xref: /linux/net/ipv4/fib_rules.c (revision 9ee0034b8f49aaaa7e7c2da8db1038915db99c19)
1 /*
2  * INET		An implementation of the TCP/IP protocol suite for the LINUX
3  *		operating system.  INET is implemented using the  BSD Socket
4  *		interface as the means of communication with the user level.
5  *
6  *		IPv4 Forwarding Information Base: policy rules.
7  *
8  * Authors:	Alexey Kuznetsov, <kuznet@ms2.inr.ac.ru>
9  *		Thomas Graf <tgraf@suug.ch>
10  *
11  *		This program is free software; you can redistribute it and/or
12  *		modify it under the terms of the GNU General Public License
13  *		as published by the Free Software Foundation; either version
14  *		2 of the License, or (at your option) any later version.
15  *
16  * Fixes:
17  *		Rani Assaf	:	local_rule cannot be deleted
18  *		Marc Boucher	:	routing by fwmark
19  */
20 
21 #include <linux/types.h>
22 #include <linux/kernel.h>
23 #include <linux/netdevice.h>
24 #include <linux/netlink.h>
25 #include <linux/inetdevice.h>
26 #include <linux/init.h>
27 #include <linux/list.h>
28 #include <linux/rcupdate.h>
29 #include <linux/export.h>
30 #include <net/ip.h>
31 #include <net/route.h>
32 #include <net/tcp.h>
33 #include <net/ip_fib.h>
34 #include <net/fib_rules.h>
35 
36 struct fib4_rule {
37 	struct fib_rule		common;
38 	u8			dst_len;
39 	u8			src_len;
40 	u8			tos;
41 	__be32			src;
42 	__be32			srcmask;
43 	__be32			dst;
44 	__be32			dstmask;
45 #ifdef CONFIG_IP_ROUTE_CLASSID
46 	u32			tclassid;
47 #endif
48 };
49 
50 int __fib_lookup(struct net *net, struct flowi4 *flp,
51 		 struct fib_result *res, unsigned int flags)
52 {
53 	struct fib_lookup_arg arg = {
54 		.result = res,
55 		.flags = flags,
56 	};
57 	int err;
58 
59 	/* update flow if oif or iif point to device enslaved to l3mdev */
60 	l3mdev_update_flow(net, flowi4_to_flowi(flp));
61 
62 	err = fib_rules_lookup(net->ipv4.rules_ops, flowi4_to_flowi(flp), 0, &arg);
63 #ifdef CONFIG_IP_ROUTE_CLASSID
64 	if (arg.rule)
65 		res->tclassid = ((struct fib4_rule *)arg.rule)->tclassid;
66 	else
67 		res->tclassid = 0;
68 #endif
69 
70 	if (err == -ESRCH)
71 		err = -ENETUNREACH;
72 
73 	return err;
74 }
75 EXPORT_SYMBOL_GPL(__fib_lookup);
76 
77 static int fib4_rule_action(struct fib_rule *rule, struct flowi *flp,
78 			    int flags, struct fib_lookup_arg *arg)
79 {
80 	int err = -EAGAIN;
81 	struct fib_table *tbl;
82 	u32 tb_id;
83 
84 	switch (rule->action) {
85 	case FR_ACT_TO_TBL:
86 		break;
87 
88 	case FR_ACT_UNREACHABLE:
89 		return -ENETUNREACH;
90 
91 	case FR_ACT_PROHIBIT:
92 		return -EACCES;
93 
94 	case FR_ACT_BLACKHOLE:
95 	default:
96 		return -EINVAL;
97 	}
98 
99 	rcu_read_lock();
100 
101 	tb_id = fib_rule_get_table(rule, arg);
102 	tbl = fib_get_table(rule->fr_net, tb_id);
103 	if (tbl)
104 		err = fib_table_lookup(tbl, &flp->u.ip4,
105 				       (struct fib_result *)arg->result,
106 				       arg->flags);
107 
108 	rcu_read_unlock();
109 	return err;
110 }
111 
112 static bool fib4_rule_suppress(struct fib_rule *rule, struct fib_lookup_arg *arg)
113 {
114 	struct fib_result *result = (struct fib_result *) arg->result;
115 	struct net_device *dev = NULL;
116 
117 	if (result->fi)
118 		dev = result->fi->fib_dev;
119 
120 	/* do not accept result if the route does
121 	 * not meet the required prefix length
122 	 */
123 	if (result->prefixlen <= rule->suppress_prefixlen)
