1 // SPDX-License-Identifier: GPL-2.0-or-later
2 /*
3 * INET An implementation of the TCP/IP protocol suite for the LINUX
4 * operating system. INET is implemented using the BSD Socket
5 * interface as the means of communication with the user level.
6 *
7 * IPv4 Forwarding Information Base: policy rules.
8 *
9 * Authors: Alexey Kuznetsov, <kuznet@ms2.inr.ac.ru>
10 * Thomas Graf <tgraf@suug.ch>
11 *
12 * Fixes:
13 * Rani Assaf : local_rule cannot be deleted
14 * Marc Boucher : routing by fwmark
15 */
16
17 #include <linux/types.h>
18 #include <linux/kernel.h>
19 #include <linux/netdevice.h>
20 #include <linux/netlink.h>
21 #include <linux/inetdevice.h>
22 #include <linux/init.h>
23 #include <linux/list.h>
24 #include <linux/rcupdate.h>
25 #include <linux/export.h>
26 #include <net/ip.h>
27 #include <net/route.h>
28 #include <net/tcp.h>
29 #include <net/ip_fib.h>
30 #include <net/nexthop.h>
31 #include <net/fib_rules.h>
32
33 struct fib4_rule {
34 struct fib_rule common;
35 u8 dst_len;
36 u8 src_len;
37 u8 tos;
38 __be32 src;
39 __be32 srcmask;
40 __be32 dst;
41 __be32 dstmask;
42 #ifdef CONFIG_IP_ROUTE_CLASSID
43 u32 tclassid;
44 #endif
45 };
46
fib4_rule_matchall(const struct fib_rule * rule)47 static bool fib4_rule_matchall(const struct fib_rule *rule)
48 {
49 struct fib4_rule *r = container_of(rule, struct fib4_rule, common);
50
51 if (r->dst_len || r->src_len || r->tos)
52 return false;
53 return fib_rule_matchall(rule);
54 }
55
fib4_rule_default(const struct fib_rule * rule)56 bool fib4_rule_default(const struct fib_rule *rule)
57 {
58 if (!fib4_rule_matchall(rule) || rule->action != FR_ACT_TO_TBL ||
59 rule->l3mdev)
60 return false;
61 if (rule->table != RT_TABLE_LOCAL && rule->table != RT_TABLE_MAIN &&
62 rule->table != RT_TABLE_DEFAULT)
63 return false;
64 return true;
65 }
66 EXPORT_SYMBOL_GPL(fib4_rule_default);
67
fib4_rules_dump(struct net * net,struct notifier_block * nb)68 int fib4_rules_dump(struct net *net, struct notifier_block *nb)
69 {
70 return fib_rules_dump(net, nb, AF_INET);
71 }
72
fib4_rules_seq_read(struct net * net)73 unsigned int fib4_rules_seq_read(struct net *net)
74 {
75 return fib_rules_seq_read(net, AF_INET);
76 }
77
__fib_lookup(struct net * net,struct flowi4 * flp,struct fib_result * res,unsigned int flags)78 int __fib_lookup(struct net *net, struct flowi4 *flp,
79 struct fib_result *res, unsigned int flags)
80 {
81 struct fib_lookup_arg arg = {
82 .result = res,
83 .flags = flags,
84 };
85 int err;
86
87 /* update flow if oif or iif point to device enslaved to l3mdev */
88 l3mdev_update_flow(net, flowi4_to_flowi(flp));
89
90 err = fib_rules_lookup(net->ipv4.rules_ops, flowi4_to_flowi(flp), 0, &arg);
91 #ifdef CONFIG_IP_ROUTE_CLASSID
92 if (arg.rule)
93 res->tclassid = ((struct fib4_rule *)arg.rule)->tclassid;
94 else
95 res->tclassid = 0;
96 #endif
97
98 if (err == -ESRCH)
99 err = -ENETUNREACH;
100
101 return err;
102 }
103 EXPORT_SYMBOL_GPL(__fib_lookup);
104
