xref: /openbmc/linux/net/ipv4/fib_rules.c (revision f3a8b664)
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 int call_fib_rule_notifiers(struct net *net,
168 				   enum fib_event_type event_type)
169 {
170 	struct fib_notifier_info info;
171 
172 	return call_fib_notifiers(net, event_type, &info);
173 }
174 
175 static const struct nla_policy fib4_rule_policy[FRA_MAX+1] = {
176 	FRA_GENERIC_POLICY,
177 	[FRA_FLOW]	= { .type = NLA_U32 },
178 };
179 
180 static int fib4_rule_configure(struct fib_rule *rule, struct sk_buff *skb,
181 			       struct fib_rule_hdr *frh,
182 			       struct nlattr **tb)
183 {
184 	struct net *net = sock_net(skb->sk);
185 	int err = -EINVAL;
186 	struct fib4_rule *rule4 = (struct fib4_rule *) rule;
187 
188 	if (frh->tos & ~IPTOS_TOS_MASK)
189 		goto errout;
190 
191 	/* split local/main if they are not already split */
192 	err = fib_unmerge(net);
193 	if (err)
194 		goto errout;
195 
196 	if (rule->table == RT_TABLE_UNSPEC && !rule->l3mdev) {
197 		if (rule->action == FR_ACT_TO_TBL) {
198 			struct fib_table *table;
199 
200 			table = fib_empty_table(net);
201 			if (!table) {
202 				err = -ENOBUFS;
203 				goto errout;
204 			}
205 
206 			rule->table = table->tb_id;
207 		}
208 	}
209 
210 	if (frh->src_len)
211 		rule4->src = nla_get_in_addr(tb[FRA_SRC]);
212 
213 	if (frh->dst_len)
214 		rule4->dst = nla_get_in_addr(tb[FRA_DST]);
215 
216 #ifdef CONFIG_IP_ROUTE_CLASSID
217 	if (tb[FRA_FLOW]) {
218 		rule4->tclassid = nla_get_u32(tb[FRA_FLOW]);
219 		if (rule4->tclassid)
220 			net->ipv4.fib_num_tclassid_users++;
221 	}
222 #endif
223 
224 	rule4->src_len = frh->src_len;
225 	rule4->srcmask = inet_make_mask(rule4->src_len);
226 	rule4->dst_len = frh->dst_len;
227 	rule4->dstmask = inet_make_mask(rule4->dst_len);
228 	rule4->tos = frh->tos;
229 
230 	net->ipv4.fib_has_custom_rules = true;
231 	call_fib_rule_notifiers(net, FIB_EVENT_RULE_ADD);
232 
233 	err = 0;
234 errout:
235 	return err;
236 }
237 
238 static int fib4_rule_delete(struct fib_rule *rule)
239 {
240 	struct net *net = rule->fr_net;
241 	int err;
242 
243 	/* split local/main if they are not already split */
244 	err = fib_unmerge(net);
245 	if (err)
246 		goto errout;
247 
248 #ifdef CONFIG_IP_ROUTE_CLASSID
249 	if (((struct fib4_rule *)rule)->tclassid)
250 		net->ipv4.fib_num_tclassid_users--;
251 #endif
252 	net->ipv4.fib_has_custom_rules = true;
253 	call_fib_rule_notifiers(net, FIB_EVENT_RULE_DEL);
254 errout:
255 	return err;
256 }
257 
258 static int fib4_rule_compare(struct fib_rule *rule, struct fib_rule_hdr *frh,
259 			     struct nlattr **tb)
260 {
261 	struct fib4_rule *rule4 = (struct fib4_rule *) rule;
262 
263 	if (frh->src_len && (rule4->src_len != frh->src_len))
264 		return 0;
265 
266 	if (frh->dst_len && (rule4->dst_len != frh->dst_len))
267 		return 0;
268 
269 	if (frh->tos && (rule4->tos != frh->tos))
270 		return 0;
271 
272 #ifdef CONFIG_IP_ROUTE_CLASSID
273 	if (tb[FRA_FLOW] && (rule4->tclassid != nla_get_u32(tb[FRA_FLOW])))
