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 <asm/uaccess.h>
23 #include <asm/system.h>
24 #include <linux/bitops.h>
25 #include <linux/types.h>
26 #include <linux/kernel.h>
27 #include <linux/sched.h>
29 #include <linux/string.h>
30 #include <linux/socket.h>
31 #include <linux/sockios.h>
32 #include <linux/errno.h>
34 #include <linux/inet.h>
35 #include <linux/inetdevice.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 #include <linux/list.h>
43 #include <linux/rcupdate.h>
46 #include <net/protocol.h>
47 #include <net/route.h>
50 #include <net/ip_fib.h>
52 #define FRprintk(a...)
56 struct hlist_node hlist;
59 unsigned char r_table;
60 unsigned char r_action;
61 unsigned char r_dst_len;
62 unsigned char r_src_len;
70 #ifdef CONFIG_IP_ROUTE_FWMARK
74 #ifdef CONFIG_NET_CLS_ROUTE
77 char r_ifname[IFNAMSIZ];
82 static struct fib_rule default_rule = {
83 .r_clntref = ATOMIC_INIT(2),
84 .r_preference = 0x7FFF,
85 .r_table = RT_TABLE_DEFAULT,
86 .r_action = RTN_UNICAST,
89 static struct fib_rule main_rule = {
90 .r_clntref = ATOMIC_INIT(2),
91 .r_preference = 0x7FFE,
92 .r_table = RT_TABLE_MAIN,
93 .r_action = RTN_UNICAST,
96 static struct fib_rule local_rule = {
97 .r_clntref = ATOMIC_INIT(2),
98 .r_table = RT_TABLE_LOCAL,
99 .r_action = RTN_UNICAST,
102 static struct hlist_head fib_rules;
104 /* writer func called from netlink -- rtnl_sem hold*/
106 static void rtmsg_rule(int, struct fib_rule *);
108 int inet_rtm_delrule(struct sk_buff *skb, struct nlmsghdr* nlh, void *arg)
110 struct rtattr **rta = arg;
111 struct rtmsg *rtm = NLMSG_DATA(nlh);
113 struct hlist_node *node;
116 hlist_for_each_entry(r, node, &fib_rules, hlist) {
117 if ((!rta[RTA_SRC-1] || memcmp(RTA_DATA(rta[RTA_SRC-1]), &r->r_src, 4) == 0) &&
118 rtm->rtm_src_len == r->r_src_len &&
119 rtm->rtm_dst_len == r->r_dst_len &&
120 (!rta[RTA_DST-1] || memcmp(RTA_DATA(rta[RTA_DST-1]), &r->r_dst, 4) == 0) &&
121 rtm->rtm_tos == r->r_tos &&
122 #ifdef CONFIG_IP_ROUTE_FWMARK
123 (!rta[RTA_PROTOINFO-1] || memcmp(RTA_DATA(rta[RTA_PROTOINFO-1]), &r->r_fwmark, 4) == 0) &&
125 (!rtm->rtm_type || rtm->rtm_type == r->r_action) &&
126 (!rta[RTA_PRIORITY-1] || memcmp(RTA_DATA(rta[RTA_PRIORITY-1]), &r->r_preference, 4) == 0) &&
127 (!rta[RTA_IIF-1] || rtattr_strcmp(rta[RTA_IIF-1], r->r_ifname) == 0) &&
128 (!rtm->rtm_table || (r && rtm->rtm_table == r->r_table))) {
130 if (r == &local_rule)
133 hlist_del_rcu(&r->hlist);
135 rtmsg_rule(RTM_DELRULE, r);
144 /* Allocate new unique table id */
146 static struct fib_table *fib_empty_table(void)
150 for (id = 1; id <= RT_TABLE_MAX; id++)
151 if (fib_tables[id] == NULL)
152 return __fib_new_table(id);
156 static inline void fib_rule_put_rcu(struct rcu_head *head)
158 struct fib_rule *r = container_of(head, struct fib_rule, rcu);
162 void fib_rule_put(struct fib_rule *r)
164 if (atomic_dec_and_test(&r->r_clntref)) {
166 call_rcu(&r->rcu, fib_rule_put_rcu);
168 printk("Freeing alive rule %p\n", r);
172 /* writer func called from netlink -- rtnl_sem hold*/
174 int inet_rtm_newrule(struct sk_buff *skb, struct nlmsghdr* nlh, void *arg)
176 struct rtattr **rta = arg;
177 struct rtmsg *rtm = NLMSG_DATA(nlh);