fib4_rule_action(struct fib_rule * rule,struct flowi * flp,int flags,struct fib_lookup_arg * arg)105 static int fib4_rule_action(struct fib_rule *rule, struct flowi *flp,
106 int flags, struct fib_lookup_arg *arg)
107 {
108 int err = -EAGAIN;
109 struct fib_table *tbl;
110 u32 tb_id;
111
112 switch (rule->action) {
113 case FR_ACT_TO_TBL:
114 break;
115
116 case FR_ACT_UNREACHABLE:
117 return -ENETUNREACH;
118
119 case FR_ACT_PROHIBIT:
120 return -EACCES;
121
122 case FR_ACT_BLACKHOLE:
123 default:
124 return -EINVAL;
125 }
126
127 rcu_read_lock();
128
129 tb_id = fib_rule_get_table(rule, arg);
130 tbl = fib_get_table(rule->fr_net, tb_id);
131 if (tbl)
132 err = fib_table_lookup(tbl, &flp->u.ip4,
133 (struct fib_result *)arg->result,
134 arg->flags);
135
136 rcu_read_unlock();
137 return err;
138 }
139
fib4_rule_suppress(struct fib_rule * rule,struct fib_lookup_arg * arg)140 static bool fib4_rule_suppress(struct fib_rule *rule, struct fib_lookup_arg *arg)
141 {
142 struct fib_result *result = (struct fib_result *) arg->result;
143 struct net_device *dev = NULL;
144
145 if (result->fi) {
146 struct fib_nh_common *nhc = fib_info_nhc(result->fi, 0);
147
148 dev = nhc->nhc_dev;
149 }
150
151 /* do not accept result if the route does
152 * not meet the required prefix length
153 */
154 if (result->prefixlen <= rule->suppress_prefixlen)
155 goto suppress_route;
156
157 /* do not accept result if the route uses a device
158 * belonging to a forbidden interface group
159 */
160 if (rule->suppress_ifgroup != -1 && dev && dev->group == rule->suppress_ifgroup)
161 goto suppress_route;
162
163 return false;
164
165 suppress_route:
166 if (!(arg->flags & FIB_LOOKUP_NOREF))
167 fib_info_put(result->fi);
168 return true;
169 }
170
fib4_rule_match(struct fib_rule * rule,struct flowi * fl,int flags)171 static int fib4_rule_match(struct fib_rule *rule, struct flowi *fl, int flags)
172 {
173 struct fib4_rule *r = (struct fib4_rule *) rule;
174 struct flowi4 *fl4 = &fl->u.ip4;
175 __be32 daddr = fl4->daddr;
176 __be32 saddr = fl4->saddr;
177
178 if (((saddr ^ r->src) & r->srcmask) ||
179 ((daddr ^ r->dst) & r->dstmask))
180 return 0;
181
182 if (r->tos && (r->tos != fl4->flowi4_tos))
183 return 0;
184
185 if (rule->ip_proto && (rule->ip_proto != fl4->flowi4_proto))
186 return 0;
187
188 if (fib_rule_port_range_set(&rule->sport_range) &&
189 !fib_rule_port_inrange(&rule->sport_range, fl4->fl4_sport))
190 return 0;
191
192 if (fib_rule_port_range_set(&rule->dport_range) &&
193 !fib_rule_port_inrange(&rule->dport_range, fl4->fl4_dport))
194 return 0;
195
196 return 1;
197 }
198
fib_empty_table(struct net * net)199 static struct fib_table *fib_empty_table(struct net *net)
200 {
201 u32 id = 1;
202
203 while (1) {
204 if (!fib_get_table(net, id))
205 return fib_new_table(net, id);
206
207 if (id++ == RT_TABLE_MAX)
208 break;
209 }
210 return NULL;
211 }
212
213 static const struct nla_policy fib4_rule_policy[FRA_MAX+1] = {
214 FRA_GENERIC_POLICY,
215 [FRA_FLOW] = { .type = NLA_U32 },
216 };
217
