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 <asm/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/netdevice.h>
37 #include <linux/if_arp.h>
38 #include <linux/proc_fs.h>
39 #include <linux/skbuff.h>
40 #include <linux/netlink.h>
41 #include <linux/init.h>
42 
43 #include <net/ip.h>
44 #include <net/protocol.h>
45 #include <net/route.h>
46 #include <net/tcp.h>
47 #include <net/sock.h>
48 #include <net/ip_fib.h>
49 
50 #define FRprintk(a...)
51 
52 struct fib_rule
53 {
54 	struct fib_rule *r_next;
55 	atomic_t	r_clntref;
56 	u32		r_preference;
57 	unsigned char	r_table;
58 	unsigned char	r_action;
59 	unsigned char	r_dst_len;
60 	unsigned char	r_src_len;
61 	u32		r_src;
62 	u32		r_srcmask;
63 	u32		r_dst;
64 	u32		r_dstmask;
65 	u32		r_srcmap;
66 	u8		r_flags;
67 	u8		r_tos;
68 #ifdef CONFIG_IP_ROUTE_FWMARK
69 	u32		r_fwmark;
70 #endif
71 	int		r_ifindex;
72 #ifdef CONFIG_NET_CLS_ROUTE
73 	__u32		r_tclassid;
74 #endif
75 	char		r_ifname[IFNAMSIZ];
76 	int		r_dead;
77 };
78 
79 static struct fib_rule default_rule = {
80 	r_clntref:	ATOMIC_INIT(2),
81 	r_preference:	0x7FFF,
82 	r_table:	RT_TABLE_DEFAULT,
83 	r_action:	RTN_UNICAST,
84 };
85 
86 static struct fib_rule main_rule = {
87 	r_next:		&default_rule,
88 	r_clntref:	ATOMIC_INIT(2),
89 	r_preference:	0x7FFE,
90 	r_table:	RT_TABLE_MAIN,
91 	r_action:	RTN_UNICAST,
92 };
93 
94 static struct fib_rule local_rule = {
95 	r_next:		&main_rule,
96 	r_clntref:	ATOMIC_INIT(2),
97 	r_table:	RT_TABLE_LOCAL,
98 	r_action:	RTN_UNICAST,
99 };
100 
101 static struct fib_rule *fib_rules = &local_rule;
102 static rwlock_t fib_rules_lock = RW_LOCK_UNLOCKED;
103 
inet_rtm_delrule(struct sk_buff * skb,struct nlmsghdr * nlh,void * arg)104 int inet_rtm_delrule(struct sk_buff *skb, struct nlmsghdr* nlh, void *arg)
105 {
106 	struct rtattr **rta = arg;
107 	struct rtmsg *rtm = NLMSG_DATA(nlh);
108 	struct fib_rule *r, **rp;
109 	int err = -ESRCH;
110 
111 	for (rp=&fib_rules; (r=*rp) != NULL; rp=&r->r_next) {
112 		if ((!rta[RTA_SRC-1] || memcmp(RTA_DATA(rta[RTA_SRC-1]), &r->r_src, 4) == 0) &&
113 		    rtm->rtm_src_len == r->r_src_len &&
114 		    rtm->rtm_dst_len == r->r_dst_len &&
115 		    (!rta[RTA_DST-1] || memcmp(RTA_DATA(rta[RTA_DST-1]), &r->r_dst, 4) == 0) &&
116 		    rtm->rtm_tos == r->r_tos &&
117 #ifdef CONFIG_IP_ROUTE_FWMARK
118 		    (!rta[RTA_PROTOINFO-1] || memcmp(RTA_DATA(rta[RTA_PROTOINFO-1]), &r->r_fwmark, 4) == 0) &&
119 #endif
120 		    (!rtm->rtm_type || rtm->rtm_type == r->r_action) &&
121 		    (!rta[RTA_PRIORITY-1] || memcmp(RTA_DATA(rta[RTA_PRIORITY-1]), &r->r_preference, 4) == 0) &&
122 		    (!rta[RTA_IIF-1] || strcmp(RTA_DATA(rta[RTA_IIF-1]), r->r_ifname) == 0) &&
123 		    (!rtm->rtm_table || (r && rtm->rtm_table == r->r_table))) {
124 			err = -EPERM;
125 			if (r == &local_rule)
126 				break;
127 
128 			write_lock_bh(&fib_rules_lock);
129 			*rp = r->r_next;
130 			r->r_dead = 1;
131 			write_unlock_bh(&fib_rules_lock);
132 			fib_rule_put(r);
133 			err = 0;
134 			break;
135 		}
136 	}
137 	return err;
138 }
139 
140 /* Allocate new unique table id */
