112 path->ops->redirect(path, sk, skb);
113}
114
115static void xfrm4_dst_destroy(struct dst_entry *dst)
116{
117 struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
118
119 dst_destroy_metrics_generic(dst);
120 rt_del_uncached_list(&xdst->u.rt);
121 xfrm_dst_destroy(xdst);
122}
123
124static struct dst_ops xfrm4_dst_ops_template = {
125 .family = AF_INET,
126 .update_pmtu = xfrm4_update_pmtu,
127 .redirect = xfrm4_redirect,
128 .cow_metrics = dst_cow_metrics_generic,
129 .destroy = xfrm4_dst_destroy,
130 .ifdown = xfrm_dst_ifdown,
131 .local_out = __ip_local_out,
132 .gc_thresh = 32768,
133};
134
135static const struct xfrm_policy_afinfo xfrm4_policy_afinfo = {
136 .dst_ops = &xfrm4_dst_ops_template,
137 .dst_lookup = xfrm4_dst_lookup,
138 .get_saddr = xfrm4_get_saddr,
139 .fill_dst = xfrm4_fill_dst,
140 .blackhole_route = ipv4_blackhole_route,
141};
142
143#ifdef CONFIG_SYSCTL
144static struct ctl_table xfrm4_policy_table[] = {
145 {
146 .procname = "xfrm4_gc_thresh",
147 .data = &init_net.xfrm.xfrm4_dst_ops.gc_thresh,
148 .maxlen = sizeof(int),
149 .mode = 0644,
150 .proc_handler = proc_dointvec,
151 },
152};
153
154static __net_init int xfrm4_net_sysctl_init(struct net *net)
155{
156 struct ctl_table *table;
157 struct ctl_table_header *hdr;
158
159 table = xfrm4_policy_table;
160 if (!net_eq(net, &init_net)) {
161 table = kmemdup(table, sizeof(xfrm4_policy_table), GFP_KERNEL);
162 if (!table)
163 goto err_alloc;
164
165 table[0].data = &net->xfrm.xfrm4_dst_ops.gc_thresh;
166 }
167
168 hdr = register_net_sysctl_sz(net, "net/ipv4", table,
169 ARRAY_SIZE(xfrm4_policy_table));
170 if (!hdr)
171 goto err_reg;
172
173 net->ipv4.xfrm4_hdr = hdr;
174 return 0;
175
176err_reg:
177 if (!net_eq(net, &init_net))
178 kfree(table);
179err_alloc:
180 return -ENOMEM;
181}
182
183static __net_exit void xfrm4_net_sysctl_exit(struct net *net)
184{
185 const struct ctl_table *table;
186
187 if (!net->ipv4.xfrm4_hdr)
188 return;
189
190 table = net->ipv4.xfrm4_hdr->ctl_table_arg;
191 unregister_net_sysctl_table(net->ipv4.xfrm4_hdr);
192 if (!net_eq(net, &init_net))
193 kfree(table);
194}
195#else /* CONFIG_SYSCTL */
196static inline int xfrm4_net_sysctl_init(struct net *net)
197{
198 return 0;
199}
200
201static inline void xfrm4_net_sysctl_exit(struct net *net)
202{
203}
204#endif
205
206static int __net_init xfrm4_net_init(struct net *net)
207{
208 int ret;
209
210 memcpy(&net->xfrm.xfrm4_dst_ops, &xfrm4_dst_ops_template,
211 sizeof(xfrm4_dst_ops_template));
212 ret = dst_entries_init(&net->xfrm.xfrm4_dst_ops);
213 if (ret)
214 return ret;
215
216 ret = xfrm4_net_sysctl_init(net);
217 if (ret)
218 dst_entries_destroy(&net->xfrm.xfrm4_dst_ops);
219
220 return ret;
221}
222
223static void __net_exit xfrm4_net_exit(struct net *net)
224{
225 xfrm4_net_sysctl_exit(net);
226 dst_entries_destroy(&net->xfrm.xfrm4_dst_ops);
227}
228
229static struct pernet_operations __net_initdata xfrm4_net_ops = {
230 .init = xfrm4_net_init,
231 .exit = xfrm4_net_exit,
232};
233
234static void __init xfrm4_policy_init(void)
235{
236 xfrm_policy_register_afinfo(&xfrm4_policy_afinfo, AF_INET);
237}
238
239void __init xfrm4_init(void)
240{
241 xfrm4_state_init();