fib4_rule_configure(struct fib_rule * rule,struct sk_buff * skb,struct fib_rule_hdr * frh,struct nlattr ** tb,struct netlink_ext_ack * extack)218 static int fib4_rule_configure(struct fib_rule *rule, struct sk_buff *skb,
219 struct fib_rule_hdr *frh,
220 struct nlattr **tb,
221 struct netlink_ext_ack *extack)
222 {
223 struct net *net = sock_net(skb->sk);
224 int err = -EINVAL;
225 struct fib4_rule *rule4 = (struct fib4_rule *) rule;
226
227 if (frh->tos & ~IPTOS_TOS_MASK) {
228 NL_SET_ERR_MSG(extack, "Invalid tos");
229 goto errout;
230 }
231
232 /* split local/main if they are not already split */
233 err = fib_unmerge(net);
234 if (err)
235 goto errout;
236
237 if (rule->table == RT_TABLE_UNSPEC && !rule->l3mdev) {
238 if (rule->action == FR_ACT_TO_TBL) {
239 struct fib_table *table;
240
241 table = fib_empty_table(net);
242 if (!table) {
243 err = -ENOBUFS;
244 goto errout;
245 }
246
247 rule->table = table->tb_id;
248 }
249 }
250
251 if (frh->src_len)
252 rule4->src = nla_get_in_addr(tb[FRA_SRC]);
253
254 if (frh->dst_len)
255 rule4->dst = nla_get_in_addr(tb[FRA_DST]);
256
257 #ifdef CONFIG_IP_ROUTE_CLASSID
258 if (tb[FRA_FLOW]) {
259 rule4->tclassid = nla_get_u32(tb[FRA_FLOW]);
260 if (rule4->tclassid)
261 net->ipv4.fib_num_tclassid_users++;
262 }
263 #endif
264
265 if (fib_rule_requires_fldissect(rule))
266 net->ipv4.fib_rules_require_fldissect++;
267
268 rule4->src_len = frh->src_len;
269 rule4->srcmask = inet_make_mask(rule4->src_len);
270 rule4->dst_len = frh->dst_len;
271 rule4->dstmask = inet_make_mask(rule4->dst_len);
272 rule4->tos = frh->tos;
273
274 net->ipv4.fib_has_custom_rules = true;
275
276 err = 0;
277 errout:
278 return err;
279 }
280
fib4_rule_delete(struct fib_rule * rule)281 static int fib4_rule_delete(struct fib_rule *rule)
282 {
283 struct net *net = rule->fr_net;
284 int err;
285
286 /* split local/main if they are not already split */
287 err = fib_unmerge(net);
288 if (err)
289 goto errout;
290
291 #ifdef CONFIG_IP_ROUTE_CLASSID
292 if (((struct fib4_rule *)rule)->tclassid)
293 net->ipv4.fib_num_tclassid_users--;
294 #endif
295 net->ipv4.fib_has_custom_rules = true;
296
297 if (net->ipv4.fib_rules_require_fldissect &&
298 fib_rule_requires_fldissect(rule))
299 net->ipv4.fib_rules_require_fldissect--;
300 errout:
301 return err;
302 }
303
fib4_rule_compare(struct fib_rule * rule,struct fib_rule_hdr * frh,struct nlattr ** tb)304 static int fib4_rule_compare(struct fib_rule *rule, struct fib_rule_hdr *frh,
305 struct nlattr **tb)
306 {
307 struct fib4_rule *rule4 = (struct fib4_rule *) rule;
308
309 if (frh->src_len && (rule4->src_len != frh->src_len))
310 return 0;
311
312 if (frh->dst_len && (rule4->dst_len != frh->dst_len))
313 return 0;
314
315 if (frh->tos && (rule4->tos != frh->tos))
316 return 0;
317
318 #ifdef CONFIG_IP_ROUTE_CLASSID
319 if (tb[FRA_FLOW] && (rule4->tclassid != nla_get_u32(tb[FRA_FLOW])))
320 return 0;
321 #endif
322
323 if (frh->src_len && (rule4->src != nla_get_in_addr(tb[FRA_SRC])))
324 return 0;
325