141 
fib_empty_table(void)142 static struct fib_table *fib_empty_table(void)
143 {
144 	int id;
145 
146 	for (id = 1; id <= RT_TABLE_MAX; id++)
147 		if (fib_tables[id] == NULL)
148 			return __fib_new_table(id);
149 	return NULL;
150 }
151 
fib_rule_put(struct fib_rule * r)152 void fib_rule_put(struct fib_rule *r)
153 {
154 	if (atomic_dec_and_test(&r->r_clntref)) {
155 		if (r->r_dead)
156 			kfree(r);
157 		else
158 			printk("Freeing alive rule %p\n", r);
159 	}
160 }
161 
inet_rtm_newrule(struct sk_buff * skb,struct nlmsghdr * nlh,void * arg)162 int inet_rtm_newrule(struct sk_buff *skb, struct nlmsghdr* nlh, void *arg)
163 {
164 	struct rtattr **rta = arg;
165 	struct rtmsg *rtm = NLMSG_DATA(nlh);
166 	struct fib_rule *r, *new_r, **rp;
167 	unsigned char table_id;
168 
169 	if (rtm->rtm_src_len > 32 || rtm->rtm_dst_len > 32 ||
170 	    (rtm->rtm_tos & ~IPTOS_TOS_MASK))
171 		return -EINVAL;
172 
173 	if (rta[RTA_IIF-1] && RTA_PAYLOAD(rta[RTA_IIF-1]) > IFNAMSIZ)
174 		return -EINVAL;
175 
176 	table_id = rtm->rtm_table;
177 	if (table_id == RT_TABLE_UNSPEC) {
178 		struct fib_table *table;
179 		if (rtm->rtm_type == RTN_UNICAST || rtm->rtm_type == RTN_NAT) {
180 			if ((table = fib_empty_table()) == NULL)
181 				return -ENOBUFS;
182 			table_id = table->tb_id;
183 		}
184 	}
185 
186 	new_r = kmalloc(sizeof(*new_r), GFP_KERNEL);
187 	if (!new_r)
188 		return -ENOMEM;
189 	memset(new_r, 0, sizeof(*new_r));
190 	if (rta[RTA_SRC-1])
191 		memcpy(&new_r->r_src, RTA_DATA(rta[RTA_SRC-1]), 4);
192 	if (rta[RTA_DST-1])
193 		memcpy(&new_r->r_dst, RTA_DATA(rta[RTA_DST-1]), 4);
194 	if (rta[RTA_GATEWAY-1])
195 		memcpy(&new_r->r_srcmap, RTA_DATA(rta[RTA_GATEWAY-1]), 4);
196 	new_r->r_src_len = rtm->rtm_src_len;
197 	new_r->r_dst_len = rtm->rtm_dst_len;
198 	new_r->r_srcmask = inet_make_mask(rtm->rtm_src_len);
199 	new_r->r_dstmask = inet_make_mask(rtm->rtm_dst_len);
200 	new_r->r_tos = rtm->rtm_tos;
201 #ifdef CONFIG_IP_ROUTE_FWMARK
202 	if (rta[RTA_PROTOINFO-1])
203 		memcpy(&new_r->r_fwmark, RTA_DATA(rta[RTA_PROTOINFO-1]), 4);
204 #endif
205 	new_r->r_action = rtm->rtm_type;
206 	new_r->r_flags = rtm->rtm_flags;
207 	if (rta[RTA_PRIORITY-1])
208 		memcpy(&new_r->r_preference, RTA_DATA(rta[RTA_PRIORITY-1]), 4);
209 	new_r->r_table = table_id;
210 	if (rta[RTA_IIF-1]) {
211 		struct net_device *dev;
212 		memcpy(new_r->r_ifname, RTA_DATA(rta[RTA_IIF-1]), IFNAMSIZ);
213 		new_r->r_ifname[IFNAMSIZ-1] = 0;
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 
fib_rules_map_destination(u32 daddr,struct fib_result * res)248 u32 fib_rules_map_destination(u32 daddr, struct fib_result *res)
249 {
250 	u32 mask = inet_make_mask(res->prefixlen);
251 	return (daddr&~mask)|res->fi->fib_nh->nh_gw;
252 }
253 
fib_rules_policy(u32 saddr,struct fib_result * res,unsigned * flags)254 u32 fib_rules_policy(u32 saddr, struct fib_result *res, unsigned *flags)
255 {
256 	struct fib_rule *r = res->r;
257 
258 	if (r->r_action == RTN_NAT) {
259 		int addrtype = inet_addr_type(r->r_srcmap);
260 
261 		if (addrtype == RTN_NAT) {