274 		return 0;
275 #endif
276 
277 	if (frh->src_len && (rule4->src != nla_get_in_addr(tb[FRA_SRC])))
278 		return 0;
279 
280 	if (frh->dst_len && (rule4->dst != nla_get_in_addr(tb[FRA_DST])))
281 		return 0;
282 
283 	return 1;
284 }
285 
286 static int fib4_rule_fill(struct fib_rule *rule, struct sk_buff *skb,
287 			  struct fib_rule_hdr *frh)
288 {
289 	struct fib4_rule *rule4 = (struct fib4_rule *) rule;
290 
291 	frh->dst_len = rule4->dst_len;
292 	frh->src_len = rule4->src_len;
293 	frh->tos = rule4->tos;
294 
295 	if ((rule4->dst_len &&
296 	     nla_put_in_addr(skb, FRA_DST, rule4->dst)) ||
297 	    (rule4->src_len &&
298 	     nla_put_in_addr(skb, FRA_SRC, rule4->src)))
299 		goto nla_put_failure;
300 #ifdef CONFIG_IP_ROUTE_CLASSID
301 	if (rule4->tclassid &&
302 	    nla_put_u32(skb, FRA_FLOW, rule4->tclassid))
303 		goto nla_put_failure;
304 #endif
305 	return 0;
306 
307 nla_put_failure:
308 	return -ENOBUFS;
309 }
310 
311 static size_t fib4_rule_nlmsg_payload(struct fib_rule *rule)
312 {
313 	return nla_total_size(4) /* dst */
314 	       + nla_total_size(4) /* src */
315 	       + nla_total_size(4); /* flow */
316 }
317 
318 static void fib4_rule_flush_cache(struct fib_rules_ops *ops)
319 {
320 	rt_cache_flush(ops->fro_net);
321 }
322 
323 static const struct fib_rules_ops __net_initconst fib4_rules_ops_template = {
324 	.family		= AF_INET,
325 	.rule_size	= sizeof(struct fib4_rule),
326 	.addr_size	= sizeof(u32),
327 	.action		= fib4_rule_action,
328 	.suppress	= fib4_rule_suppress,
329 	.match		= fib4_rule_match,
330 	.configure	= fib4_rule_configure,
331 	.delete		= fib4_rule_delete,
332 	.compare	= fib4_rule_compare,
333 	.fill		= fib4_rule_fill,
334 	.nlmsg_payload	= fib4_rule_nlmsg_payload,
335 	.flush_cache	= fib4_rule_flush_cache,
336 	.nlgroup	= RTNLGRP_IPV4_RULE,
337 	.policy		= fib4_rule_policy,
338 	.owner		= THIS_MODULE,
339 };
340 
341 static int fib_default_rules_init(struct fib_rules_ops *ops)
342 {
343 	int err;
344 
345 	err = fib_default_rule_add(ops, 0, RT_TABLE_LOCAL, 0);
346 	if (err < 0)
347 		return err;
348 	err = fib_default_rule_add(ops, 0x7FFE, RT_TABLE_MAIN, 0);
349 	if (err < 0)
350 		return err;
351 	err = fib_default_rule_add(ops, 0x7FFF, RT_TABLE_DEFAULT, 0);
352 	if (err < 0)
353 		return err;
354 	return 0;
355 }
356 
357 int __net_init fib4_rules_init(struct net *net)
358 {
359 	int err;
360 	struct fib_rules_ops *ops;
361 
362 	ops = fib_rules_register(&fib4_rules_ops_template, net);
363 	if (IS_ERR(ops))
364 		return PTR_ERR(ops);
365 
366 	err = fib_default_rules_init(ops);
367 	if (err < 0)
368 		goto fail;
369 	net->ipv4.rules_ops = ops;
370 	net->ipv4.fib_has_custom_rules = false;
371 	return 0;
372 
373 fail:
374 	/* also cleans all rules already added */
375 	fib_rules_unregister(ops);
376 	return err;
377 }
378 
379 void __net_exit fib4_rules_exit(struct net *net)
380 {
381 	fib_rules_unregister(net->ipv4.rules_ops);
382 }
383