178 struct fib_rule *r, *new_r, *last = NULL;
179 struct hlist_node *node = NULL;
180 unsigned char table_id;
182 if (rtm->rtm_src_len > 32 || rtm->rtm_dst_len > 32 ||
183 (rtm->rtm_tos & ~IPTOS_TOS_MASK))
186 if (rta[RTA_IIF-1] && RTA_PAYLOAD(rta[RTA_IIF-1]) > IFNAMSIZ)
189 table_id = rtm->rtm_table;
190 if (table_id == RT_TABLE_UNSPEC) {
191 struct fib_table *table;
192 if (rtm->rtm_type == RTN_UNICAST) {
193 if ((table = fib_empty_table()) == NULL)
195 table_id = table->tb_id;
199 new_r = kzalloc(sizeof(*new_r), GFP_KERNEL);
204 memcpy(&new_r->r_src, RTA_DATA(rta[RTA_SRC-1]), 4);
206 memcpy(&new_r->r_dst, RTA_DATA(rta[RTA_DST-1]), 4);
207 if (rta[RTA_GATEWAY-1])
208 memcpy(&new_r->r_srcmap, RTA_DATA(rta[RTA_GATEWAY-1]), 4);
209 new_r->r_src_len = rtm->rtm_src_len;
210 new_r->r_dst_len = rtm->rtm_dst_len;
211 new_r->r_srcmask = inet_make_mask(rtm->rtm_src_len);
212 new_r->r_dstmask = inet_make_mask(rtm->rtm_dst_len);
213 new_r->r_tos = rtm->rtm_tos;
214 #ifdef CONFIG_IP_ROUTE_FWMARK
215 if (rta[RTA_PROTOINFO-1])
216 memcpy(&new_r->r_fwmark, RTA_DATA(rta[RTA_PROTOINFO-1]), 4);
218 new_r->r_action = rtm->rtm_type;
219 new_r->r_flags = rtm->rtm_flags;
220 if (rta[RTA_PRIORITY-1])
221 memcpy(&new_r->r_preference, RTA_DATA(rta[RTA_PRIORITY-1]), 4);
222 new_r->r_table = table_id;
223 if (rta[RTA_IIF-1]) {
224 struct net_device *dev;
225 rtattr_strlcpy(new_r->r_ifname, rta[RTA_IIF-1], IFNAMSIZ);
226 new_r->r_ifindex = -1;
227 dev = __dev_get_by_name(new_r->r_ifname);
229 new_r->r_ifindex = dev->ifindex;
231 #ifdef CONFIG_NET_CLS_ROUTE
233 memcpy(&new_r->r_tclassid, RTA_DATA(rta[RTA_FLOW-1]), 4);
235 r = container_of(fib_rules.first, struct fib_rule, hlist);
237 if (!new_r->r_preference) {
238 if (r && r->hlist.next != NULL) {
239 r = container_of(r->hlist.next, struct fib_rule, hlist);
241 new_r->r_preference = r->r_preference - 1;
245 hlist_for_each_entry(r, node, &fib_rules, hlist) {
246 if (r->r_preference > new_r->r_preference)
250 atomic_inc(&new_r->r_clntref);
253 hlist_add_after_rcu(&last->hlist, &new_r->hlist);
255 hlist_add_before_rcu(&new_r->hlist, &r->hlist);
257 rtmsg_rule(RTM_NEWRULE, new_r);
261 #ifdef CONFIG_NET_CLS_ROUTE
262 u32 fib_rules_tclass(struct fib_result *res)
265 return res->r->r_tclassid;
270 /* callers should hold rtnl semaphore */
272 static void fib_rules_detach(struct net_device *dev)
274 struct hlist_node *node;
277 hlist_for_each_entry(r, node, &fib_rules, hlist) {
278 if (r->r_ifindex == dev->ifindex)
284 /* callers should hold rtnl semaphore */
286 static void fib_rules_attach(struct net_device *dev)
288 struct hlist_node *node;
291 hlist_for_each_entry(r, node, &fib_rules, hlist) {
292 if (r->r_ifindex == -1 && strcmp(dev->name, r->r_ifname) == 0)
293 r->r_ifindex = dev->ifindex;
297 int fib_lookup(const struct flowi *flp, struct fib_result *res)
300 struct fib_rule *r, *policy;
301 struct fib_table *tb;
302 struct hlist_node *node;
304 u32 daddr = flp->fl4_dst;
305 u32 saddr = flp->fl4_src;
307 FRprintk("Lookup: %u.%u.%u.%u <- %u.%u.%u.%u ",
308 NIPQUAD(flp->fl4_dst), NIPQUAD(flp->fl4_src));
312 hlist_for_each_entry_rcu(r, node, &fib_rules, hlist) {
313 if (((saddr^r->r_src) & r->r_srcmask) ||
