tjh.dev
Linux kernel mirror (for testing)
git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git
kernel
os
linux
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/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
52struct 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
79static 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
86static 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
94static 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
101static struct fib_rule *fib_rules = &local_rule;
102static DEFINE_RWLOCK(fib_rules_lock);
103
104int 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] || rtattr_strcmp(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
142static 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
152void 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
162int 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) {
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 rtattr_strlcpy(new_r->r_ifname, rta[RTA_IIF-1], IFNAMSIZ);
213 new_r->r_ifindex = -1;
214 dev = __dev_get_by_name(new_r->r_ifname);
215 if (dev)
216 new_r->r_ifindex = dev->ifindex;
217 }
218#ifdef CONFIG_NET_CLS_ROUTE
219 if (rta[RTA_FLOW-1])
220 memcpy(&new_r->r_tclassid, RTA_DATA(rta[RTA_FLOW-1]), 4);
221#endif
222
223 rp = &fib_rules;
224 if (!new_r->r_preference) {
225 r = fib_rules;
226 if (r && (r = r->r_next) != NULL) {
227 rp = &fib_rules->r_next;
228 if (r->r_preference)
229 new_r->r_preference = r->r_preference - 1;
230 }
231 }
232
233 while ( (r = *rp) != NULL ) {
234 if (r->r_preference > new_r->r_preference)
235 break;
236 rp = &r->r_next;
237 }
238
239 new_r->r_next = r;
240 atomic_inc(&new_r->r_clntref);
241 write_lock_bh(&fib_rules_lock);
242 *rp = new_r;
243 write_unlock_bh(&fib_rules_lock);
244 return 0;
245}
246
247#ifdef CONFIG_NET_CLS_ROUTE
248u32 fib_rules_tclass(struct fib_result *res)
249{
250 if (res->r)
251 return res->r->r_tclassid;
252 return 0;
253}
254#endif
255
256
257static void fib_rules_detach(struct net_device *dev)
258{
259 struct fib_rule *r;
260
261 for (r=fib_rules; r; r=r->r_next) {
262 if (r->r_ifindex == dev->ifindex) {
263 write_lock_bh(&fib_rules_lock);
264 r->r_ifindex = -1;
265 write_unlock_bh(&fib_rules_lock);
266 }
267 }
268}
269
270static void fib_rules_attach(struct net_device *dev)
271{
272 struct fib_rule *r;
273
274 for (r=fib_rules; r; r=r->r_next) {
275 if (r->r_ifindex == -1 && strcmp(dev->name, r->r_ifname) == 0) {
276 write_lock_bh(&fib_rules_lock);
277 r->r_ifindex = dev->ifindex;
278 write_unlock_bh(&fib_rules_lock);
279 }
280 }
281}
282
283int fib_lookup(const struct flowi *flp, struct fib_result *res)
284{
285 int err;
286 struct fib_rule *r, *policy;
287 struct fib_table *tb;
288
289 u32 daddr = flp->fl4_dst;
290 u32 saddr = flp->fl4_src;
291
292FRprintk("Lookup: %u.%u.%u.%u <- %u.%u.%u.%u ",
293 NIPQUAD(flp->fl4_dst), NIPQUAD(flp->fl4_src));
294 read_lock(&fib_rules_lock);
295 for (r = fib_rules; r; r=r->r_next) {
296 if (((saddr^r->r_src) & r->r_srcmask) ||
297 ((daddr^r->r_dst) & r->r_dstmask) ||
298 (r->r_tos && r->r_tos != flp->fl4_tos) ||
299#ifdef CONFIG_IP_ROUTE_FWMARK
300 (r->r_fwmark && r->r_fwmark != flp->fl4_fwmark) ||
301#endif
302 (r->r_ifindex && r->r_ifindex != flp->iif))
303 continue;
304
305FRprintk("tb %d r %d ", r->r_table, r->r_action);
