xref: /linux/net/ipv4/fib_rules.c (revision 4f1933620f57145212cdbb1ac6ce099eeeb21c5a)
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  * Version:	$Id: fib_rules.c,v 1.17 2001/10/31 21:55:54 davem Exp $
9  *
10  * Authors:	Alexey Kuznetsov, <kuznet@ms2.inr.ac.ru>
11  *
12  *		This program is free software; you can redistribute it and/or
13  *		modify it under the terms of the GNU General Public License
14  *		as published by the Free Software Foundation; either version
15  *		2 of the License, or (at your option) any later version.
16  *
17  * Fixes:
18  * 		Rani Assaf	:	local_rule cannot be deleted
19  *		Marc Boucher	:	routing by fwmark
20  */
21 
22 #include <linux/config.h>
23 #include <asm/uaccess.h>
24 #include <asm/system.h>
25 #include <linux/bitops.h>
26 #include <linux/types.h>
27 #include <linux/kernel.h>
28 #include <linux/sched.h>
29 #include <linux/mm.h>
30 #include <linux/string.h>
31 #include <linux/socket.h>
32 #include <linux/sockios.h>
33 #include <linux/errno.h>
34 #include <linux/in.h>
35 #include <linux/inet.h>
36 #include <linux/inetdevice.h>
37 #include <linux/netdevice.h>
38 #include <linux/if_arp.h>
39 #include <linux/proc_fs.h>
40 #include <linux/skbuff.h>
41 #include <linux/netlink.h>
42 #include <linux/init.h>
43 
44 #include <net/ip.h>
45 #include <net/protocol.h>
46 #include <net/route.h>
47 #include <net/tcp.h>
48 #include <net/sock.h>
49 #include <net/ip_fib.h>
50 
51 #define FRprintk(a...)
52 
53 struct fib_rule
54 {
55 	struct fib_rule *r_next;
56 	atomic_t	r_clntref;
57 	u32		r_preference;
58 	unsigned char	r_table;
59 	unsigned char	r_action;
60 	unsigned char	r_dst_len;
61 	unsigned char	r_src_len;
62 	u32		r_src;
63 	u32		r_srcmask;
64 	u32		r_dst;
65 	u32		r_dstmask;
66 	u32		r_srcmap;
67 	u8		r_flags;
68 	u8		r_tos;
69 #ifdef CONFIG_IP_ROUTE_FWMARK
70 	u32		r_fwmark;
71 #endif
72 	int		r_ifindex;
73 #ifdef CONFIG_NET_CLS_ROUTE
74 	__u32		r_tclassid;
75 #endif
76 	char		r_ifname[IFNAMSIZ];
77 	int		r_dead;
78 };
79 
80 static struct fib_rule default_rule = {
81 	.r_clntref =	ATOMIC_INIT(2),
82 	.r_preference =	0x7FFF,
83 	.r_table =	RT_TABLE_DEFAULT,
84 	.r_action =	RTN_UNICAST,
85 };
86 
87 static struct fib_rule main_rule = {
88 	.r_next =	&default_rule,
89 	.r_clntref =	ATOMIC_INIT(2),
90 	.r_preference =	0x7FFE,
91 	.r_table =	RT_TABLE_MAIN,
92 	.r_action =	RTN_UNICAST,
93 };
94 
95 static struct fib_rule local_rule = {
96 	.r_next =	&main_rule,
97 	.r_clntref =	ATOMIC_INIT(2),
98 	.r_table =	RT_TABLE_LOCAL,
99 	.r_action =	RTN_UNICAST,
100 };
101 
102 static struct fib_rule *fib_rules = &local_rule;
103 static DEFINE_RWLOCK(fib_rules_lock);
104 
105 int inet_rtm_delrule(struct sk_buff *skb, struct nlmsghdr* nlh, void *arg)
106 {
107 	struct rtattr **rta = arg;
108 	struct rtmsg *rtm = NLMSG_DATA(nlh);
109 	struct fib_rule *r, **rp;
110 	int err = -ESRCH;
111 
112 	for (rp=&fib_rules; (r=*rp) != NULL; rp=&r->r_next) {
113 		if ((!rta[RTA_SRC-1] || memcmp(RTA_DATA(rta[RTA_SRC-1]), &r->r_src, 4) == 0) &&
114 		    rtm->rtm_src_len == r->r_src_len &&
115 		    rtm->rtm_dst_len == r->r_dst_len &&