326 if (frh->dst_len && (rule4->dst != nla_get_in_addr(tb[FRA_DST])))
327 return 0;
328
329 return 1;
330 }
331
fib4_rule_fill(struct fib_rule * rule,struct sk_buff * skb,struct fib_rule_hdr * frh)332 static int fib4_rule_fill(struct fib_rule *rule, struct sk_buff *skb,
333 struct fib_rule_hdr *frh)
334 {
335 struct fib4_rule *rule4 = (struct fib4_rule *) rule;
336
337 frh->dst_len = rule4->dst_len;
338 frh->src_len = rule4->src_len;
339 frh->tos = rule4->tos;
340
341 if ((rule4->dst_len &&
342 nla_put_in_addr(skb, FRA_DST, rule4->dst)) ||
343 (rule4->src_len &&
344 nla_put_in_addr(skb, FRA_SRC, rule4->src)))
345 goto nla_put_failure;
346 #ifdef CONFIG_IP_ROUTE_CLASSID
347 if (rule4->tclassid &&
348 nla_put_u32(skb, FRA_FLOW, rule4->tclassid))
349 goto nla_put_failure;
350 #endif
351 return 0;
352
353 nla_put_failure:
354 return -ENOBUFS;
355 }
356
fib4_rule_nlmsg_payload(struct fib_rule * rule)357 static size_t fib4_rule_nlmsg_payload(struct fib_rule *rule)
358 {
359 return nla_total_size(4) /* dst */
360 + nla_total_size(4) /* src */
361 + nla_total_size(4); /* flow */
362 }
363
fib4_rule_flush_cache(struct fib_rules_ops * ops)364 static void fib4_rule_flush_cache(struct fib_rules_ops *ops)
365 {
366 rt_cache_flush(ops->fro_net);
367 }
368
369 static const struct fib_rules_ops __net_initconst fib4_rules_ops_template = {
370 .family = AF_INET,
371 .rule_size = sizeof(struct fib4_rule),
372 .addr_size = sizeof(u32),
373 .action = fib4_rule_action,
374 .suppress = fib4_rule_suppress,
375 .match = fib4_rule_match,
376 .configure = fib4_rule_configure,
377 .delete = fib4_rule_delete,
378 .compare = fib4_rule_compare,
379 .fill = fib4_rule_fill,
380 .nlmsg_payload = fib4_rule_nlmsg_payload,
381 .flush_cache = fib4_rule_flush_cache,
382 .nlgroup = RTNLGRP_IPV4_RULE,
383 .policy = fib4_rule_policy,
384 .owner = THIS_MODULE,
385 };
386
fib_default_rules_init(struct fib_rules_ops * ops)387 static int fib_default_rules_init(struct fib_rules_ops *ops)
388 {
389 int err;
390
391 err = fib_default_rule_add(ops, 0, RT_TABLE_LOCAL, 0);
392 if (err < 0)
393 return err;
394 err = fib_default_rule_add(ops, 0x7FFE, RT_TABLE_MAIN, 0);
395 if (err < 0)
396 return err;
397 err = fib_default_rule_add(ops, 0x7FFF, RT_TABLE_DEFAULT, 0);
398 if (err < 0)
399 return err;
400 return 0;
401 }
402
fib4_rules_init(struct net * net)403 int __net_init fib4_rules_init(struct net *net)
404 {
405 int err;
406 struct fib_rules_ops *ops;
407
408 ops = fib_rules_register(&fib4_rules_ops_template, net);
409 if (IS_ERR(ops))
410 return PTR_ERR(ops);
411
412 err = fib_default_rules_init(ops);
413 if (err < 0)
414 goto fail;
415 net->ipv4.rules_ops = ops;
416 net->ipv4.fib_has_custom_rules = false;
417 net->ipv4.fib_rules_require_fldissect = 0;
418 return 0;
419
420 fail:
421 /* also cleans all rules already added */
422 fib_rules_unregister(ops);
423 return err;
424 }
425
fib4_rules_exit(struct net * net)426 void __net_exit fib4_rules_exit(struct net *net)
427 {
428 fib_rules_unregister(net->ipv4.rules_ops);
429 }
430