xref: /linux/net/ipv4/fib_rules.c (revision b7019ac550eb3916f34d79db583e9b7ea2524afa)
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/fib_rules.h>
31 
32 struct fib4_rule {
33 	struct fib_rule		common;
34 	u8			dst_len;
35 	u8			src_len;
36 	u8			tos;
37 	__be32			src;
38 	__be32			srcmask;
39 	__be32			dst;
40 	__be32			dstmask;
41 #ifdef CONFIG_IP_ROUTE_CLASSID
42 	u32			tclassid;
43 #endif
44 };
45 
46 static bool fib4_rule_matchall(const struct fib_rule *rule)
47 {
48 	struct fib4_rule *r = container_of(rule, struct fib4_rule, common);
49 
50 	if (r->dst_len || r->src_len || r->tos)
51 		return false;
52 	return fib_rule_matchall(rule);
53 }
54 
55 bool fib4_rule_default(const struct fib_rule *rule)
56 {
57 	if (!fib4_rule_matchall(rule) || rule->action != FR_ACT_TO_TBL ||
58 	    rule->l3mdev)
59 		return false;
60 	if (rule->table != RT_TABLE_LOCAL && rule->table != RT_TABLE_MAIN &&
61 	    rule->table != RT_TABLE_DEFAULT)
62 		return false;
63 	return true;
64 }
65 EXPORT_SYMBOL_GPL(fib4_rule_default);
66 
67 int fib4_rules_dump(struct net *net, struct notifier_block *nb)
68 {
69 	return fib_rules_dump(net, nb, AF_INET);
70 }
71 
72 unsigned int fib4_rules_seq_read(struct net *net)
73 {
74 	return fib_rules_seq_read(net, AF_INET);
75 }
76 
77 int __fib_lookup(struct net *net, struct flowi4 *flp,
78 		 struct fib_result *res, unsigned int flags)
79 {
80 	struct fib_lookup_arg arg = {
81 		.result = res,
82 		.flags = flags,
83 	};
84 	int err;
85 
86 	/* update flow if oif or iif point to device enslaved to l3mdev */
87 	l3mdev_update_flow(net, flowi4_to_flowi(flp));
88 
89 	err = fib_rules_lookup(net->ipv4.rules_ops, flowi4_to_flowi(flp), 0, &arg);
90 #ifdef CONFIG_IP_ROUTE_CLASSID
91 	if (arg.rule)
92 		res->tclassid = ((struct fib4_rule *)arg.rule)->tclassid;
93 	else
94 		res->tclassid = 0;
95 #endif
96 
97 	if (err == -ESRCH)
98 		err = -ENETUNREACH;
99 
100 	return err;
101 }
102 EXPORT_SYMBOL_GPL(__fib_lookup);
103 
104 static int fib4_rule_action(struct fib_rule *rule, struct flowi *flp,
105 			    int flags, struct fib_lookup_arg *arg)
106 {
107 	int err = -EAGAIN;
108 	struct fib_table *tbl;
109 	u32 tb_id;
110 
111 	switch (rule->action) {
112 	case FR_ACT_TO_TBL:
113 		break;
114 
115 	case FR_ACT_UNREACHABLE:
116 		return -ENETUNREACH;
117 
118 	case FR_ACT_PROHIBIT:
119 		return -EACCES;
120 
121 	case FR_ACT_BLACKHOLE:
122 	default:
123 		return -EINVAL;
124 	}
125 
126 	rcu_read_lock();
127 
128 	tb_id = fib_rule_get_table(rule, arg);
129 	tbl = fib_get_table(rule->fr_net, tb_id);
130 	if (tbl)
131 		err = fib_table_lookup(tbl, &flp->u.ip4,
132 				       (struct fib_result *)arg->result,
133 				       arg->flags);
134 
135 	rcu_read_unlock();
136 	return err;
137 }
138 
139 static bool fib4_rule_suppress(struct fib_rule *rule, struct fib_lookup_arg *arg)
140 {
141 	struct fib_result *result = (struct fib_result *) arg->result;
142 	struct net_device *dev = NULL;
143 
144 	if (result->fi)
145 		dev = result->fi->fib_dev;
146 
147 	/* do not accept result if the route does
148 	 * not meet the required prefix length
149 	 */
150 	if (result->prefixlen <= rule->suppress_prefixlen)
151 		goto suppress_route;
152 
153 	/* do not accept result if the route uses a device
154 	 * belonging to a forbidden interface group
155 	 */
156 	if (rule->suppress_ifgroup != -1 && dev && dev->group == rule->suppress_ifgroup)
157 		goto suppress_route;
158 
159 	return false;
160 
161 suppress_route:
162 	if (!(arg->flags & FIB_LOOKUP_NOREF))
