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