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.
6 * IPv4 Forwarding Information Base: policy rules.
8 * Version: $Id: fib_rules.c,v 1.17 2001/10/31 21:55:54 davem Exp $
10 * Authors: Alexey Kuznetsov, <kuznet@ms2.inr.ac.ru>
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.
18 * Rani Assaf : local_rule cannot be deleted
19 * Marc Boucher : routing by fwmark
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>
30 #include <linux/string.h>
31 #include <linux/socket.h>
32 #include <linux/sockios.h>
33 #include <linux/errno.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>
44 #include <net/protocol.h>
45 #include <net/route.h>
48 #include <net/ip_fib.h>
50 #define FRprintk(a...)
54 struct fib_rule *r_next;
57 unsigned char r_table;
58 unsigned char r_action;
59 unsigned char r_dst_len;
60 unsigned char r_src_len;
68 #ifdef CONFIG_IP_ROUTE_FWMARK
72 #ifdef CONFIG_NET_CLS_ROUTE
75 char r_ifname[IFNAMSIZ];
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,
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,
94 static struct fib_rule local_rule = {
96 .r_clntref = ATOMIC_INIT(2),
97 .r_table = RT_TABLE_LOCAL,
98 .r_action = RTN_UNICAST,
101 static struct fib_rule *fib_rules = &local_rule;
102 static rwlock_t fib_rules_lock = RW_LOCK_UNLOCKED;
104 int inet_rtm_delrule(struct sk_buff *skb, struct nlmsghdr* nlh, void *arg)
106 struct rtattr **rta = arg;
107 struct rtmsg *rtm = NLMSG_DATA(nlh);
108 struct fib_rule *r, **rp;
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) &&
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))) {
125 if (r == &local_rule)
128 write_lock_bh(&fib_rules_lock);
131 write_unlock_bh(&fib_rules_lock);
140 /* Allocate new unique table id */
142 static struct fib_table *fib_empty_table(void)
146 for (id = 1; id <= RT_TABLE_MAX; id++)
147 if (fib_tables[id] == NULL)
148 return __fib_new_table(id);
152 void fib_rule_put(struct fib_rule *r)
154 if (atomic_dec_and_test(&r->r_clntref)) {
158 printk("Freeing alive rule %p\n", r);
162 int inet_rtm_newrule(struct sk_buff *skb, struct nlmsghdr* nlh, void *arg)
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;
169 if (rtm->rtm_src_len > 32 || rtm->rtm_dst_len > 32 ||
170 (rtm->rtm_tos & ~IPTOS_TOS_MASK))
173 if (rta[RTA_IIF-1] && RTA_PAYLOAD(rta[RTA_IIF-1]) > IFNAMSIZ)
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)
182 table_id = table->tb_id;
186 new_r = kmalloc(sizeof(*new_r), GFP_KERNEL);
189 memset(new_r, 0, sizeof(*new_r));
191 memcpy(&new_r->r_src, RTA_DATA(rta[RTA_SRC-1]), 4);
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);
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);
217 new_r->r_ifindex = dev->ifindex;
219 #ifdef CONFIG_NET_CLS_ROUTE
221 memcpy(&new_r->r_tclassid, RTA_DATA(rta[RTA_FLOW-1]), 4);
225 if (!new_r->r_preference) {
227 if (r && (r = r->r_next) != NULL) {
228 rp = &fib_rules->r_next;
230 new_r->r_preference = r->r_preference - 1;
234 while ( (r = *rp) != NULL ) {
235 if (r->r_preference > new_r->r_preference)
241 atomic_inc(&new_r->r_clntref);
242 write_lock_bh(&fib_rules_lock);
244 write_unlock_bh(&fib_rules_lock);
248 u32 fib_rules_map_destination(u32 daddr, struct fib_result *res)
250 u32 mask = inet_make_mask(res->prefixlen);
251 return (daddr&~mask)|res->fi->fib_nh->nh_gw;
254 u32 fib_rules_policy(u32 saddr, struct fib_result *res, unsigned *flags)
256 struct fib_rule *r = res->r;
