xref: /openbmc/linux/net/ipv4/fib_rules.c (revision b737eecd)
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/nexthop.h>
31 #include <net/fib_rules.h>
32 #include <linux/indirect_call_wrapper.h>
33 
34 struct fib4_rule {
35 	struct fib_rule		common;
36 	u8			dst_len;
37 	u8			src_len;
38 	u8			tos;
39 	__be32			src;
40 	__be32			srcmask;
41 	__be32			dst;
42 	__be32			dstmask;
43 #ifdef CONFIG_IP_ROUTE_CLASSID
44 	u32			tclassid;
45 #endif
46 };
47 
48 static bool fib4_rule_matchall(const struct fib_rule *rule)
49 {
50 	struct fib4_rule *r = container_of(rule, struct fib4_rule, common);
51 
52 	if (r->dst_len || r->src_len || r->tos)
53 		return false;
54 	return fib_rule_matchall(rule);
55 }
56 
57 bool fib4_rule_default(const struct fib_rule *rule)
58 {
59 	if (!fib4_rule_matchall(rule) || rule->action != FR_ACT_TO_TBL ||
60 	    rule->l3mdev)
61 		return false;
62 	if (rule->table != RT_TABLE_LOCAL && rule->table != RT_TABLE_MAIN &&
63 	    rule->table != RT_TABLE_DEFAULT)
64 		return false;
65 	return true;
66 }
67 EXPORT_SYMBOL_GPL(fib4_rule_default);
68 
69 int fib4_rules_dump(struct net *net, struct notifier_block *nb,
70 		    struct netlink_ext_ack *extack)
71 {
72 	return fib_rules_dump(net, nb, AF_INET, extack);
73 }
74 
75 unsigned int fib4_rules_seq_read(struct net *net)
76 {
77 	return fib_rules_seq_read(net, AF_INET);
78 }
79 
80 int __fib_lookup(struct net *net, struct flowi4 *flp,
81 		 struct fib_result *res, unsigned int flags)
82 {
83 	struct fib_lookup_arg arg = {
84 		.result = res,
85 		.flags = flags,
86 	};
87 	int err;
88 
89 	/* update flow if oif or iif point to device enslaved to l3mdev */
90 	l3mdev_update_flow(net, flowi4_to_flowi(flp));
91 
92 	err = fib_rules_lookup(net->ipv4.rules_ops, flowi4_to_flowi(flp), 0, &arg);
93 #ifdef CONFIG_IP_ROUTE_CLASSID
94 	if (arg.rule)
95 		res->tclassid = ((struct fib4_rule *)arg.rule)->tclassid;
96 	else
97 		res->tclassid = 0;
98 #endif
99 
100 	if (err == -ESRCH)
101 		err = -ENETUNREACH;
102 
103 	return err;
104 }
105 EXPORT_SYMBOL_GPL(__fib_lookup);
106 
107 INDIRECT_CALLABLE_SCOPE int fib4_rule_action(struct fib_rule *rule,
108 					     struct flowi *flp, int flags,
109 					     struct fib_lookup_arg *arg)
110 {
111 	int err = -EAGAIN;
112 	struct fib_table *tbl;
113 	u32 tb_id;
114 
115 	switch (rule->action) {
116 	case FR_ACT_TO_TBL:
117 		break;
118 
119 	case FR_ACT_UNREACHABLE:
120 		return -ENETUNREACH;
121 
122 	case FR_ACT_PROHIBIT:
123 		return -EACCES;
124 
125 	case FR_ACT_BLACKHOLE:
126 	default:
127 		return -EINVAL;
128 	}
129 
130 	rcu_read_lock();
131 
132 	tb_id = fib_rule_get_table(rule, arg);
133 	tbl = fib_get_table(rule->fr_net, tb_id);
134 	if (tbl)
135 		err = fib_table_lookup(tbl, &flp->u.ip4,
136 				       (struct fib_result *)arg->result,
137 				       arg->flags);
138 
139 	rcu_read_unlock();
140 	return err;
141 }
142 
143 INDIRECT_CALLABLE_SCOPE bool fib4_rule_suppress(struct fib_rule *rule,
144 						struct fib_lookup_arg *arg)
145 {
146 	struct fib_result *result = (struct fib_result *) arg->result;
147 	struct net_device *dev = NULL;
148 
149 	if (result->fi) {
150 		struct fib_nh_common *nhc = fib_info_nhc(result->fi, 0);
151 
152 		dev = nhc->nhc_dev;
153 	}
154 
155 	/* do not accept result if the route does
156 	 * not meet the required prefix length
157 	 */
158 	if (result->prefixlen <= rule->suppress_prefixlen)
159 		goto suppress_route;
160 
161 	/* do not accept result if the route uses a device
162 	 * belonging to a forbidden interface group
163 	 */
164 	if (rule->suppress_ifgroup != -1 && dev && dev->group == rule->suppress_ifgroup)
