]>
Commit | Line | Data |
---|---|---|
e905a9ed | 1 | /* |
1da177e4 LT |
2 | * xfrm4_policy.c |
3 | * | |
4 | * Changes: | |
5 | * Kazunori MIYAZAWA @USAGI | |
6 | * YOSHIFUJI Hideaki @USAGI | |
7 | * Split up af-specific portion | |
e905a9ed | 8 | * |
1da177e4 LT |
9 | */ |
10 | ||
66cdb3ca HX |
11 | #include <linux/err.h> |
12 | #include <linux/kernel.h> | |
aabc9761 | 13 | #include <linux/inetdevice.h> |
cc9ff19d | 14 | #include <linux/if_tunnel.h> |
45ff5a3f | 15 | #include <net/dst.h> |
1da177e4 LT |
16 | #include <net/xfrm.h> |
17 | #include <net/ip.h> | |
18 | ||
1da177e4 LT |
19 | static struct xfrm_policy_afinfo xfrm4_policy_afinfo; |
20 | ||
c5b3cf46 | 21 | static struct dst_entry *xfrm4_dst_lookup(struct net *net, int tos, |
5e6b930f DM |
22 | const xfrm_address_t *saddr, |
23 | const xfrm_address_t *daddr) | |
1da177e4 | 24 | { |
9d6ec938 DM |
25 | struct flowi4 fl4 = { |
26 | .daddr = daddr->a4, | |
27 | .flowi4_tos = tos, | |
a1e59abf | 28 | }; |
66cdb3ca | 29 | struct rtable *rt; |
a1e59abf | 30 | |
66cdb3ca | 31 | if (saddr) |
9d6ec938 | 32 | fl4.saddr = saddr->a4; |
66cdb3ca | 33 | |
9d6ec938 | 34 | rt = __ip_route_output_key(net, &fl4); |
b23dd4fe DM |
35 | if (!IS_ERR(rt)) |
36 | return &rt->dst; | |
37 | ||
38 | return ERR_CAST(rt); | |
66cdb3ca HX |
39 | } |
40 | ||
fbda33b2 AD |
41 | static int xfrm4_get_saddr(struct net *net, |
42 | xfrm_address_t *saddr, xfrm_address_t *daddr) | |
66cdb3ca HX |
43 | { |
44 | struct dst_entry *dst; | |
45 | struct rtable *rt; | |
46 | ||
fbda33b2 | 47 | dst = xfrm4_dst_lookup(net, 0, NULL, daddr); |
66cdb3ca HX |
48 | if (IS_ERR(dst)) |
49 | return -EHOSTUNREACH; | |
50 | ||
51 | rt = (struct rtable *)dst; | |
52 | saddr->a4 = rt->rt_src; | |
53 | dst_release(dst); | |
54 | return 0; | |
a1e59abf PM |
55 | } |
56 | ||
05d84025 | 57 | static int xfrm4_get_tos(const struct flowi *fl) |
1da177e4 | 58 | { |
7e1dc7b6 | 59 | return IPTOS_RT_MASK & fl->u.ip4.flowi4_tos; /* Strip ECN bits */ |
25ee3286 | 60 | } |
1da177e4 | 61 | |
a1b05140 MN |
62 | static int xfrm4_init_path(struct xfrm_dst *path, struct dst_entry *dst, |
63 | int nfheader_len) | |
64 | { | |
65 | return 0; | |
66 | } | |
67 | ||
87c1e12b | 68 | static int xfrm4_fill_dst(struct xfrm_dst *xdst, struct net_device *dev, |
0c7b3eef | 69 | const struct flowi *fl) |
25ee3286 HX |
70 | { |
71 | struct rtable *rt = (struct rtable *)xdst->route; | |
7e1dc7b6 | 72 | const struct flowi4 *fl4 = &fl->u.ip4; |
1da177e4 | 73 | |
7e1dc7b6 DM |
74 | rt->rt_key_dst = fl4->daddr; |
75 | rt->rt_key_src = fl4->saddr; | |
475949d8 | 76 | rt->rt_key_tos = fl4->flowi4_tos; |
1b86a58f | 77 | rt->rt_route_iif = fl4->flowi4_iif; |
7e1dc7b6 DM |
78 | rt->rt_iif = fl4->flowi4_iif; |