163 		fib_info_put(result->fi);
164 	return true;
165 }
166 
167 static int fib4_rule_match(struct fib_rule *rule, struct flowi *fl, int flags)
168 {
169 	struct fib4_rule *r = (struct fib4_rule *) rule;
170 	struct flowi4 *fl4 = &fl->u.ip4;
171 	__be32 daddr = fl4->daddr;
172 	__be32 saddr = fl4->saddr;
173 
174 	if (((saddr ^ r->src) & r->srcmask) ||
175 	    ((daddr ^ r->dst) & r->dstmask))
176 		return 0;
177 
178 	if (r->tos && (r->tos != fl4->flowi4_tos))
179 		return 0;
180 
181 	if (rule->ip_proto && (rule->ip_proto != fl4->flowi4_proto))
182 		return 0;
183 
184 	if (fib_rule_port_range_set(&rule->sport_range) &&
185 	    !fib_rule_port_inrange(&rule->sport_range, fl4->fl4_sport))
186 		return 0;
187 
188 	if (fib_rule_port_range_set(&rule->dport_range) &&
189 	    !fib_rule_port_inrange(&rule->dport_range, fl4->fl4_dport))
190 		return 0;
191 
192 	return 1;
193 }
194 
195 static struct fib_table *fib_empty_table(struct net *net)
196 {
197 	u32 id = 1;
198 
199 	while (1) {
200 		if (!fib_get_table(net, id))
201 			return fib_new_table(net, id);
202 
203 		if (id++ == RT_TABLE_MAX)
204 			break;
205 	}
206 	return NULL;
207 }
208 
209 static const struct nla_policy fib4_rule_policy[FRA_MAX+1] = {
210 	FRA_GENERIC_POLICY,
211 	[FRA_FLOW]	= { .type = NLA_U32 },
212 };
213 
214 static int fib4_rule_configure(struct fib_rule *rule, struct sk_buff *skb,
215 			       struct fib_rule_hdr *frh,
216 			       struct nlattr **tb,
217 			       struct netlink_ext_ack *extack)
218 {
219 	struct net *net = sock_net(skb->sk);
220 	int err = -EINVAL;
221 	struct fib4_rule *rule4 = (struct fib4_rule *) rule;
222 
223 	if (frh->tos & ~IPTOS_TOS_MASK) {
224 		NL_SET_ERR_MSG(extack, "Invalid tos");
225 		goto errout;
226 	}
227 
228 	/* split local/main if they are not already split */
229 	err = fib_unmerge(net);
230 	if (err)
231 		goto errout;
232 
233 	if (rule->table == RT_TABLE_UNSPEC && !rule->l3mdev) {
234 		if (rule->action == FR_ACT_TO_TBL) {
235 			struct fib_table *table;
236 
237 			table = fib_empty_table(net);
238 			if (!table) {
239 				err = -ENOBUFS;
240 				goto errout;
241 			}
242 
243 			rule->table = table->tb_id;
244 		}
245 	}
246 
247 	if (frh->src_len)
248 		rule4->src = nla_get_in_addr(tb[FRA_SRC]);
249 
250 	if (frh->dst_len)
251 		rule4->dst = nla_get_in_addr(tb[FRA_DST]);
252 
253 #ifdef CONFIG_IP_ROUTE_CLASSID
254 	if (tb[FRA_FLOW]) {
255 		rule4->tclassid = nla_get_u32(tb[FRA_FLOW]);
256 		if (rule4->tclassid)
257 			net->ipv4.fib_num_tclassid_users++;
258 	}
259 #endif
260 
261 	if (fib_rule_requires_fldissect(rule))
262 		net->ipv4.fib_rules_require_fldissect++;
263 
264 	rule4->src_len = frh->src_len;
265 	rule4->srcmask = inet_make_mask(rule4->src_len);
266 	rule4->dst_len = frh->dst_len;
267 	rule4->dstmask = inet_make_mask(rule4->dst_len);
268 	rule4->tos = frh->tos;
269 
270 	net->ipv4.fib_has_custom_rules = true;
271 
272 	err = 0;
273 errout:
274 	return err;
275 }
276 
277 static int fib4_rule_delete(struct fib_rule *rule)
278 {
279 	struct net *net = rule->fr_net;
280 	int err;
281 
282 	/* split local/main if they are not already split */
283 	err = fib_unmerge(net);
284 	if (err)
285 		goto errout;
286 
287 #ifdef CONFIG_IP_ROUTE_CLASSID
288 	if (((struct fib4_rule *)rule)->tclassid)
289 		net->ipv4.fib_num_tclassid_users--;
290 #endif
291 	net->ipv4.fib_has_custom_rules = true;
292 
293 	if (net->ipv4.fib_rules_require_fldissect &&
294 	    fib_rule_requires_fldissect(rule))
295 		net->ipv4.fib_rules_require_fldissect--;
296 errout:
297 	return err;
298 }
299 
300 static int fib4_rule_compare(struct fib_rule *rule, struct fib_rule_hdr *frh,
301 			     struct nlattr **tb)
302 {
303 	struct fib4_rule *rule4 = (struct fib4_rule *) rule;
304 
305 	if (frh->src_len && (rule4->src_len != frh->src_len))
306 		return 0;
307 
308 	if (frh->dst_len && (rule4->dst_len != frh->dst_len))
309 		return 0;
310 
311 	if (frh->tos && (rule4->tos != frh->tos))
312 		return 0;
313 
314 #ifdef CONFIG_IP_ROUTE_CLASSID
315 	if (tb[FRA_FLOW] && (rule4->tclassid != nla_get_u32(tb[FRA_FLOW])))
316 		return 0;
317 #endif
318 
319 	if (frh->src_len && (rule4->src != nla_get_in_addr(tb[FRA_SRC])))
320 		return 0;
321 
322 	if (frh->dst_len && (rule4->dst != nla_get_in_addr(tb[FRA_DST])))
323 		return 0;
324 
325 	return 1;
326 }
327 
328 static int fib4_rule_fill(struct fib_rule *rule, struct sk_buff *skb,
329 			  struct fib_rule_hdr *frh)
330 {
331 	struct fib4_rule *rule4 = (struct fib4_rule *) rule;
332 
333 	frh->dst_len = rule4->dst_len;
334 	frh->src_len = rule4->src_len;
335 	frh->tos = rule4->tos;
336 
337 	if ((rule4->dst_len &&
338 	     nla_put_in_addr(skb, FRA_DST, rule4->dst)) ||
339 	    (rule4->src_len &&
340 	     nla_put_in_addr(skb, FRA_SRC, rule4->src)))
341 		goto nla_put_failure;
342 #ifdef CONFIG_IP_ROUTE_CLASSID
343 	if (rule4->tclassid &&
344 	    nla_put_u32(skb, FRA_FLOW, rule4->tclassid))
345 		goto nla_put_failure;
346 #endif
347 	return 0;
348 
349 nla_put_failure:
350 	return -ENOBUFS;
351 }
352 
353 static size_t fib4_rule_nlmsg_payload(struct fib_rule *rule)
354 {
355 	return nla_total_size(4) /* dst */
356 	       + nla_total_size(4) /* src */
357 	       + nla_total_size(4); /* flow */
358 }
359 
360 static void fib4_rule_flush_cache(struct fib_rules_ops *ops)
361 {
362 	rt_cache_flush(ops->fro_net);
363 }
364 
365 static const struct fib_rules_ops __net_initconst fib4_rules_ops_template = {
366 	.family		= AF_INET,
367 	.rule_size	= sizeof(struct fib4_rule),
368 	.addr_size	= sizeof(u32),
369 	.action		= fib4_rule_action,
370 	.suppress	= fib4_rule_suppress,
371 	.match		= fib4_rule_match,
372 	.configure	= fib4_rule_configure,
373 	.delete		= fib4_rule_delete,
374 	.compare	= fib4_rule_compare,
375 	.fill		= fib4_rule_fill,
376 	.nlmsg_payload	= fib4_rule_nlmsg_payload,
377 	.flush_cache	= fib4_rule_flush_cache,
378 	.nlgroup	= RTNLGRP_IPV4_RULE,
379 	.policy		= fib4_rule_policy,
380 	.owner		= THIS_MODULE,
381 };
382 
383 static int fib_default_rules_init(struct fib_rules_ops *ops)
384 {
385 	int err;
386 
387 	err = fib_default_rule_add(ops, 0, RT_TABLE_LOCAL, 0);
388 	if (err < 0)
389 		return err;
390 	err = fib_default_rule_add(ops, 0x7FFE, RT_TABLE_MAIN, 0);
391 	if (err < 0)
392 		return err;
393 	err = fib_default_rule_add(ops, 0x7FFF, RT_TABLE_DEFAULT, 0);
394 	if (err < 0)
395 		return err;
396 	return 0;
397 }
398 
399 int __net_init fib4_rules_init(struct net *net)
400 {
401 	int err;
402 	struct fib_rules_ops *ops;
403 
404 	ops = fib_rules_register(&fib4_rules_ops_template, net);
405 	if (IS_ERR(ops))
406 		return PTR_ERR(ops);
407 
408 	err = fib_default_rules_init(ops);
409 	if (err < 0)
410 		goto fail;
411 	net->ipv4.rules_ops = ops;
412 	net->ipv4.fib_has_custom_rules = false;
413 	net->ipv4.fib_rules_require_fldissect = 0;
414 	return 0;
415 
416 fail:
417 	/* also cleans all rules already added */
418 	fib_rules_unregister(ops);
419 	return err;
420 }
421 
422 void __net_exit fib4_rules_exit(struct net *net)
423 {
424 	fib_rules_unregister(net->ipv4.rules_ops);
425 }
426