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/inet_dscp.h>
27#include <net/ip.h>
28#include <net/route.h>
29#include <net/tcp.h>
30#include <net/ip_fib.h>
31#include <net/nexthop.h>
32#include <net/fib_rules.h>
33#include <linux/indirect_call_wrapper.h>
34
35struct fib4_rule {
36 struct fib_rule common;
37 u8 dst_len;
38 u8 src_len;
39 dscp_t dscp;
40 __be32 src;
41 __be32 srcmask;
42 __be32 dst;
43 __be32 dstmask;
44#ifdef CONFIG_IP_ROUTE_CLASSID
45 u32 tclassid;
46#endif
47};
48
49static bool fib4_rule_matchall(const struct fib_rule *rule)
50{
51 struct fib4_rule *r = container_of(rule, struct fib4_rule, common);
52
53 if (r->dst_len || r->src_len || r->dscp)
54 return false;
55 return fib_rule_matchall(rule);
56}
57
58bool fib4_rule_default(const struct fib_rule *rule)
59{
60 if (!fib4_rule_matchall(rule) || rule->action != FR_ACT_TO_TBL ||
61 rule->l3mdev)
62 return false;
63 if (rule->table != RT_TABLE_LOCAL && rule->table != RT_TABLE_MAIN &&
64 rule->table != RT_TABLE_DEFAULT)
65 return false;
66 return true;
67}
68EXPORT_SYMBOL_GPL(fib4_rule_default);
69
70int fib4_rules_dump(struct net *net, struct notifier_block *nb,
71 struct netlink_ext_ack *extack)
72{
73 return fib_rules_dump(net, nb, AF_INET, extack);
74}
75
76unsigned int fib4_rules_seq_read(struct net *net)
77{
78 return fib_rules_seq_read(net, AF_INET);
79}
80
81int __fib_lookup(struct net *net, struct flowi4 *flp,
82 struct fib_result *res, unsigned int flags)
83{
84 struct fib_lookup_arg arg = {
85 .result = res,
86 .flags = flags,
87 };
88 int err;
89
90 /* update flow if oif or iif point to device enslaved to l3mdev */
91 l3mdev_update_flow(net, fl: flowi4_to_flowi(fl4: flp));
92
93 err = fib_rules_lookup(net->ipv4.rules_ops, flowi4_to_flowi(fl4: flp), flags: 0, &arg);
94#ifdef CONFIG_IP_ROUTE_CLASSID
95 if (arg.rule)
96 res->tclassid = ((struct fib4_rule *)arg.rule)->tclassid;
97 else
98 res->tclassid = 0;
99#endif
100
101 if (err == -ESRCH)
102 err = -ENETUNREACH;
103
104 return err;
105}
106EXPORT_SYMBOL_GPL(__fib_lookup);
107
108INDIRECT_CALLABLE_SCOPE int fib4_rule_action(struct fib_rule *rule,
109 struct flowi *flp, int flags,
110 struct fib_lookup_arg *arg)
111{
112 int err = -EAGAIN;
113 struct fib_table *tbl;
114 u32 tb_id;
115
116 switch (rule->action) {
117 case FR_ACT_TO_TBL:
118 break;
119
120 case FR_ACT_UNREACHABLE:
121 return -ENETUNREACH;
122
123 case FR_ACT_PROHIBIT:
124 return -EACCES;
125
126 case FR_ACT_BLACKHOLE:
127 default:
128 return -EINVAL;
129 }
130
131 rcu_read_lock();
132
133 tb_id = fib_rule_get_table(rule, arg);
134 tbl = fib_get_table(net: rule->fr_net, id: tb_id);
135 if (tbl)
136 err = fib_table_lookup(tb: tbl, flp: &flp->u.ip4,
137 res: (struct fib_result *)arg->result,
138 fib_flags: arg->flags);
139
140 rcu_read_unlock();
141 return err;
142}
143
144INDIRECT_CALLABLE_SCOPE bool fib4_rule_suppress(struct fib_rule *rule,
145 int flags,
146 struct fib_lookup_arg *arg)
147{
148 struct fib_result *result = arg->result;
149 struct net_device *dev = NULL;
150
151 if (result->fi) {
152 struct fib_nh_common *nhc = fib_info_nhc(fi: result->fi, nhsel: 0);
153
154 dev = nhc->nhc_dev;
155 }
156
157 /* do not accept result if the route does
158 * not meet the required prefix length
159 */
160 if (result->prefixlen <= rule->suppress_prefixlen)
161 goto suppress_route;
162
163 /* do not accept result if the route uses a device
164 * belonging to a forbidden interface group
165 */
166 if (rule->suppress_ifgroup != -1 && dev && dev->group == rule->suppress_ifgroup)
167 goto suppress_route;
168
169 return false;
170
171suppress_route:
172 if (!(arg->flags & FIB_LOOKUP_NOREF))
173 fib_info_put(fi: result->fi);
174 return true;
175}
176
177INDIRECT_CALLABLE_SCOPE int fib4_rule_match(struct fib_rule *rule,
178 struct flowi *fl, int flags)
179{
180 struct fib4_rule *r = (struct fib4_rule *) rule;
181 struct flowi4 *fl4 = &fl->u.ip4;
182 __be32 daddr = fl4->daddr;
183 __be32 saddr = fl4->saddr;
184
185 if (((saddr ^ r->src) & r->srcmask) ||
186 ((daddr ^ r->dst) & r->dstmask))
187 return 0;
188
189 if (r->dscp && r->dscp != inet_dsfield_to_dscp(dsfield: fl4->flowi4_tos))
190 return 0;
191
192 if (rule->ip_proto && (rule->ip_proto != fl4->flowi4_proto))
193 return 0;
194
195 if (fib_rule_port_range_set(range: &rule->sport_range) &&
196 !fib_rule_port_inrange(a: &rule->sport_range, port: fl4->fl4_sport))
197 return 0;
198
199 if (fib_rule_port_range_set(range: &rule->dport_range) &&
200 !fib_rule_port_inrange(a: &rule->dport_range, port: fl4->fl4_dport))
201 return 0;
202
203 return 1;
204}
205
206static struct fib_table *fib_empty_table(struct net *net)
207{
208 u32 id = 1;
209
210 while (1) {
211 if (!fib_get_table(net, id))
212 return fib_new_table(net, id);
213
214 if (id++ == RT_TABLE_MAX)
215 break;
216 }
217 return NULL;
218}
219
220static int fib4_rule_configure(struct fib_rule *rule, struct sk_buff *skb,
221 struct fib_rule_hdr *frh,
222 struct nlattr **tb,
223 struct netlink_ext_ack *extack)
224{
225 struct net *net = sock_net(sk: skb->sk);
226 int err = -EINVAL;
227 struct fib4_rule *rule4 = (struct fib4_rule *) rule;
228
229 if (!inet_validate_dscp(val: frh->tos)) {
230 NL_SET_ERR_MSG(extack,
231 "Invalid dsfield (tos): ECN bits must be 0");
232 goto errout;
233 }
234 /* IPv4 currently doesn't handle high order DSCP bits correctly */
235 if (frh->tos & ~IPTOS_TOS_MASK) {
236 NL_SET_ERR_MSG(extack, "Invalid tos");
237 goto errout;
238 }
239 rule4->dscp = inet_dsfield_to_dscp(dsfield: frh->tos);
240
241 /* split local/main if they are not already split */
242 err = fib_unmerge(net);
243 if (err)
244 goto errout;
245
246 if (rule->table == RT_TABLE_UNSPEC && !rule->l3mdev) {
247 if (rule->action == FR_ACT_TO_TBL) {
248 struct fib_table *table;
249
250 table = fib_empty_table(net);
251 if (!table) {
252 err = -ENOBUFS;
253 goto errout;
254 }
255
256 rule->table = table->tb_id;
257 }
258 }
259
260 if (frh->src_len)
261 rule4->src = nla_get_in_addr(nla: tb[FRA_SRC]);
262
263 if (frh->dst_len)
264 rule4->dst = nla_get_in_addr(nla: tb[FRA_DST]);
265
266#ifdef CONFIG_IP_ROUTE_CLASSID
267 if (tb[FRA_FLOW]) {
268 rule4->tclassid = nla_get_u32(nla: tb[FRA_FLOW]);
269 if (rule4->tclassid)
270 atomic_inc(v: &net->ipv4.fib_num_tclassid_users);
271 }
272#endif
273
274 if (fib_rule_requires_fldissect(rule))
275 net->ipv4.fib_rules_require_fldissect++;
276
277 rule4->src_len = frh->src_len;
278 rule4->srcmask = inet_make_mask(logmask: rule4->src_len);
279 rule4->dst_len = frh->dst_len;
280 rule4->dstmask = inet_make_mask(logmask: rule4->dst_len);
281
282 net->ipv4.fib_has_custom_rules = true;
283
284 err = 0;
285errout:
286 return err;
287}
288
289static int fib4_rule_delete(struct fib_rule *rule)
290{
291 struct net *net = rule->fr_net;
292 int err;
293
294 /* split local/main if they are not already split */
295 err = fib_unmerge(net);
296 if (err)
297 goto errout;
298
299#ifdef CONFIG_IP_ROUTE_CLASSID
300 if (((struct fib4_rule *)rule)->tclassid)
301 atomic_dec(v: &net->ipv4.fib_num_tclassid_users);
302#endif
303 net->ipv4.fib_has_custom_rules = true;
304
305 if (net->ipv4.fib_rules_require_fldissect &&
306 fib_rule_requires_fldissect(rule))
307 net->ipv4.fib_rules_require_fldissect--;