79 | rt->rt_oif = fl4->flowi4_oif; | |
80 | rt->rt_mark = fl4->flowi4_mark; | |
1da177e4 | 81 | |
25ee3286 HX |
82 | xdst->u.dst.dev = dev; |
83 | dev_hold(dev); | |
43372262 | 84 | |
25ee3286 HX |
85 | xdst->u.rt.peer = rt->peer; |
86 | if (rt->peer) | |
87 | atomic_inc(&rt->peer->refcnt); | |
66cdb3ca | 88 | |
25ee3286 HX |
89 | /* Sheit... I remember I did this right. Apparently, |
90 | * it was magically lost, so this code needs audit */ | |
91 | xdst->u.rt.rt_flags = rt->rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST | | |
92 | RTCF_LOCAL); | |
93 | xdst->u.rt.rt_type = rt->rt_type; | |
94 | xdst->u.rt.rt_src = rt->rt_src; | |
95 | xdst->u.rt.rt_dst = rt->rt_dst; | |
96 | xdst->u.rt.rt_gateway = rt->rt_gateway; | |
97 | xdst->u.rt.rt_spec_dst = rt->rt_spec_dst; | |
1da177e4 | 98 | |
1da177e4 | 99 | return 0; |
1da177e4 LT |
100 | } |
101 | ||
102 | static void | |
d5422efe | 103 | _decode_session4(struct sk_buff *skb, struct flowi *fl, int reverse) |
1da177e4 | 104 | { |
b71d1d42 | 105 | const struct iphdr *iph = ip_hdr(skb); |
d56f90a7 | 106 | u8 *xprth = skb_network_header(skb) + iph->ihl * 4; |
7e1dc7b6 | 107 | struct flowi4 *fl4 = &fl->u.ip4; |
1da177e4 | 108 | |
7e1dc7b6 DM |
109 | memset(fl4, 0, sizeof(struct flowi4)); |
110 | fl4->flowi4_mark = skb->mark; | |
44b451f1 | 111 | |
1da177e4 LT |
112 | if (!(iph->frag_off & htons(IP_MF | IP_OFFSET))) { |
113 | switch (iph->protocol) { | |
114 | case IPPROTO_UDP: | |
ba4e58ec | 115 | case IPPROTO_UDPLITE: |
1da177e4 LT |
116 | case IPPROTO_TCP: |
117 | case IPPROTO_SCTP: | |
9e999993 | 118 | case IPPROTO_DCCP: |
c615c9f3 WY |
119 | if (xprth + 4 < skb->data || |
120 | pskb_may_pull(skb, xprth + 4 - skb->data)) { | |
8c689a6e | 121 | __be16 *ports = (__be16 *)xprth; |
1da177e4 | 122 | |
9cce96df DM |
123 | fl4->fl4_sport = ports[!!reverse]; |
124 | fl4->fl4_dport = ports[!reverse]; | |
1da177e4 LT |
125 | } |
126 | break; | |
127 | ||
128 | case IPPROTO_ICMP: | |
129 | if (pskb_may_pull(skb, xprth + 2 - skb->data)) { | |
130 | u8 *icmp = xprth; | |
131 | ||
9cce96df DM |
132 | fl4->fl4_icmp_type = icmp[0]; |
133 | fl4->fl4_icmp_code = icmp[1]; | |
1da177e4 LT |
134 | } |
135 | break; | |
136 | ||
137 | case IPPROTO_ESP: | |
138 | if (pskb_may_pull(skb, xprth + 4 - skb->data)) { | |
4324a174 | 139 | __be32 *ehdr = (__be32 *)xprth; |
1da177e4 | 140 | |
9cce96df | 141 | fl4->fl4_ipsec_spi = ehdr[0]; |
1da177e4 LT |
142 | } |
143 | break; | |
144 | ||
145 | case IPPROTO_AH: | |
146 | if (pskb_may_pull(skb, xprth + 8 - skb->data)) { | |
4324a174 | 147 | __be32 *ah_hdr = (__be32*)xprth; |
1da177e4 | 148 | |
9cce96df | 149 | fl4->fl4_ipsec_spi = ah_hdr[1]; |
1da177e4 LT |
150 | } |
151 | break; | |
152 | ||
153 | case IPPROTO_COMP: | |