165 		goto suppress_route;
166 
167 	return false;
168 
169 suppress_route:
170 	if (!(arg->flags & FIB_LOOKUP_NOREF))
171 		fib_info_put(result->fi);
172 	return true;
173 }
174 
175 INDIRECT_CALLABLE_SCOPE int fib4_rule_match(struct fib_rule *rule,
176 					    struct flowi *fl, int flags)
177 {
178 	struct fib4_rule *r = (struct fib4_rule *) rule;
179 	struct flowi4 *fl4 = &fl->u.ip4;
180 	__be32 daddr = fl4->daddr;
181 	__be32 saddr = fl4->saddr;
182 
183 	if (((saddr ^ r->src) & r->srcmask) ||
184 	    ((daddr ^ r->dst) & r->dstmask))
185 		return 0;
186 
187 	if (r->tos && (r->tos != fl4->flowi4_tos))
188 		return 0;
189 
190 	if (rule->ip_proto && (rule->ip_proto != fl4->flowi4_proto))
191 		return 0;
192 
193 	if (fib_rule_port_range_set(&rule->sport_range) &&
194 	    !fib_rule_port_inrange(&rule->sport_range, fl4->fl4_sport))
195 		return 0;
196 
197 	if (fib_rule_port_range_set(&rule->dport_range) &&
198 	    !fib_rule_port_inrange(&rule->dport_range, fl4->fl4_dport))
199 		return 0;
200 
201 	return 1;
202 }
203 
204 static struct fib_table *fib_empty_table(struct net *net)
205 {
206 	u32 id = 1;
207 
208 	while (1) {
209 		if (!fib_get_table(net, id))
210 			return fib_new_table(net, id);
211 
212 		if (id++ == RT_TABLE_MAX)
213 			break;
214 	}
215 	return NULL;
216 }
217 
218 static const struct nla_policy fib4_rule_policy[FRA_MAX+1] = {
219 	FRA_GENERIC_POLICY,
220 	[FRA_FLOW]	= { .type = NLA_U32 },
221 };
222 
223 static int fib4_rule_configure(struct fib_rule *rule, struct sk_buff *skb,
224 			       struct fib_rule_hdr *frh,
225 			       struct nlattr **tb,
226 			       struct netlink_ext_ack *extack)
227 {
228 	struct net *net = sock_net(skb->sk);
229 	int err = -EINVAL;
230 	struct fib4_rule *rule4 = (struct fib4_rule *) rule;
231 
232 	if (frh->tos & ~IPTOS_TOS_MASK) {
233 		NL_SET_ERR_MSG(extack, "Invalid tos");
234 		goto errout;
235 	}
236 
237 	/* split local/main if they are not already split */
238 	err = fib_unmerge(net);
239 	if (err)
240 		goto errout;
241 
242 	if (rule->table == RT_TABLE_UNSPEC && !rule->l3mdev) {
243 		if (rule->action == FR_ACT_TO_TBL) {
244 			struct fib_table *table;
245 
246 			table = fib_empty_table(net);
247 			if (!table) {
248 				err = -ENOBUFS;
249 				goto errout;
250 			}
251 
252 			rule->table = table->tb_id;
253 		}
254 	}
255 
256 	if (frh->src_len)
257 		rule4->src = nla_get_in_addr(tb[FRA_SRC]);
258 
259 	if (frh->dst_len)
260 		rule4->dst = nla_get_in_addr(tb[FRA_DST]);
261 
262 #ifdef CONFIG_IP_ROUTE_CLASSID
263 	if (tb[FRA_FLOW]) {
264 		rule4->tclassid = nla_get_u32(tb[FRA_FLOW]);
265 		if (rule4->tclassid)
266 			net->ipv4.fib_num_tclassid_users++;
267 	}
268 #endif
269 
270 	if (fib_rule_requires_fldissect(rule))
271 		net->ipv4.fib_rules_require_fldissect++;
272 
273 	rule4->src_len = frh->src_len;
274 	rule4->srcmask = inet_make_mask(rule4->src_len);
275 	rule4->dst_len = frh->dst_len;
276 	rule4->dstmask = inet_make_mask(rule4->dst_len);
277 	rule4->tos = frh->tos;
278 
279 	net->ipv4.fib_has_custom_rules = true;
280 
281 	err = 0;
282 errout:
283 	return err;
284 }
285 
286 static int fib4_rule_delete(struct fib_rule *rule)
287 {
288 	struct net *net = rule->fr_net;
289 	int err;
290 
291 	/* split local/main if they are not already split */
292 	err = fib_unmerge(net);
293 	if (err)
294 		goto errout;
295 
296 #ifdef CONFIG_IP_ROUTE_CLASSID
297 	if (((struct fib4_rule *)rule)->tclassid)
298 		net->ipv4.fib_num_tclassid_users--;
299 #endif
300 	net->ipv4.fib_has_custom_rules = true;
301 
302 	if (net->ipv4.fib_rules_require_fldissect &&
303 	    fib_rule_requires_fldissect(rule))