124 		goto suppress_route;
125 
126 	/* do not accept result if the route uses a device
127 	 * belonging to a forbidden interface group
128 	 */
129 	if (rule->suppress_ifgroup != -1 && dev && dev->group == rule->suppress_ifgroup)
130 		goto suppress_route;
131 
132 	return false;
133 
134 suppress_route:
135 	if (!(arg->flags & FIB_LOOKUP_NOREF))
136 		fib_info_put(result->fi);
137 	return true;
138 }
139 
140 static int fib4_rule_match(struct fib_rule *rule, struct flowi *fl, int flags)
141 {
142 	struct fib4_rule *r = (struct fib4_rule *) rule;
143 	struct flowi4 *fl4 = &fl->u.ip4;
144 	__be32 daddr = fl4->daddr;
145 	__be32 saddr = fl4->saddr;
146 
147 	if (((saddr ^ r->src) & r->srcmask) ||
148 	    ((daddr ^ r->dst) & r->dstmask))
149 		return 0;
150 
151 	if (r->tos && (r->tos != fl4->flowi4_tos))
152 		return 0;
153 
154 	return 1;
155 }
156 
157 static struct fib_table *fib_empty_table(struct net *net)
158 {
159 	u32 id;
160 
161 	for (id = 1; id <= RT_TABLE_MAX; id++)
162 		if (!fib_get_table(net, id))
163 			return fib_new_table(net, id);
164 	return NULL;
165 }
166 
167 static const struct nla_policy fib4_rule_policy[FRA_MAX+1] = {
168 	FRA_GENERIC_POLICY,
169 	[FRA_FLOW]	= { .type = NLA_U32 },
170 };
171 
172 static int fib4_rule_configure(struct fib_rule *rule, struct sk_buff *skb,
173 			       struct fib_rule_hdr *frh,
174 			       struct nlattr **tb)
175 {
176 	struct net *net = sock_net(skb->sk);
177 	int err = -EINVAL;
178 	struct fib4_rule *rule4 = (struct fib4_rule *) rule;
179 
180 	if (frh->tos & ~IPTOS_TOS_MASK)
181 		goto errout;
182 
183 	/* split local/main if they are not already split */
184 	err = fib_unmerge(net);
185 	if (err)
186 		goto errout;
187 
188 	if (rule->table == RT_TABLE_UNSPEC && !rule->l3mdev) {
189 		if (rule->action == FR_ACT_TO_TBL) {
190 			struct fib_table *table;
191 
192 			table = fib_empty_table(net);
193 			if (!table) {
194 				err = -ENOBUFS;
195 				goto errout;
196 			}
197 
198 			rule->table = table->tb_id;
199 		}
200 	}
201 
202 	if (frh->src_len)
203 		rule4->src = nla_get_in_addr(tb[FRA_SRC]);
204 
205 	if (frh->dst_len)
206 		rule4->dst = nla_get_in_addr(tb[FRA_DST]);
207 
208 #ifdef CONFIG_IP_ROUTE_CLASSID
209 	if (tb[FRA_FLOW]) {
210 		rule4->tclassid = nla_get_u32(tb[FRA_FLOW]);
211 		if (rule4->tclassid)
212 			net->ipv4.fib_num_tclassid_users++;
213 	}
214 #endif
215 
216 	rule4->src_len = frh->src_len;
217 	rule4->srcmask = inet_make_mask(rule4->src_len);
218 	rule4->dst_len = frh->dst_len;
219 	rule4->dstmask = inet_make_mask(rule4->dst_len);
220 	rule4->tos = frh->tos;
221 
222 	net->ipv4.fib_has_custom_rules = true;
223 	fib_flush_external(rule->fr_net);
224 
225 	err = 0;
226 errout:
227 	return err;
228 }
229 
230 static int fib4_rule_delete(struct fib_rule *rule)
231 {
232 	struct net *net = rule->fr_net;
233 	int err;
234 
235 	/* split local/main if they are not already split */
236 	err = fib_unmerge(net);
237 	if (err)
238 		goto errout;
239 
240 #ifdef CONFIG_IP_ROUTE_CLASSID
241 	if (((struct fib4_rule *)rule)->tclassid)
242 		net->ipv4.fib_num_tclassid_users--;
243 #endif
244 	net->ipv4.fib_has_custom_rules = true;
245 	fib_flush_external(rule->fr_net);
246 errout:
247 	return err;
248 }
249 
250 static int fib4_rule_compare(struct fib_rule *rule, struct fib_rule_hdr *frh,
251 			     struct nlattr **tb)