154 | if (pskb_may_pull(skb, xprth + 4 - skb->data)) { | |
4324a174 | 155 | __be16 *ipcomp_hdr = (__be16 *)xprth; |
1da177e4 | 156 | |
9cce96df | 157 | fl4->fl4_ipsec_spi = htonl(ntohs(ipcomp_hdr[1])); |
1da177e4 LT |
158 | } |
159 | break; | |
cc9ff19d TT |
160 | |
161 | case IPPROTO_GRE: | |
162 | if (pskb_may_pull(skb, xprth + 12 - skb->data)) { | |
163 | __be16 *greflags = (__be16 *)xprth; | |
164 | __be32 *gre_hdr = (__be32 *)xprth; | |
165 | ||
166 | if (greflags[0] & GRE_KEY) { | |
167 | if (greflags[0] & GRE_CSUM) | |
168 | gre_hdr++; | |
9cce96df | 169 | fl4->fl4_gre_key = gre_hdr[1]; |
cc9ff19d TT |
170 | } |
171 | } | |
172 | break; | |
173 | ||
1da177e4 | 174 | default: |
9cce96df | 175 | fl4->fl4_ipsec_spi = 0; |
1da177e4 | 176 | break; |
3ff50b79 | 177 | } |
1da177e4 | 178 | } |
7e1dc7b6 DM |
179 | fl4->flowi4_proto = iph->protocol; |
180 | fl4->daddr = reverse ? iph->saddr : iph->daddr; | |
181 | fl4->saddr = reverse ? iph->daddr : iph->saddr; | |
182 | fl4->flowi4_tos = iph->tos; | |
1da177e4 LT |
183 | } |
184 | ||
569d3645 | 185 | static inline int xfrm4_garbage_collect(struct dst_ops *ops) |
1da177e4 | 186 | { |
d7c7544c AD |
187 | struct net *net = container_of(ops, struct net, xfrm.xfrm4_dst_ops); |
188 | ||
189 | xfrm4_policy_afinfo.garbage_collect(net); | |
fc66f95c | 190 | return (dst_entries_get_slow(ops) > ops->gc_thresh * 2); |
1da177e4 LT |
191 | } |
192 | ||
193 | static void xfrm4_update_pmtu(struct dst_entry *dst, u32 mtu) | |
194 | { | |
195 | struct xfrm_dst *xdst = (struct xfrm_dst *)dst; | |
196 | struct dst_entry *path = xdst->route; | |
197 | ||
198 | path->ops->update_pmtu(path, mtu); | |
199 | } | |
200 | ||
aabc9761 HX |
201 | static void xfrm4_dst_destroy(struct dst_entry *dst) |
202 | { | |
203 | struct xfrm_dst *xdst = (struct xfrm_dst *)dst; | |
204 | ||
62fa8a84 DM |
205 | dst_destroy_metrics_generic(dst); |
206 | ||
ed3e37dd | 207 | if (likely(xdst->u.rt.peer)) |
26db1677 | 208 | inet_putpeer(xdst->u.rt.peer); |
62fa8a84 | 209 | |
aabc9761 HX |
210 | xfrm_dst_destroy(xdst); |
211 | } | |
212 | ||
213 | static void xfrm4_dst_ifdown(struct dst_entry *dst, struct net_device *dev, | |
214 | int unregister) | |
215 | { | |
aabc9761 HX |
216 | if (!unregister) |
217 | return; | |
218 | ||
aabc9761 HX |
219 | xfrm_dst_ifdown(dst, dev); |
220 | } | |
221 | ||
1da177e4 LT |
222 | static struct dst_ops xfrm4_dst_ops = { |
223 | .family = AF_INET, | |
09640e63 | 224 | .protocol = cpu_to_be16(ETH_P_IP), |
1da177e4 LT |
225 | .gc = xfrm4_garbage_collect, |
226 | .update_pmtu = xfrm4_update_pmtu, | |
62fa8a84 | 227 | .cow_metrics = dst_cow_metrics_generic, |
aabc9761 HX |
228 | .destroy = xfrm4_dst_destroy, |
229 | .ifdown = xfrm4_dst_ifdown, | |