242 xfrm4_policy_init();
243 xfrm4_protocol_init();
244 register_pernet_subsys(&xfrm4_net_ops);
245}
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/err.h>
12#include <linux/kernel.h>
13#include <linux/inetdevice.h>
14#include <linux/if_tunnel.h>
15#include <net/dst.h>
16#include <net/xfrm.h>
17#include <net/ip.h>
18#include <net/l3mdev.h>
19
20static struct xfrm_policy_afinfo xfrm4_policy_afinfo;
21
22static struct dst_entry *__xfrm4_dst_lookup(struct net *net, struct flowi4 *fl4,
23 int tos, int oif,
24 const xfrm_address_t *saddr,
25 const xfrm_address_t *daddr)
26{
27 struct rtable *rt;
28
29 memset(fl4, 0, sizeof(*fl4));
30 fl4->daddr = daddr->a4;
31 fl4->flowi4_tos = tos;
32 fl4->flowi4_oif = oif;
33 if (saddr)
34 fl4->saddr = saddr->a4;
35
36 fl4->flowi4_flags = FLOWI_FLAG_SKIP_NH_OIF;
37
38 rt = __ip_route_output_key(net, fl4);
39 if (!IS_ERR(rt))
40 return &rt->dst;
41
42 return ERR_CAST(rt);
43}
44
45static struct dst_entry *xfrm4_dst_lookup(struct net *net, int tos, int oif,
46 const xfrm_address_t *saddr,
47 const xfrm_address_t *daddr)
48{
49 struct flowi4 fl4;
50
51 return __xfrm4_dst_lookup(net, &fl4, tos, oif, saddr, daddr);
52}
53
54static int xfrm4_get_saddr(struct net *net, int oif,
55 xfrm_address_t *saddr, xfrm_address_t *daddr)
56{
57 struct dst_entry *dst;
58 struct flowi4 fl4;
59
60 dst = __xfrm4_dst_lookup(net, &fl4, 0, oif, NULL, daddr);
61 if (IS_ERR(dst))
62 return -EHOSTUNREACH;
63
64 saddr->a4 = fl4.saddr;
65 dst_release(dst);
66 return 0;
67}
68
69static int xfrm4_get_tos(const struct flowi *fl)
70{
71 return IPTOS_RT_MASK & fl->u.ip4.flowi4_tos; /* Strip ECN bits */
72}
73
74static int xfrm4_init_path(struct xfrm_dst *path, struct dst_entry *dst,
75 int nfheader_len)
76{
77 return 0;
78}
79
80static int xfrm4_fill_dst(struct xfrm_dst *xdst, struct net_device *dev,
81 const struct flowi *fl)
82{
83 struct rtable *rt = (struct rtable *)xdst->route;
84 const struct flowi4 *fl4 = &fl->u.ip4;
85
86 xdst->u.rt.rt_iif = fl4->flowi4_iif;
87
88 xdst->u.dst.dev = dev;
89 dev_hold(dev);
90
91 /* Sheit... I remember I did this right. Apparently,
92 * it was magically lost, so this code needs audit */
93 xdst->u.rt.rt_is_input = rt->rt_is_input;
94 xdst->u.rt.rt_flags = rt->rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST |
95 RTCF_LOCAL);
96 xdst->u.rt.rt_type = rt->rt_type;
97 xdst->u.rt.rt_gateway = rt->rt_gateway;
98 xdst->u.rt.rt_uses_gateway = rt->rt_uses_gateway;
99 xdst->u.rt.rt_pmtu = rt->rt_pmtu;
100 xdst->u.rt.rt_table_id = rt->rt_table_id;
101 INIT_LIST_HEAD(&xdst->u.rt.rt_uncached);
102
103 return 0;
104}
105
106static void
107_decode_session4(struct sk_buff *skb, struct flowi *fl, int reverse)
108{
109 const struct iphdr *iph = ip_hdr(skb);
110 u8 *xprth = skb_network_header(skb) + iph->ihl * 4;
111 struct flowi4 *fl4 = &fl->u.ip4;
112 int oif = 0;
113
114 if (skb_dst(skb))
115 oif = l3mdev_fib_oif(skb_dst(skb)->dev);
116
117 memset(fl4, 0, sizeof(struct flowi4));
118 fl4->flowi4_mark = skb->mark;
119 fl4->flowi4_oif = reverse ? skb->skb_iif : oif;
120
121 if (!ip_is_fragment(iph)) {
122 switch (iph->protocol) {
123 case IPPROTO_UDP:
124 case IPPROTO_UDPLITE:
125 case IPPROTO_TCP:
126 case IPPROTO_SCTP:
127 case IPPROTO_DCCP:
128 if (xprth + 4 < skb->data ||
129 pskb_may_pull(skb, xprth + 4 - skb->data)) {
130 __be16 *ports;
131
132 xprth = skb_network_header(skb) + iph->ihl * 4;
133 ports = (__be16 *)xprth;
134
135 fl4->fl4_sport = ports[!!reverse];
136 fl4->fl4_dport = ports[!reverse];
137 }
138 break;
139
140 case IPPROTO_ICMP:
141 if (xprth + 2 < skb->data ||
142 pskb_may_pull(skb, xprth + 2 - skb->data)) {
143 u8 *icmp;
144
145 xprth = skb_network_header(skb) + iph->ihl * 4;
146 icmp = xprth;
147
148 fl4->fl4_icmp_type = icmp[0];
149 fl4->fl4_icmp_code = icmp[1];
150 }
151 break;
152
153 case IPPROTO_ESP:
154 if (xprth + 4 < skb->data ||
155 pskb_may_pull(skb, xprth + 4 - skb->data)) {
156 __be32 *ehdr;
157
158 xprth = skb_network_header(skb) + iph->ihl * 4;
159 ehdr = (__be32 *)xprth;
160
161 fl4->fl4_ipsec_spi = ehdr[0];
162 }
163 break;
164
165 case IPPROTO_AH:
166 if (xprth + 8 < skb->data ||
167 pskb_may_pull(skb, xprth + 8 - skb->data)) {
168 __be32 *ah_hdr;
169
170 xprth = skb_network_header(skb) + iph->ihl * 4;
171 ah_hdr = (__be32 *)xprth;
172
173 fl4->fl4_ipsec_spi = ah_hdr[1];
174 }
175 break;
176
177 case IPPROTO_COMP:
178 if (xprth + 4 < skb->data ||
179 pskb_may_pull(skb, xprth + 4 - skb->data)) {
180 __be16 *ipcomp_hdr;
181
182 xprth = skb_network_header(skb) + iph->ihl * 4;
183 ipcomp_hdr = (__be16 *)xprth;
184
185 fl4->fl4_ipsec_spi = htonl(ntohs(ipcomp_hdr[1]));
186 }
187 break;
188
189 case IPPROTO_GRE:
190 if (xprth + 12 < skb->data ||
191 pskb_may_pull(skb, xprth + 12 - skb->data)) {
192 __be16 *greflags;
193 __be32 *gre_hdr;
194
195 xprth = skb_network_header(skb) + iph->ihl * 4;
196 greflags = (__be16 *)xprth;
197 gre_hdr = (__be32 *)xprth;
198
199 if (greflags[0] & GRE_KEY) {
200 if (greflags[0] & GRE_CSUM)
201 gre_hdr++;
202 fl4->fl4_gre_key = gre_hdr[1];
203 }
204 }
205 break;
206
207 default:
208 fl4->fl4_ipsec_spi = 0;
209 break;
210 }
211 }
212 fl4->flowi4_proto = iph->protocol;
213 fl4->daddr = reverse ? iph->saddr : iph->daddr;
214 fl4->saddr = reverse ? iph->daddr : iph->saddr;
215 fl4->flowi4_tos = iph->tos;
216}
217
218static inline int xfrm4_garbage_collect(struct dst_ops *ops)
219{
220 struct net *net = container_of(ops, struct net, xfrm.xfrm4_dst_ops);
221
222 xfrm4_policy_afinfo.garbage_collect(net);
223 return (dst_entries_get_slow(ops) > ops->gc_thresh * 2);
224}
225
226static void xfrm4_update_pmtu(struct dst_entry *dst, struct sock *sk,
227 struct sk_buff *skb, u32 mtu)
228{
229 struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
230 struct dst_entry *path = xdst->route;
231
232 path->ops->update_pmtu(path, sk, skb, mtu);
233}
234
235static void xfrm4_redirect(struct dst_entry *dst, struct sock *sk,
236 struct sk_buff *skb)
237{
238 struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
239 struct dst_entry *path = xdst->route;
240
241 path->ops->redirect(path, sk, skb);
242}
243
244static void xfrm4_dst_destroy(struct dst_entry *dst)
245{
246 struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
247
248 dst_destroy_metrics_generic(dst);
249
250 xfrm_dst_destroy(xdst);
251}
252
253static void xfrm4_dst_ifdown(struct dst_entry *dst, struct net_device *dev,
254 int unregister)