116 		    (!rta[RTA_DST-1] || memcmp(RTA_DATA(rta[RTA_DST-1]), &r->r_dst, 4) == 0) &&
117 		    rtm->rtm_tos == r->r_tos &&
118 #ifdef CONFIG_IP_ROUTE_FWMARK
119 		    (!rta[RTA_PROTOINFO-1] || memcmp(RTA_DATA(rta[RTA_PROTOINFO-1]), &r->r_fwmark, 4) == 0) &&
120 #endif
121 		    (!rtm->rtm_type || rtm->rtm_type == r->r_action) &&
122 		    (!rta[RTA_PRIORITY-1] || memcmp(RTA_DATA(rta[RTA_PRIORITY-1]), &r->r_preference, 4) == 0) &&
123 		    (!rta[RTA_IIF-1] || rtattr_strcmp(rta[RTA_IIF-1], r->r_ifname) == 0) &&
124 		    (!rtm->rtm_table || (r && rtm->rtm_table == r->r_table))) {
125 			err = -EPERM;
126 			if (r == &local_rule)
127 				break;
128 
129 			write_lock_bh(&fib_rules_lock);
130 			*rp = r->r_next;
131 			r->r_dead = 1;
132 			write_unlock_bh(&fib_rules_lock);
133 			fib_rule_put(r);
134 			err = 0;
135 			break;
136 		}
137 	}
138 	return err;
139 }
140 
141 /* Allocate new unique table id */
142 
143 static struct fib_table *fib_empty_table(void)
144 {
145 	int id;
146 
147 	for (id = 1; id <= RT_TABLE_MAX; id++)
148 		if (fib_tables[id] == NULL)
149 			return __fib_new_table(id);
150 	return NULL;
151 }
152 
153 void fib_rule_put(struct fib_rule *r)
154 {
155 	if (atomic_dec_and_test(&r->r_clntref)) {
156 		if (r->r_dead)
157 			kfree(r);
158 		else
159 			printk("Freeing alive rule %p\n", r);
160 	}
161 }
162 
163 int inet_rtm_newrule(struct sk_buff *skb, struct nlmsghdr* nlh, void *arg)
164 {
165 	struct rtattr **rta = arg;
166 	struct rtmsg *rtm = NLMSG_DATA(nlh);
167 	struct fib_rule *r, *new_r, **rp;
168 	unsigned char table_id;
169 
170 	if (rtm->rtm_src_len > 32 || rtm->rtm_dst_len > 32 ||
171 	    (rtm->rtm_tos & ~IPTOS_TOS_MASK))
172 		return -EINVAL;
173 
174 	if (rta[RTA_IIF-1] && RTA_PAYLOAD(rta[RTA_IIF-1]) > IFNAMSIZ)
175 		return -EINVAL;
176 
177 	table_id = rtm->rtm_table;
178 	if (table_id == RT_TABLE_UNSPEC) {
179 		struct fib_table *table;
180 		if (rtm->rtm_type == RTN_UNICAST) {
181 			if ((table = fib_empty_table()) == NULL)
182 				return -ENOBUFS;
183 			table_id = table->tb_id;
184 		}
185 	}
186 
187 	new_r = kmalloc(sizeof(*new_r), GFP_KERNEL);
188 	if (!new_r)
189 		return -ENOMEM;
190 	memset(new_r, 0, sizeof(*new_r));
191 	if (rta[RTA_SRC-1])
192 		memcpy(&new_r->r_src, RTA_DATA(rta[RTA_SRC-1]), 4);
193 	if (rta[RTA_DST-1])
194 		memcpy(&new_r->r_dst, RTA_DATA(rta[RTA_DST-1]), 4);
195 	if (rta[RTA_GATEWAY-1])
196 		memcpy(&new_r->r_srcmap, RTA_DATA(rta[RTA_GATEWAY-1]), 4);
197 	new_r->r_src_len = rtm->rtm_src_len;
198 	new_r->r_dst_len = rtm->rtm_dst_len;
199 	new_r->r_srcmask = inet_make_mask(rtm->rtm_src_len);
200 	new_r->r_dstmask = inet_make_mask(rtm->rtm_dst_len);
201 	new_r->r_tos = rtm->rtm_tos;
202 #ifdef CONFIG_IP_ROUTE_FWMARK
203 	if (rta[RTA_PROTOINFO-1])
204 		memcpy(&new_r->r_fwmark, RTA_DATA(rta[RTA_PROTOINFO-1]), 4);
205 #endif
206 	new_r->r_action = rtm->rtm_type;
207 	new_r->r_flags = rtm->rtm_flags;
208 	if (rta[RTA_PRIORITY-1])
209 		memcpy(&new_r->r_preference, RTA_DATA(rta[RTA_PRIORITY-1]), 4);
210 	new_r->r_table = table_id;
211 	if (rta[RTA_IIF-1]) {