258 if (r->r_action == RTN_NAT) {
259 int addrtype = inet_addr_type(r->r_srcmap);
261 if (addrtype == RTN_NAT) {
262 /* Packet is from translated source; remember it */
263 saddr = (saddr&~r->r_srcmask)|r->r_srcmap;
265 } else if (addrtype == RTN_LOCAL || r->r_srcmap == 0) {
266 /* Packet is from masqueraded source; remember it */
274 #ifdef CONFIG_NET_CLS_ROUTE
275 u32 fib_rules_tclass(struct fib_result *res)
278 return res->r->r_tclassid;
284 static void fib_rules_detach(struct net_device *dev)
288 for (r=fib_rules; r; r=r->r_next) {
289 if (r->r_ifindex == dev->ifindex) {
290 write_lock_bh(&fib_rules_lock);
292 write_unlock_bh(&fib_rules_lock);
297 static void fib_rules_attach(struct net_device *dev)
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);
310 int fib_lookup(const struct flowi *flp, struct fib_result *res)
313 struct fib_rule *r, *policy;
314 struct fib_table *tb;
316 u32 daddr = flp->fl4_dst;
317 u32 saddr = flp->fl4_src;
319 FRprintk("Lookup: %u.%u.%u.%u <- %u.%u.%u.%u ",
320 NIPQUAD(flp->fl4_dst), NIPQUAD(flp->fl4_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 != flp->fl4_tos) ||
328 #ifdef CONFIG_IP_ROUTE_FWMARK
329 (r->r_fwmark && r->r_fwmark != flp->fl4_fwmark) ||
331 (r->r_ifindex && r->r_ifindex != flp->iif))
334 FRprintk("tb %d r %d ", r->r_table, r->r_action);
335 switch (r->r_action) {
340 case RTN_UNREACHABLE:
341 read_unlock(&fib_rules_lock);
345 read_unlock(&fib_rules_lock);
348 read_unlock(&fib_rules_lock);
352 if ((tb = fib_get_table(r->r_table)) == NULL)
354 err = tb->tb_lookup(tb, flp, res);
358 atomic_inc(&policy->r_clntref);
359 read_unlock(&fib_rules_lock);
362 if (err < 0 && err != -EAGAIN) {
363 read_unlock(&fib_rules_lock);
367 FRprintk("FAILURE\n");
368 read_unlock(&fib_rules_lock);
372 void fib_select_default(const struct flowi *flp, struct fib_result *res)
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, flp, res);
382 static int fib_rules_event(struct notifier_block *this, unsigned long event, void *ptr)
384 struct net_device *dev = ptr;
386 if (event == NETDEV_UNREGISTER)
387 fib_rules_detach(dev);
388 else if (event == NETDEV_REGISTER)
389 fib_rules_attach(dev);
394 struct notifier_block fib_rules_notifier = {
395 .notifier_call =fib_rules_event,
398 static __inline__ int inet_fill_rule(struct sk_buff *skb,
400 struct netlink_callback *cb)
403 struct nlmsghdr *nlh;
404 unsigned char *b = skb->tail;
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
414 RTA_PUT(skb, RTA_PROTOINFO, 4, &r->r_fwmark);
416 rtm->rtm_table = r->r_table;
417 rtm->rtm_protocol = 0;
419 rtm->rtm_type = r->r_action;
420 rtm->rtm_flags = r->r_flags;
423 RTA_PUT(skb, RTA_DST, 4, &r->r_dst);
425 RTA_PUT(skb, RTA_SRC, 4, &r->r_src);
427 RTA_PUT(skb, RTA_IIF, IFNAMSIZ, &r->r_ifname);
429 RTA_PUT(skb, RTA_PRIORITY, 4, &r->r_preference);
431 RTA_PUT(skb, RTA_GATEWAY, 4, &r->r_srcmap);
432 #ifdef CONFIG_NET_CLS_ROUTE
434 RTA_PUT(skb, RTA_FLOW, 4, &r->r_tclassid);
436 nlh->nlmsg_len = skb->tail - b;
441 skb_trim(skb, b - skb->data);
445 int inet_dump_rules(struct sk_buff *skb, struct netlink_callback *cb)
448 int s_idx = cb->args[0];
451 read_lock(&fib_rules_lock);
452 for (r=fib_rules, idx=0; r; r = r->r_next, idx++) {
455 if (inet_fill_rule(skb, r, cb) < 0)
458 read_unlock(&fib_rules_lock);
464 void __init fib_rules_init(void)
466 register_netdevice_notifier(&fib_rules_notifier);