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
19static struct xfrm_policy_afinfo xfrm4_policy_afinfo;
20
21static struct dst_entry *__xfrm4_dst_lookup(struct net *net, struct flowi4 *fl4,
22 int tos,
23 const xfrm_address_t *saddr,
24 const xfrm_address_t *daddr)
25{
26 struct rtable *rt;
27
28 memset(fl4, 0, sizeof(*fl4));
29 fl4->daddr = daddr->a4;
30 fl4->flowi4_tos = tos;
31 if (saddr)
32 fl4->saddr = saddr->a4;
33
34 rt = __ip_route_output_key(net, fl4);
35 if (!IS_ERR(rt))
36 return &rt->dst;
37
38 return ERR_CAST(rt);
39}
40
41static struct dst_entry *xfrm4_dst_lookup(struct net *net, int tos,
42 const xfrm_address_t *saddr,
43 const xfrm_address_t *daddr)
44{
45 struct flowi4 fl4;
46
47 return __xfrm4_dst_lookup(net, &fl4, tos, saddr, daddr);
48}
49
50static int xfrm4_get_saddr(struct net *net,
51 xfrm_address_t *saddr, xfrm_address_t *daddr)
52{
53 struct dst_entry *dst;
54 struct flowi4 fl4;
55
56 dst = __xfrm4_dst_lookup(net, &fl4, 0, NULL, daddr);
57 if (IS_ERR(dst))
58 return -EHOSTUNREACH;
59
60 saddr->a4 = fl4.saddr;
61 dst_release(dst);
62 return 0;
63}
64
65static int xfrm4_get_tos(const struct flowi *fl)
66{
67 return IPTOS_RT_MASK & fl->u.ip4.flowi4_tos; /* Strip ECN bits */
68}
69
70static int xfrm4_init_path(struct xfrm_dst *path, struct dst_entry *dst,
71 int nfheader_len)
72{
73 return 0;
74}
75
76static int xfrm4_fill_dst(struct xfrm_dst *xdst, struct net_device *dev,
77 const struct flowi *fl)
78{
79 struct rtable *rt = (struct rtable *)xdst->route;
80 const struct flowi4 *fl4 = &fl->u.ip4;
81
82 xdst->u.rt.rt_key_dst = fl4->daddr;
83 xdst->u.rt.rt_key_src = fl4->saddr;
84 xdst->u.rt.rt_key_tos = fl4->flowi4_tos;
85 xdst->u.rt.rt_route_iif = fl4->flowi4_iif;
86 xdst->u.rt.rt_iif = fl4->flowi4_iif;
87 xdst->u.rt.rt_oif = fl4->flowi4_oif;
88 xdst->u.rt.rt_mark = fl4->flowi4_mark;
89
90 xdst->u.dst.dev = dev;
91 dev_hold(dev);
92
93 xdst->u.rt.peer = rt->peer;
94 if (rt->peer)
95 atomic_inc(&rt->peer->refcnt);
96
97 /* Sheit... I remember I did this right. Apparently,
98 * it was magically lost, so this code needs audit */
99 xdst->u.rt.rt_flags = rt->rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST |
100 RTCF_LOCAL);
101 xdst->u.rt.rt_type = rt->rt_type;
102 xdst->u.rt.rt_src = rt->rt_src;
103 xdst->u.rt.rt_dst = rt->rt_dst;
104 xdst->u.rt.rt_gateway = rt->rt_gateway;
105 xdst->u.rt.rt_spec_dst = rt->rt_spec_dst;
106
107 return 0;
108}
109
110static void
111_decode_session4(struct sk_buff *skb, struct flowi *fl, int reverse)
112{
113 const struct iphdr *iph = ip_hdr(skb);
114 u8 *xprth = skb_network_header(skb) + iph->ihl * 4;
115 struct flowi4 *fl4 = &fl->u.ip4;
116
117 memset(fl4, 0, sizeof(struct flowi4));
118 fl4->flowi4_mark = skb->mark;
119
120 if (!ip_is_fragment(iph)) {
121 switch (iph->protocol) {
122 case IPPROTO_UDP:
123 case IPPROTO_UDPLITE:
124 case IPPROTO_TCP:
125 case IPPROTO_SCTP:
126 case IPPROTO_DCCP:
127 if (xprth + 4 < skb->data ||
128 pskb_may_pull(skb, xprth + 4 - skb->data)) {
129 __be16 *ports = (__be16 *)xprth;
130
131 fl4->fl4_sport = ports[!!reverse];
132 fl4->fl4_dport = ports[!reverse];
133 }
134 break;
135
136 case IPPROTO_ICMP:
137 if (pskb_may_pull(skb, xprth + 2 - skb->data)) {
138 u8 *icmp = xprth;
139
140 fl4->fl4_icmp_type = icmp[0];
141 fl4->fl4_icmp_code = icmp[1];
142 }
143 break;
144
145 case IPPROTO_ESP:
146 if (pskb_may_pull(skb, xprth + 4 - skb->data)) {
147 __be32 *ehdr = (__be32 *)xprth;
148
149 fl4->fl4_ipsec_spi = ehdr[0];
150 }
151 break;
152
153 case IPPROTO_AH:
154 if (pskb_may_pull(skb, xprth + 8 - skb->data)) {
155 __be32 *ah_hdr = (__be32 *)xprth;
156
157 fl4->fl4_ipsec_spi = ah_hdr[1];
158 }
159 break;
160
161 case IPPROTO_COMP:
162 if (pskb_may_pull(skb, xprth + 4 - skb->data)) {
163 __be16 *ipcomp_hdr = (__be16 *)xprth;
164