212 		struct net_device *dev;
213 		rtattr_strlcpy(new_r->r_ifname, rta[RTA_IIF-1], IFNAMSIZ);
214 		new_r->r_ifindex = -1;
215 		dev = __dev_get_by_name(new_r->r_ifname);
216 		if (dev)
217 			new_r->r_ifindex = dev->ifindex;
218 	}
219 #ifdef CONFIG_NET_CLS_ROUTE
220 	if (rta[RTA_FLOW-1])
221 		memcpy(&new_r->r_tclassid, RTA_DATA(rta[RTA_FLOW-1]), 4);
222 #endif
223 
224 	rp = &fib_rules;
225 	if (!new_r->r_preference) {
226 		r = fib_rules;
227 		if (r && (r = r->r_next) != NULL) {
228 			rp = &fib_rules->r_next;
229 			if (r->r_preference)
230 				new_r->r_preference = r->r_preference - 1;
231 		}
232 	}
233 
234 	while ( (r = *rp) != NULL ) {
235 		if (r->r_preference > new_r->r_preference)
236 			break;
237 		rp = &r->r_next;
238 	}
239 
240 	new_r->r_next = r;
241 	atomic_inc(&new_r->r_clntref);
242 	write_lock_bh(&fib_rules_lock);
243 	*rp = new_r;
244 	write_unlock_bh(&fib_rules_lock);
245 	return 0;
246 }
247 
248 #ifdef CONFIG_NET_CLS_ROUTE
249 u32 fib_rules_tclass(struct fib_result *res)
250 {
251 	if (res->r)
252 		return res->r->r_tclassid;
253 	return 0;
254 }
255 #endif
256 
257 
258 static void fib_rules_detach(struct net_device *dev)
259 {
260 	struct fib_rule *r;
261 
262 	for (r=fib_rules; r; r=r->r_next) {
263 		if (r->r_ifindex == dev->ifindex) {
264 			write_lock_bh(&fib_rules_lock);
265 			r->r_ifindex = -1;
266 			write_unlock_bh(&fib_rules_lock);
267 		}
268 	}
269 }
270 
271 static void fib_rules_attach(struct net_device *dev)
272 {
273 	struct fib_rule *r;
274 
275 	for (r=fib_rules; r; r=r->r_next) {
276 		if (r->r_ifindex == -1 && strcmp(dev->name, r->r_ifname) == 0) {
277 			write_lock_bh(&fib_rules_lock);
278 			r->r_ifindex = dev->ifindex;
279 			write_unlock_bh(&fib_rules_lock);
280 		}
281 	}
282 }
283 
284 int fib_lookup(const struct flowi *flp, struct fib_result *res)
285 {
286 	int err;
287 	struct fib_rule *r, *policy;
288 	struct fib_table *tb;
289 
290 	u32 daddr = flp->fl4_dst;
291 	u32 saddr = flp->fl4_src;
292 
293 FRprintk("Lookup: %u.%u.%u.%u <- %u.%u.%u.%u ",
294 	NIPQUAD(flp->fl4_dst), NIPQUAD(flp->fl4_src));
295 	read_lock(&fib_rules_lock);
296 	for (r = fib_rules; r; r=r->r_next) {
297 		if (((saddr^r->r_src) & r->r_srcmask) ||
298 		    ((daddr^r->r_dst) & r->r_dstmask) ||
299 		    (r->r_tos && r->r_tos != flp->fl4_tos) ||
300 #ifdef CONFIG_IP_ROUTE_FWMARK
301 		    (r->r_fwmark && r->r_fwmark != flp->fl4_fwmark) ||
302 #endif
303 		    (r->r_ifindex && r->r_ifindex != flp->iif))
304 			continue;
305 
306 FRprintk("tb %d r %d ", r->r_table, r->r_action);
307 		switch (r->r_action) {
308 		case RTN_UNICAST:
309 			policy = r;
310 			break;
311 		case RTN_UNREACHABLE:
312 			read_unlock(&fib_rules_lock);
313 			return -ENETUNREACH;
314 		default:
315 		case RTN_BLACKHOLE:
316 			read_unlock(&fib_rules_lock);
317 			return -EINVAL;
318 		case RTN_PROHIBIT:
319 			read_unlock(&fib_rules_lock);
320 			return -EACCES;
321 		}
322 
323 		if ((tb = fib_get_table(r->r_table)) == NULL)
324 			continue;
325 		err = tb->tb_lookup(tb, flp, res);
326 		if (err == 0) {
327 			res->r = policy;
328 			if (policy)