262 			/* Packet is from  translated source; remember it */
263 			saddr = (saddr&~r->r_srcmask)|r->r_srcmap;
264 			*flags |= RTCF_SNAT;
265 		} else if (addrtype == RTN_LOCAL || r->r_srcmap == 0) {
266 			/* Packet is from masqueraded source; remember it */
267 			saddr = r->r_srcmap;
268 			*flags |= RTCF_MASQ;
269 		}
270 	}
271 	return saddr;
272 }
273 
274 #ifdef CONFIG_NET_CLS_ROUTE
fib_rules_tclass(struct fib_result * res)275 u32 fib_rules_tclass(struct fib_result *res)
276 {
277 	if (res->r)
278 		return res->r->r_tclassid;
279 	return 0;
280 }
281 #endif
282 
283 
fib_rules_detach(struct net_device * dev)284 static void fib_rules_detach(struct net_device *dev)
285 {
286 	struct fib_rule *r;
287 
288 	for (r=fib_rules; r; r=r->r_next) {
289 		if (r->r_ifindex == dev->ifindex) {
290 			write_lock_bh(&fib_rules_lock);
291 			r->r_ifindex = -1;
292 			write_unlock_bh(&fib_rules_lock);
293 		}
294 	}
295 }
296 
fib_rules_attach(struct net_device * dev)297 static void fib_rules_attach(struct net_device *dev)
298 {
299 	struct fib_rule *r;
300 
301 	for (r=fib_rules; r; r=r->r_next) {
302 		if (r->r_ifindex == -1 && strcmp(dev->name, r->r_ifname) == 0) {
303 			write_lock_bh(&fib_rules_lock);
304 			r->r_ifindex = dev->ifindex;
305 			write_unlock_bh(&fib_rules_lock);
306 		}
307 	}
308 }
309 
fib_lookup(const struct rt_key * key,struct fib_result * res)310 int fib_lookup(const struct rt_key *key, struct fib_result *res)
311 {
312 	int err;
313 	struct fib_rule *r, *policy;
314 	struct fib_table *tb;
315 
316 	u32 daddr = key->dst;
317 	u32 saddr = key->src;
318 
319 FRprintk("Lookup: %u.%u.%u.%u <- %u.%u.%u.%u ",
320 	NIPQUAD(key->dst), NIPQUAD(key->src));
321 	read_lock(&fib_rules_lock);
322 	for (r = fib_rules; r; r=r->r_next) {
323 		if (((saddr^r->r_src) & r->r_srcmask) ||
324 		    ((daddr^r->r_dst) & r->r_dstmask) ||
325 #ifdef CONFIG_IP_ROUTE_TOS
326 		    (r->r_tos && r->r_tos != key->tos) ||
327 #endif
328 #ifdef CONFIG_IP_ROUTE_FWMARK
329 		    (r->r_fwmark && r->r_fwmark != key->fwmark) ||
330 #endif
331 		    (r->r_ifindex && r->r_ifindex != key->iif))
332 			continue;
333 
334 FRprintk("tb %d r %d ", r->r_table, r->r_action);
335 		switch (r->r_action) {
336 		case RTN_UNICAST:
337 		case RTN_NAT:
338 			policy = r;
339 			break;
340 		case RTN_UNREACHABLE:
341 			read_unlock(&fib_rules_lock);
342 			return -ENETUNREACH;
343 		default:
344 		case RTN_BLACKHOLE:
345 			read_unlock(&fib_rules_lock);
346 			return -EINVAL;
347 		case RTN_PROHIBIT:
348 			read_unlock(&fib_rules_lock);
349 			return -EACCES;
350 		}
351 
352 		if ((tb = fib_get_table(r->r_table)) == NULL)
353 			continue;
354 		err = tb->tb_lookup(tb, key, res);
355 		if (err == 0) {
356 			res->r = policy;
357 			if (policy)
358 				atomic_inc(&policy->r_clntref);
359 			read_unlock(&fib_rules_lock);
360 			return 0;
361 		}
362 		if (err < 0 && err != -EAGAIN) {
363 			read_unlock(&fib_rules_lock);
364 			return err;
365 		}
366 	}
367 FRprintk("FAILURE\n");
368 	read_unlock(&fib_rules_lock);
369 	return -ENETUNREACH;
370 }
371 