304 		net->ipv4.fib_rules_require_fldissect--;
305 errout:
306 	return err;
307 }
308 
309 static int fib4_rule_compare(struct fib_rule *rule, struct fib_rule_hdr *frh,
310 			     struct nlattr **tb)
311 {
312 	struct fib4_rule *rule4 = (struct fib4_rule *) rule;
313 
314 	if (frh->src_len && (rule4->src_len != frh->src_len))
315 		return 0;
316 
317 	if (frh->dst_len && (rule4->dst_len != frh->dst_len))
318 		return 0;
319 
320 	if (frh->tos && (rule4->tos != frh->tos))
321 		return 0;
322 
323 #ifdef CONFIG_IP_ROUTE_CLASSID
324 	if (tb[FRA_FLOW] && (rule4->tclassid != nla_get_u32(tb[FRA_FLOW])))
325 		return 0;
326 #endif
327 
328 	if (frh->src_len && (rule4->src != nla_get_in_addr(tb[FRA_SRC])))
329 		return 0;
330 
331 	if (frh->dst_len && (rule4->dst != nla_get_in_addr(tb[FRA_DST])))
332 		return 0;
333 
334 	return 1;
335 }
336 
337 static int fib4_rule_fill(struct fib_rule *rule, struct sk_buff *skb,
338 			  struct fib_rule_hdr *frh)
339 {
340 	struct fib4_rule *rule4 = (struct fib4_rule *) rule;
341 
342 	frh->dst_len = rule4->dst_len;
343 	frh->src_len = rule4->src_len;
344 	frh->tos = rule4->tos;
345 
346 	if ((rule4->dst_len &&
347 	     nla_put_in_addr(skb, FRA_DST, rule4->dst)) ||
348 	    (rule4->src_len &&
349 	     nla_put_in_addr(skb, FRA_SRC, rule4->src)))
350 		goto nla_put_failure;
351 #ifdef CONFIG_IP_ROUTE_CLASSID
352 	if (rule4->tclassid &&
353 	    nla_put_u32(skb, FRA_FLOW, rule4->tclassid))
354 		goto nla_put_failure;
355 #endif
356 	return 0;
357 
358 nla_put_failure:
359 	return -ENOBUFS;
360 }
361 
362 static size_t fib4_rule_nlmsg_payload(struct fib_rule *rule)
363 {
364 	return nla_total_size(4) /* dst */
365 	       + nla_total_size(4) /* src */
366 	       + nla_total_size(4); /* flow */
367 }
368 
369 static void fib4_rule_flush_cache(struct fib_rules_ops *ops)
370 {
371 	rt_cache_flush(ops->fro_net);
372 }
373 
374 static const struct fib_rules_ops __net_initconst fib4_rules_ops_template = {
375 	.family		= AF_INET,
376 	.rule_size	= sizeof(struct fib4_rule),
377 	.addr_size	= sizeof(u32),
378 	.action		= fib4_rule_action,
379 	.suppress	= fib4_rule_suppress,
380 	.match		= fib4_rule_match,
381 	.configure	= fib4_rule_configure,
382 	.delete		= fib4_rule_delete,
383 	.compare	= fib4_rule_compare,
384 	.fill		= fib4_rule_fill,
385 	.nlmsg_payload	= fib4_rule_nlmsg_payload,
386 	.flush_cache	= fib4_rule_flush_cache,
387 	.nlgroup	= RTNLGRP_IPV4_RULE,
388 	.policy		= fib4_rule_policy,
389 	.owner		= THIS_MODULE,
390 };
391 
392 static int fib_default_rules_init(struct fib_rules_ops *ops)
393 {
394 	int err;
395 
396 	err = fib_default_rule_add(ops, 0, RT_TABLE_LOCAL, 0);
397 	if (err < 0)
398 		return err;
399 	err = fib_default_rule_add(ops, 0x7FFE, RT_TABLE_MAIN, 0);
400 	if (err < 0)
401 		return err;
402 	err = fib_default_rule_add(ops, 0x7FFF, RT_TABLE_DEFAULT, 0);
403 	if (err < 0)
404 		return err;
405 	return 0;
406 }
407 
408 int __net_init fib4_rules_init(struct net *net)
409 {
410 	int err;
411 	struct fib_rules_ops *ops;
412 
413 	ops = fib_rules_register(&fib4_rules_ops_template, net);
414 	if (IS_ERR(ops))
415 		return PTR_ERR(ops);
416 
417 	err = fib_default_rules_init(ops);
418 	if (err < 0)
419 		goto fail;
420 	net->ipv4.rules_ops = ops;
421 	net->ipv4.fib_has_custom_rules = false;
422 	net->ipv4.fib_rules_require_fldissect = 0;
423 	return 0;
424 
425 fail:
426 	/* also cleans all rules already added */
427 	fib_rules_unregister(ops);
428 	return err;
429 }
430 
431 void __net_exit fib4_rules_exit(struct net *net)
432 {
433 	fib_rules_unregister(net->ipv4.rules_ops);
434 }
435