329 				atomic_inc(&policy->r_clntref);
330 			read_unlock(&fib_rules_lock);
331 			return 0;
332 		}
333 		if (err < 0 && err != -EAGAIN) {
334 			read_unlock(&fib_rules_lock);
335 			return err;
336 		}
337 	}
338 FRprintk("FAILURE\n");
339 	read_unlock(&fib_rules_lock);
340 	return -ENETUNREACH;
341 }
342 
343 void fib_select_default(const struct flowi *flp, struct fib_result *res)
344 {
345 	if (res->r && res->r->r_action == RTN_UNICAST &&
346 	    FIB_RES_GW(*res) && FIB_RES_NH(*res).nh_scope == RT_SCOPE_LINK) {
347 		struct fib_table *tb;
348 		if ((tb = fib_get_table(res->r->r_table)) != NULL)
349 			tb->tb_select_default(tb, flp, res);
350 	}
351 }
352 
353 static int fib_rules_event(struct notifier_block *this, unsigned long event, void *ptr)
354 {
355 	struct net_device *dev = ptr;
356 
357 	if (event == NETDEV_UNREGISTER)
358 		fib_rules_detach(dev);
359 	else if (event == NETDEV_REGISTER)
360 		fib_rules_attach(dev);
361 	return NOTIFY_DONE;
362 }
363 
364 
365 static struct notifier_block fib_rules_notifier = {
366 	.notifier_call =fib_rules_event,
367 };
368 
369 static __inline__ int inet_fill_rule(struct sk_buff *skb,
370 				     struct fib_rule *r,
371 				     struct netlink_callback *cb,
372 				     unsigned int flags)
373 {
374 	struct rtmsg *rtm;
375 	struct nlmsghdr  *nlh;
376 	unsigned char	 *b = skb->tail;
377 
378 	nlh = NLMSG_NEW_ANSWER(skb, cb, RTM_NEWRULE, sizeof(*rtm), flags);
379 	rtm = NLMSG_DATA(nlh);
380 	rtm->rtm_family = AF_INET;
381 	rtm->rtm_dst_len = r->r_dst_len;
382 	rtm->rtm_src_len = r->r_src_len;
383 	rtm->rtm_tos = r->r_tos;
384 #ifdef CONFIG_IP_ROUTE_FWMARK
385 	if (r->r_fwmark)
386 		RTA_PUT(skb, RTA_PROTOINFO, 4, &r->r_fwmark);
387 #endif
388 	rtm->rtm_table = r->r_table;
389 	rtm->rtm_protocol = 0;
390 	rtm->rtm_scope = 0;
391 	rtm->rtm_type = r->r_action;
392 	rtm->rtm_flags = r->r_flags;
393 
394 	if (r->r_dst_len)
395 		RTA_PUT(skb, RTA_DST, 4, &r->r_dst);
396 	if (r->r_src_len)
397 		RTA_PUT(skb, RTA_SRC, 4, &r->r_src);
398 	if (r->r_ifname[0])
399 		RTA_PUT(skb, RTA_IIF, IFNAMSIZ, &r->r_ifname);
400 	if (r->r_preference)
401 		RTA_PUT(skb, RTA_PRIORITY, 4, &r->r_preference);
402 	if (r->r_srcmap)
403 		RTA_PUT(skb, RTA_GATEWAY, 4, &r->r_srcmap);
404 #ifdef CONFIG_NET_CLS_ROUTE
405 	if (r->r_tclassid)
406 		RTA_PUT(skb, RTA_FLOW, 4, &r->r_tclassid);
407 #endif
408 	nlh->nlmsg_len = skb->tail - b;
409 	return skb->len;
410 
411 nlmsg_failure:
412 rtattr_failure:
413 	skb_trim(skb, b - skb->data);
414 	return -1;
415 }
416 
417 int inet_dump_rules(struct sk_buff *skb, struct netlink_callback *cb)
418 {
419 	int idx;
420 	int s_idx = cb->args[0];
421 	struct fib_rule *r;
422 
423 	read_lock(&fib_rules_lock);
424 	for (r=fib_rules, idx=0; r; r = r->r_next, idx++) {
425 		if (idx < s_idx)
426 			continue;
427 		if (inet_fill_rule(skb, r, cb, NLM_F_MULTI) < 0)
428 			break;
429 	}
430 	read_unlock(&fib_rules_lock);
431 	cb->args[0] = idx;
432 
433 	return skb->len;
434 }
435 
436 void __init fib_rules_init(void)
437 {
438 	register_netdevice_notifier(&fib_rules_notifier);
439 }
440