fib_select_default(const struct rt_key * key,struct fib_result * res)372 void fib_select_default(const struct rt_key *key, struct fib_result *res)
373 {
374 	if (res->r && res->r->r_action == RTN_UNICAST &&
375 	    FIB_RES_GW(*res) && FIB_RES_NH(*res).nh_scope == RT_SCOPE_LINK) {
376 		struct fib_table *tb;
377 		if ((tb = fib_get_table(res->r->r_table)) != NULL)
378 			tb->tb_select_default(tb, key, res);
379 	}
380 }
381 
fib_rules_event(struct notifier_block * this,unsigned long event,void * ptr)382 static int fib_rules_event(struct notifier_block *this, unsigned long event, void *ptr)
383 {
384 	struct net_device *dev = ptr;
385 
386 	if (event == NETDEV_UNREGISTER)
387 		fib_rules_detach(dev);
388 	else if (event == NETDEV_REGISTER)
389 		fib_rules_attach(dev);
390 	return NOTIFY_DONE;
391 }
392 
393 
394 struct notifier_block fib_rules_notifier = {
395 	notifier_call:	fib_rules_event,
396 };
397 
inet_fill_rule(struct sk_buff * skb,struct fib_rule * r,struct netlink_callback * cb)398 static __inline__ int inet_fill_rule(struct sk_buff *skb,
399 				     struct fib_rule *r,
400 				     struct netlink_callback *cb)
401 {
402 	struct rtmsg *rtm;
403 	struct nlmsghdr  *nlh;
404 	unsigned char	 *b = skb->tail;
405 
406 	nlh = NLMSG_PUT(skb, NETLINK_CREDS(cb->skb)->pid, cb->nlh->nlmsg_seq, RTM_NEWRULE, sizeof(*rtm));
407 	rtm = NLMSG_DATA(nlh);
408 	rtm->rtm_family = AF_INET;
409 	rtm->rtm_dst_len = r->r_dst_len;
410 	rtm->rtm_src_len = r->r_src_len;
411 	rtm->rtm_tos = r->r_tos;
412 #ifdef CONFIG_IP_ROUTE_FWMARK
413 	if (r->r_fwmark)
414 		RTA_PUT(skb, RTA_PROTOINFO, 4, &r->r_fwmark);
415 #endif
416 	rtm->rtm_table = r->r_table;
417 	rtm->rtm_protocol = 0;
418 	rtm->rtm_scope = 0;
419 	rtm->rtm_type = r->r_action;
420 	rtm->rtm_flags = r->r_flags;
421 
422 	if (r->r_dst_len)
423 		RTA_PUT(skb, RTA_DST, 4, &r->r_dst);
424 	if (r->r_src_len)
425 		RTA_PUT(skb, RTA_SRC, 4, &r->r_src);
426 	if (r->r_ifname[0])
427 		RTA_PUT(skb, RTA_IIF, IFNAMSIZ, &r->r_ifname);
428 	if (r->r_preference)
429 		RTA_PUT(skb, RTA_PRIORITY, 4, &r->r_preference);
430 	if (r->r_srcmap)
431 		RTA_PUT(skb, RTA_GATEWAY, 4, &r->r_srcmap);
432 #ifdef CONFIG_NET_CLS_ROUTE
433 	if (r->r_tclassid)
434 		RTA_PUT(skb, RTA_FLOW, 4, &r->r_tclassid);
435 #endif
436 	nlh->nlmsg_len = skb->tail - b;
437 	return skb->len;
438 
439 nlmsg_failure:
440 rtattr_failure:
441 	skb_trim(skb, b - skb->data);
442 	return -1;
443 }
444 
inet_dump_rules(struct sk_buff * skb,struct netlink_callback * cb)445 int inet_dump_rules(struct sk_buff *skb, struct netlink_callback *cb)
446 {
447 	int idx;
448 	int s_idx = cb->args[0];
449 	struct fib_rule *r;
450 
451 	read_lock(&fib_rules_lock);
452 	for (r=fib_rules, idx=0; r; r = r->r_next, idx++) {
453 		if (idx < s_idx)
454 			continue;
455 		if (inet_fill_rule(skb, r, cb) < 0)
456 			break;
457 	}
458 	read_unlock(&fib_rules_lock);
459 	cb->args[0] = idx;
460 
461 	return skb->len;
462 }
463 
fib_rules_init(void)464 void __init fib_rules_init(void)
465 {
466 	register_netdevice_notifier(&fib_rules_notifier);
467 }
468