862b82c6 | 230 | .local_out = __ip_local_out, |
1da177e4 | 231 | .gc_thresh = 1024, |
1da177e4 LT |
232 | }; |
233 | ||
234 | static struct xfrm_policy_afinfo xfrm4_policy_afinfo = { | |
235 | .family = AF_INET, | |
1da177e4 LT |
236 | .dst_ops = &xfrm4_dst_ops, |
237 | .dst_lookup = xfrm4_dst_lookup, | |
a1e59abf | 238 | .get_saddr = xfrm4_get_saddr, |
1da177e4 | 239 | .decode_session = _decode_session4, |
25ee3286 | 240 | .get_tos = xfrm4_get_tos, |
a1b05140 | 241 | .init_path = xfrm4_init_path, |
25ee3286 | 242 | .fill_dst = xfrm4_fill_dst, |
2774c131 | 243 | .blackhole_route = ipv4_blackhole_route, |
1da177e4 LT |
244 | }; |
245 | ||
f816700a | 246 | #ifdef CONFIG_SYSCTL |
a44a4a00 NH |
247 | static struct ctl_table xfrm4_policy_table[] = { |
248 | { | |
a44a4a00 | 249 | .procname = "xfrm4_gc_thresh", |
d7c7544c | 250 | .data = &init_net.xfrm.xfrm4_dst_ops.gc_thresh, |
a44a4a00 NH |
251 | .maxlen = sizeof(int), |
252 | .mode = 0644, | |
253 | .proc_handler = proc_dointvec, | |
254 | }, | |
255 | { } | |
256 | }; | |
257 | ||
258 | static struct ctl_table_header *sysctl_hdr; | |
f816700a | 259 | #endif |
a44a4a00 | 260 | |
1da177e4 LT |
261 | static void __init xfrm4_policy_init(void) |
262 | { | |
263 | xfrm_policy_register_afinfo(&xfrm4_policy_afinfo); | |
264 | } | |
265 | ||
266 | static void __exit xfrm4_policy_fini(void) | |
267 | { | |
f816700a | 268 | #ifdef CONFIG_SYSCTL |
a44a4a00 NH |
269 | if (sysctl_hdr) |
270 | unregister_net_sysctl_table(sysctl_hdr); | |
f816700a | 271 | #endif |
1da177e4 LT |
272 | xfrm_policy_unregister_afinfo(&xfrm4_policy_afinfo); |
273 | } | |
274 | ||
a33bc5c1 | 275 | void __init xfrm4_init(int rt_max_size) |
1da177e4 | 276 | { |
a33bc5c1 NH |
277 | /* |
278 | * Select a default value for the gc_thresh based on the main route | |
279 | * table hash size. It seems to me the worst case scenario is when | |
280 | * we have ipsec operating in transport mode, in which we create a | |
281 | * dst_entry per socket. The xfrm gc algorithm starts trying to remove | |
282 | * entries at gc_thresh, and prevents new allocations as 2*gc_thresh | |
283 | * so lets set an initial xfrm gc_thresh value at the rt_max_size/2. | |
284 | * That will let us store an ipsec connection per route table entry, | |
285 | * and start cleaning when were 1/2 full | |
286 | */ | |
287 | xfrm4_dst_ops.gc_thresh = rt_max_size/2; | |
fc66f95c | 288 | dst_entries_init(&xfrm4_dst_ops); |
d7c7544c AD |
289 | |
290 | xfrm4_state_init(); | |
291 | xfrm4_policy_init(); | |
f816700a | 292 | #ifdef CONFIG_SYSCTL |
a44a4a00 NH |
293 | sysctl_hdr = register_net_sysctl_table(&init_net, net_ipv4_ctl_path, |
294 | xfrm4_policy_table); | |
f816700a | 295 | #endif |
1da177e4 LT |
296 | } |
297 |