308errout:
309 return err;
310}
311
312static int fib4_rule_compare(struct fib_rule *rule, struct fib_rule_hdr *frh,
313 struct nlattr **tb)
314{
315 struct fib4_rule *rule4 = (struct fib4_rule *) rule;
316
317 if (frh->src_len && (rule4->src_len != frh->src_len))
318 return 0;
319
320 if (frh->dst_len && (rule4->dst_len != frh->dst_len))
321 return 0;
322
323 if (frh->tos && inet_dscp_to_dsfield(dscp: rule4->dscp) != frh->tos)
324 return 0;
325
326#ifdef CONFIG_IP_ROUTE_CLASSID
327 if (tb[FRA_FLOW] && (rule4->tclassid != nla_get_u32(nla: tb[FRA_FLOW])))
328 return 0;
329#endif
330
331 if (frh->src_len && (rule4->src != nla_get_in_addr(nla: tb[FRA_SRC])))
332 return 0;
333
334 if (frh->dst_len && (rule4->dst != nla_get_in_addr(nla: tb[FRA_DST])))
335 return 0;
336
337 return 1;
338}
339
340static int fib4_rule_fill(struct fib_rule *rule, struct sk_buff *skb,
341 struct fib_rule_hdr *frh)
342{
343 struct fib4_rule *rule4 = (struct fib4_rule *) rule;
344
345 frh->dst_len = rule4->dst_len;
346 frh->src_len = rule4->src_len;
347 frh->tos = inet_dscp_to_dsfield(dscp: rule4->dscp);
348
349 if ((rule4->dst_len &&
350 nla_put_in_addr(skb, attrtype: FRA_DST, addr: rule4->dst)) ||
351 (rule4->src_len &&
352 nla_put_in_addr(skb, attrtype: FRA_SRC, addr: rule4->src)))
353 goto nla_put_failure;
354#ifdef CONFIG_IP_ROUTE_CLASSID
355 if (rule4->tclassid &&
356 nla_put_u32(skb, attrtype: FRA_FLOW, value: rule4->tclassid))
357 goto nla_put_failure;
358#endif
359 return 0;
360
361nla_put_failure:
362 return -ENOBUFS;
363}
364
365static size_t fib4_rule_nlmsg_payload(struct fib_rule *rule)
366{
367 return nla_total_size(payload: 4) /* dst */
368 + nla_total_size(payload: 4) /* src */
369 + nla_total_size(payload: 4); /* flow */
370}
371
372static void fib4_rule_flush_cache(struct fib_rules_ops *ops)
373{
374 rt_cache_flush(net: ops->fro_net);
375}
376
377static const struct fib_rules_ops __net_initconst fib4_rules_ops_template = {
378 .family = AF_INET,
379 .rule_size = sizeof(struct fib4_rule),
380 .addr_size = sizeof(u32),
381 .action = fib4_rule_action,
382 .suppress = fib4_rule_suppress,
383 .match = fib4_rule_match,
384 .configure = fib4_rule_configure,
385 .delete = fib4_rule_delete,
386 .compare = fib4_rule_compare,
387 .fill = fib4_rule_fill,
388 .nlmsg_payload = fib4_rule_nlmsg_payload,
389 .flush_cache = fib4_rule_flush_cache,
390 .nlgroup = RTNLGRP_IPV4_RULE,
391 .owner = THIS_MODULE,
392};
393
394static int fib_default_rules_init(struct fib_rules_ops *ops)
395{
396 int err;
397
398 err = fib_default_rule_add(ops, pref: 0, table: RT_TABLE_LOCAL, flags: 0);
399 if (err < 0)
400 return err;
401 err = fib_default_rule_add(ops, pref: 0x7FFE, table: RT_TABLE_MAIN, flags: 0);
402 if (err < 0)
403 return err;
404 err = fib_default_rule_add(ops, pref: 0x7FFF, table: RT_TABLE_DEFAULT, flags: 0);
405 if (err < 0)
406 return err;
407 return 0;
408}
409
410int __net_init fib4_rules_init(struct net *net)
411{
412 int err;
413 struct fib_rules_ops *ops;
414
415 ops = fib_rules_register(&fib4_rules_ops_template, net);
416 if (IS_ERR(ptr: ops))
417 return PTR_ERR(ptr: ops);
418
419 err = fib_default_rules_init(ops);
420 if (err < 0)
421 goto fail;
422 net->ipv4.rules_ops = ops;
423 net->ipv4.fib_has_custom_rules = false;
424 net->ipv4.fib_rules_require_fldissect = 0;
425 return 0;
426
427fail:
428 /* also cleans all rules already added */
429 fib_rules_unregister(ops);
430 return err;
431}
432
433void __net_exit fib4_rules_exit(struct net *net)
434{
435 fib_rules_unregister(net->ipv4.rules_ops);
436}
437

source code of linux/net/ipv4/fib_rules.c