255{
256 if (!unregister)
257 return;
258
259 xfrm_dst_ifdown(dst, dev);
260}
261
262static struct dst_ops xfrm4_dst_ops_template = {
263 .family = AF_INET,
264 .gc = xfrm4_garbage_collect,
265 .update_pmtu = xfrm4_update_pmtu,
266 .redirect = xfrm4_redirect,
267 .cow_metrics = dst_cow_metrics_generic,
268 .destroy = xfrm4_dst_destroy,
269 .ifdown = xfrm4_dst_ifdown,
270 .local_out = __ip_local_out,
271 .gc_thresh = INT_MAX,
272};
273
274static struct xfrm_policy_afinfo xfrm4_policy_afinfo = {
275 .family = AF_INET,
276 .dst_ops = &xfrm4_dst_ops_template,
277 .dst_lookup = xfrm4_dst_lookup,
278 .get_saddr = xfrm4_get_saddr,
279 .decode_session = _decode_session4,
280 .get_tos = xfrm4_get_tos,
281 .init_path = xfrm4_init_path,
282 .fill_dst = xfrm4_fill_dst,
283 .blackhole_route = ipv4_blackhole_route,
284};
285
286#ifdef CONFIG_SYSCTL
287static struct ctl_table xfrm4_policy_table[] = {
288 {
289 .procname = "xfrm4_gc_thresh",
290 .data = &init_net.xfrm.xfrm4_dst_ops.gc_thresh,
291 .maxlen = sizeof(int),
292 .mode = 0644,
293 .proc_handler = proc_dointvec,
294 },
295 { }
296};
297
298static int __net_init xfrm4_net_sysctl_init(struct net *net)
299{
300 struct ctl_table *table;
301 struct ctl_table_header *hdr;
302
303 table = xfrm4_policy_table;
304 if (!net_eq(net, &init_net)) {
305 table = kmemdup(table, sizeof(xfrm4_policy_table), GFP_KERNEL);
306 if (!table)
307 goto err_alloc;
308
309 table[0].data = &net->xfrm.xfrm4_dst_ops.gc_thresh;
310 }
311
312 hdr = register_net_sysctl(net, "net/ipv4", table);
313 if (!hdr)
314 goto err_reg;
315
316 net->ipv4.xfrm4_hdr = hdr;
317 return 0;
318
319err_reg:
320 if (!net_eq(net, &init_net))
321 kfree(table);
322err_alloc:
323 return -ENOMEM;
324}
325
326static void __net_exit xfrm4_net_sysctl_exit(struct net *net)
327{
328 struct ctl_table *table;
329
330 if (!net->ipv4.xfrm4_hdr)
331 return;
332
333 table = net->ipv4.xfrm4_hdr->ctl_table_arg;
334 unregister_net_sysctl_table(net->ipv4.xfrm4_hdr);
335 if (!net_eq(net, &init_net))
336 kfree(table);
337}
338#else /* CONFIG_SYSCTL */
339static int inline xfrm4_net_sysctl_init(struct net *net)
340{
341 return 0;
342}
343
344static void inline xfrm4_net_sysctl_exit(struct net *net)
345{
346}
347#endif
348
349static int __net_init xfrm4_net_init(struct net *net)
350{
351 int ret;
352
353 memcpy(&net->xfrm.xfrm4_dst_ops, &xfrm4_dst_ops_template,
354 sizeof(xfrm4_dst_ops_template));
355 ret = dst_entries_init(&net->xfrm.xfrm4_dst_ops);
356 if (ret)
357 return ret;
358
359 ret = xfrm4_net_sysctl_init(net);
360 if (ret)
361 dst_entries_destroy(&net->xfrm.xfrm4_dst_ops);
362
363 return ret;
364}
365
366static void __net_exit xfrm4_net_exit(struct net *net)
367{
368 xfrm4_net_sysctl_exit(net);
369 dst_entries_destroy(&net->xfrm.xfrm4_dst_ops);
370}
371
372static struct pernet_operations __net_initdata xfrm4_net_ops = {
373 .init = xfrm4_net_init,
374 .exit = xfrm4_net_exit,
375};
376
377static void __init xfrm4_policy_init(void)
378{
379 xfrm_policy_register_afinfo(&xfrm4_policy_afinfo);
380}
381
382void __init xfrm4_init(void)
383{
384 xfrm4_state_init();
385 xfrm4_policy_init();
386 xfrm4_protocol_init();
387 register_pernet_subsys(&xfrm4_net_ops);
388}
389