165 fl4->fl4_ipsec_spi = htonl(ntohs(ipcomp_hdr[1]));
166 }
167 break;
168
169 case IPPROTO_GRE:
170 if (pskb_may_pull(skb, xprth + 12 - skb->data)) {
171 __be16 *greflags = (__be16 *)xprth;
172 __be32 *gre_hdr = (__be32 *)xprth;
173
174 if (greflags[0] & GRE_KEY) {
175 if (greflags[0] & GRE_CSUM)
176 gre_hdr++;
177 fl4->fl4_gre_key = gre_hdr[1];
178 }
179 }
180 break;
181
182 default:
183 fl4->fl4_ipsec_spi = 0;
184 break;
185 }
186 }
187 fl4->flowi4_proto = iph->protocol;
188 fl4->daddr = reverse ? iph->saddr : iph->daddr;
189 fl4->saddr = reverse ? iph->daddr : iph->saddr;
190 fl4->flowi4_tos = iph->tos;
191}
192
193static inline int xfrm4_garbage_collect(struct dst_ops *ops)
194{
195 struct net *net = container_of(ops, struct net, xfrm.xfrm4_dst_ops);
196
197 xfrm4_policy_afinfo.garbage_collect(net);
198 return (dst_entries_get_slow(ops) > ops->gc_thresh * 2);
199}
200
201static void xfrm4_update_pmtu(struct dst_entry *dst, u32 mtu)
202{
203 struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
204 struct dst_entry *path = xdst->route;
205
206 path->ops->update_pmtu(path, mtu);
207}
208
209static void xfrm4_dst_destroy(struct dst_entry *dst)
210{
211 struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
212
213 dst_destroy_metrics_generic(dst);
214
215 if (likely(xdst->u.rt.peer))
216 inet_putpeer(xdst->u.rt.peer);
217
218 xfrm_dst_destroy(xdst);
219}
220
221static void xfrm4_dst_ifdown(struct dst_entry *dst, struct net_device *dev,
222 int unregister)
223{
224 if (!unregister)
225 return;
226
227 xfrm_dst_ifdown(dst, dev);
228}
229
230static struct dst_ops xfrm4_dst_ops = {
231 .family = AF_INET,
232 .protocol = cpu_to_be16(ETH_P_IP),
233 .gc = xfrm4_garbage_collect,
234 .update_pmtu = xfrm4_update_pmtu,
235 .cow_metrics = dst_cow_metrics_generic,
236 .destroy = xfrm4_dst_destroy,
237 .ifdown = xfrm4_dst_ifdown,
238 .local_out = __ip_local_out,
239 .gc_thresh = 1024,
240};
241
242static struct xfrm_policy_afinfo xfrm4_policy_afinfo = {
243 .family = AF_INET,
244 .dst_ops = &xfrm4_dst_ops,
245 .dst_lookup = xfrm4_dst_lookup,
246 .get_saddr = xfrm4_get_saddr,
247 .decode_session = _decode_session4,
248 .get_tos = xfrm4_get_tos,
249 .init_path = xfrm4_init_path,
250 .fill_dst = xfrm4_fill_dst,
251 .blackhole_route = ipv4_blackhole_route,
252};
253
254#ifdef CONFIG_SYSCTL
255static struct ctl_table xfrm4_policy_table[] = {
256 {
257 .procname = "xfrm4_gc_thresh",
258 .data = &init_net.xfrm.xfrm4_dst_ops.gc_thresh,
259 .maxlen = sizeof(int),
260 .mode = 0644,
261 .proc_handler = proc_dointvec,
262 },
263 { }
264};
265
266static struct ctl_table_header *sysctl_hdr;
267#endif
268
269static void __init xfrm4_policy_init(void)
270{
271 xfrm_policy_register_afinfo(&xfrm4_policy_afinfo);
272}
273
274static void __exit xfrm4_policy_fini(void)
275{
276#ifdef CONFIG_SYSCTL
277 if (sysctl_hdr)
278 unregister_net_sysctl_table(sysctl_hdr);
279#endif
280 xfrm_policy_unregister_afinfo(&xfrm4_policy_afinfo);
281}
282
283void __init xfrm4_init(int rt_max_size)
284{
285 /*
286 * Select a default value for the gc_thresh based on the main route
287 * table hash size. It seems to me the worst case scenario is when
288 * we have ipsec operating in transport mode, in which we create a
289 * dst_entry per socket. The xfrm gc algorithm starts trying to remove
290 * entries at gc_thresh, and prevents new allocations as 2*gc_thresh
291 * so lets set an initial xfrm gc_thresh value at the rt_max_size/2.
292 * That will let us store an ipsec connection per route table entry,
293 * and start cleaning when were 1/2 full
294 */
295 xfrm4_dst_ops.gc_thresh = rt_max_size/2;
296 dst_entries_init(&xfrm4_dst_ops);
297
298 xfrm4_state_init();
299 xfrm4_policy_init();
300#ifdef CONFIG_SYSCTL
301 sysctl_hdr = register_net_sysctl(&init_net, "net/ipv4",
302 xfrm4_policy_table);
303#endif
304}
305