314 ((daddr^r->r_dst) & r->r_dstmask) ||
315 (r->r_tos && r->r_tos != flp->fl4_tos) ||
316 #ifdef CONFIG_IP_ROUTE_FWMARK
317 (r->r_fwmark && r->r_fwmark != flp->fl4_fwmark) ||
319 (r->r_ifindex && r->r_ifindex != flp->iif))
322 FRprintk("tb %d r %d ", r->r_table, r->r_action);
323 switch (r->r_action) {
327 case RTN_UNREACHABLE:
339 if ((tb = fib_get_table(r->r_table)) == NULL)
341 err = tb->tb_lookup(tb, flp, res);
345 atomic_inc(&policy->r_clntref);
349 if (err < 0 && err != -EAGAIN) {
354 FRprintk("FAILURE\n");
359 void fib_select_default(const struct flowi *flp, struct fib_result *res)
361 if (res->r && res->r->r_action == RTN_UNICAST &&
362 FIB_RES_GW(*res) && FIB_RES_NH(*res).nh_scope == RT_SCOPE_LINK) {
363 struct fib_table *tb;
364 if ((tb = fib_get_table(res->r->r_table)) != NULL)
365 tb->tb_select_default(tb, flp, res);
369 static int fib_rules_event(struct notifier_block *this, unsigned long event, void *ptr)
371 struct net_device *dev = ptr;
373 if (event == NETDEV_UNREGISTER)
374 fib_rules_detach(dev);
375 else if (event == NETDEV_REGISTER)
376 fib_rules_attach(dev);
381 static struct notifier_block fib_rules_notifier = {
382 .notifier_call =fib_rules_event,
385 static __inline__ int inet_fill_rule(struct sk_buff *skb,
387 u32 pid, u32 seq, int event,
391 struct nlmsghdr *nlh;
392 unsigned char *b = skb->tail;
394 nlh = NLMSG_NEW(skb, pid, seq, event, sizeof(*rtm), flags);
395 rtm = NLMSG_DATA(nlh);
396 rtm->rtm_family = AF_INET;
397 rtm->rtm_dst_len = r->r_dst_len;
398 rtm->rtm_src_len = r->r_src_len;
399 rtm->rtm_tos = r->r_tos;
400 #ifdef CONFIG_IP_ROUTE_FWMARK
402 RTA_PUT(skb, RTA_PROTOINFO, 4, &r->r_fwmark);
404 rtm->rtm_table = r->r_table;
405 rtm->rtm_protocol = 0;
407 rtm->rtm_type = r->r_action;
408 rtm->rtm_flags = r->r_flags;
411 RTA_PUT(skb, RTA_DST, 4, &r->r_dst);
413 RTA_PUT(skb, RTA_SRC, 4, &r->r_src);
415 RTA_PUT(skb, RTA_IIF, IFNAMSIZ, &r->r_ifname);
417 RTA_PUT(skb, RTA_PRIORITY, 4, &r->r_preference);
419 RTA_PUT(skb, RTA_GATEWAY, 4, &r->r_srcmap);
420 #ifdef CONFIG_NET_CLS_ROUTE
422 RTA_PUT(skb, RTA_FLOW, 4, &r->r_tclassid);
424 nlh->nlmsg_len = skb->tail - b;
429 skb_trim(skb, b - skb->data);
433 /* callers should hold rtnl semaphore */
435 static void rtmsg_rule(int event, struct fib_rule *r)
437 int size = NLMSG_SPACE(sizeof(struct rtmsg) + 128);
438 struct sk_buff *skb = alloc_skb(size, GFP_KERNEL);
441 netlink_set_err(rtnl, 0, RTNLGRP_IPV4_RULE, ENOBUFS);
442 else if (inet_fill_rule(skb, r, 0, 0, event, 0) < 0) {
444 netlink_set_err(rtnl, 0, RTNLGRP_IPV4_RULE, EINVAL);
446 netlink_broadcast(rtnl, skb, 0, RTNLGRP_IPV4_RULE, GFP_KERNEL);
450 int inet_dump_rules(struct sk_buff *skb, struct netlink_callback *cb)
453 int s_idx = cb->args[0];
455 struct hlist_node *node;
458 hlist_for_each_entry(r, node, &fib_rules, hlist) {
461 if (inet_fill_rule(skb, r, NETLINK_CB(cb->skb).pid,
463 RTM_NEWRULE, NLM_F_MULTI) < 0)
474 void __init fib_rules_init(void)
476 INIT_HLIST_HEAD(&fib_rules);
477 hlist_add_head(&local_rule.hlist, &fib_rules);
478 hlist_add_after(&local_rule.hlist, &main_rule.hlist);
479 hlist_add_after(&main_rule.hlist, &default_rule.hlist);
480 register_netdevice_notifier(&fib_rules_notifier);