252 {
253 	struct fib4_rule *rule4 = (struct fib4_rule *) rule;
254 
255 	if (frh->src_len && (rule4->src_len != frh->src_len))
256 		return 0;
257 
258 	if (frh->dst_len && (rule4->dst_len != frh->dst_len))
259 		return 0;
260 
261 	if (frh->tos && (rule4->tos != frh->tos))
262 		return 0;
263 
264 #ifdef CONFIG_IP_ROUTE_CLASSID
265 	if (tb[FRA_FLOW] && (rule4->tclassid != nla_get_u32(tb[FRA_FLOW])))
266 		return 0;
267 #endif
268 
269 	if (frh->src_len && (rule4->src != nla_get_in_addr(tb[FRA_SRC])))
270 		return 0;
271 
272 	if (frh->dst_len && (rule4->dst != nla_get_in_addr(tb[FRA_DST])))
273 		return 0;
274 
275 	return 1;
276 }
277 
278 static int fib4_rule_fill(struct fib_rule *rule, struct sk_buff *skb,
279 			  struct fib_rule_hdr *frh)
280 {
281 	struct fib4_rule *rule4 = (struct fib4_rule *) rule;
282 
283 	frh->dst_len = rule4->dst_len;
284 	frh->src_len = rule4->src_len;
285 	frh->tos = rule4->tos;
286 
287 	if ((rule4->dst_len &&
288 	     nla_put_in_addr(skb, FRA_DST, rule4->dst)) ||
289 	    (rule4->src_len &&
290 	     nla_put_in_addr(skb, FRA_SRC, rule4->src)))
291 		goto nla_put_failure;
292 #ifdef CONFIG_IP_ROUTE_CLASSID
293 	if (rule4->tclassid &&
294 	    nla_put_u32(skb, FRA_FLOW, rule4->tclassid))
295 		goto nla_put_failure;
296 #endif
297 	return 0;
298 
299 nla_put_failure:
300 	return -ENOBUFS;
301 }
302 
303 static size_t fib4_rule_nlmsg_payload(struct fib_rule *rule)
304 {
305 	return nla_total_size(4) /* dst */
306 	       + nla_total_size(4) /* src */
307 	       + nla_total_size(4); /* flow */
308 }
309 
310 static void fib4_rule_flush_cache(struct fib_rules_ops *ops)
311 {
312 	rt_cache_flush(ops->fro_net);
313 }
314 
315 static const struct fib_rules_ops __net_initconst fib4_rules_ops_template = {
316 	.family		= AF_INET,
317 	.rule_size	= sizeof(struct fib4_rule),
318 	.addr_size	= sizeof(u32),
319 	.action		= fib4_rule_action,
320 	.suppress	= fib4_rule_suppress,
321 	.match		= fib4_rule_match,
322 	.configure	= fib4_rule_configure,
323 	.delete		= fib4_rule_delete,
324 	.compare	= fib4_rule_compare,
325 	.fill		= fib4_rule_fill,
326 	.nlmsg_payload	= fib4_rule_nlmsg_payload,
327 	.flush_cache	= fib4_rule_flush_cache,
328 	.nlgroup	= RTNLGRP_IPV4_RULE,
329 	.policy		= fib4_rule_policy,
330 	.owner		= THIS_MODULE,
331 };
332 
333 static int fib_default_rules_init(struct fib_rules_ops *ops)
334 {
335 	int err;
336 
337 	err = fib_default_rule_add(ops, 0, RT_TABLE_LOCAL, 0);
338 	if (err < 0)
339 		return err;
340 	err = fib_default_rule_add(ops, 0x7FFE, RT_TABLE_MAIN, 0);
341 	if (err < 0)
342 		return err;
343 	err = fib_default_rule_add(ops, 0x7FFF, RT_TABLE_DEFAULT, 0);
344 	if (err < 0)
345 		return err;
346 	return 0;
347 }
348 
349 int __net_init fib4_rules_init(struct net *net)
350 {
351 	int err;
352 	struct fib_rules_ops *ops;
353 
354 	ops = fib_rules_register(&fib4_rules_ops_template, net);
355 	if (IS_ERR(ops))
356 		return PTR_ERR(ops);
357 
358 	err = fib_default_rules_init(ops);
359 	if (err < 0)
360 		goto fail;
361 	net->ipv4.rules_ops = ops;
362 	net->ipv4.fib_has_custom_rules = false;
363 	return 0;
364 
365 fail:
366 	/* also cleans all rules already added */
367 	fib_rules_unregister(ops);
368 	return err;
369 }
370 
371 void __net_exit fib4_rules_exit(struct net *net)
372 {
373 	fib_rules_unregister(net->ipv4.rules_ops);
374 }
375