306 switch (r->r_action) {
307 case RTN_UNICAST:
308 policy = r;
309 break;
310 case RTN_UNREACHABLE:
311 read_unlock(&fib_rules_lock);
312 return -ENETUNREACH;
313 default:
314 case RTN_BLACKHOLE:
315 read_unlock(&fib_rules_lock);
316 return -EINVAL;
317 case RTN_PROHIBIT:
318 read_unlock(&fib_rules_lock);
319 return -EACCES;
320 }
321
322 if ((tb = fib_get_table(r->r_table)) == NULL)
323 continue;
324 err = tb->tb_lookup(tb, flp, res);
325 if (err == 0) {
326 res->r = policy;
327 if (policy)
328 atomic_inc(&policy->r_clntref);
329 read_unlock(&fib_rules_lock);
330 return 0;
331 }
332 if (err < 0 && err != -EAGAIN) {
333 read_unlock(&fib_rules_lock);
334 return err;
335 }
336 }
337FRprintk("FAILURE\n");
338 read_unlock(&fib_rules_lock);
339 return -ENETUNREACH;
340}
341
342void fib_select_default(const struct flowi *flp, struct fib_result *res)
343{
344 if (res->r && res->r->r_action == RTN_UNICAST &&
345 FIB_RES_GW(*res) && FIB_RES_NH(*res).nh_scope == RT_SCOPE_LINK) {
346 struct fib_table *tb;
347 if ((tb = fib_get_table(res->r->r_table)) != NULL)
348 tb->tb_select_default(tb, flp, res);
349 }
350}
351
352static int fib_rules_event(struct notifier_block *this, unsigned long event, void *ptr)
353{
354 struct net_device *dev = ptr;
355
356 if (event == NETDEV_UNREGISTER)
357 fib_rules_detach(dev);
358 else if (event == NETDEV_REGISTER)
359 fib_rules_attach(dev);
360 return NOTIFY_DONE;
361}
362
363
364static struct notifier_block fib_rules_notifier = {
365 .notifier_call =fib_rules_event,
366};
367
368static __inline__ int inet_fill_rule(struct sk_buff *skb,
369 struct fib_rule *r,
370 struct netlink_callback *cb)
371{
372 struct rtmsg *rtm;
373 struct nlmsghdr *nlh;
374 unsigned char *b = skb->tail;
375
376 nlh = NLMSG_PUT(skb, NETLINK_CREDS(cb->skb)->pid, cb->nlh->nlmsg_seq, RTM_NEWRULE, sizeof(*rtm));
377 rtm = NLMSG_DATA(nlh);
378 rtm->rtm_family = AF_INET;
379 rtm->rtm_dst_len = r->r_dst_len;
380 rtm->rtm_src_len = r->r_src_len;
381 rtm->rtm_tos = r->r_tos;
382#ifdef CONFIG_IP_ROUTE_FWMARK
383 if (r->r_fwmark)
384 RTA_PUT(skb, RTA_PROTOINFO, 4, &r->r_fwmark);
385#endif
386 rtm->rtm_table = r->r_table;
387 rtm->rtm_protocol = 0;
388 rtm->rtm_scope = 0;
389 rtm->rtm_type = r->r_action;
390 rtm->rtm_flags = r->r_flags;
391
392 if (r->r_dst_len)
393 RTA_PUT(skb, RTA_DST, 4, &r->r_dst);
394 if (r->r_src_len)
395 RTA_PUT(skb, RTA_SRC, 4, &r->r_src);
396 if (r->r_ifname[0])
397 RTA_PUT(skb, RTA_IIF, IFNAMSIZ, &r->r_ifname);
398 if (r->r_preference)
399 RTA_PUT(skb, RTA_PRIORITY, 4, &r->r_preference);
400 if (r->r_srcmap)
401 RTA_PUT(skb, RTA_GATEWAY, 4, &r->r_srcmap);
402#ifdef CONFIG_NET_CLS_ROUTE
403 if (r->r_tclassid)
404 RTA_PUT(skb, RTA_FLOW, 4, &r->r_tclassid);
405#endif
406 nlh->nlmsg_len = skb->tail - b;
407 return skb->len;
408
409nlmsg_failure:
410rtattr_failure:
411 skb_trim(skb, b - skb->data);
412 return -1;
413}
414
415int inet_dump_rules(struct sk_buff *skb, struct netlink_callback *cb)
416{
417 int idx;
418 int s_idx = cb->args[0];
419 struct fib_rule *r;
420
421 read_lock(&fib_rules_lock);
422 for (r=fib_rules, idx=0; r; r = r->r_next, idx++) {
423 if (idx < s_idx)
424 continue;
425 if (inet_fill_rule(skb, r, cb) < 0)
426 break;
427 }
428 read_unlock(&fib_rules_lock);
429 cb->args[0] = idx;
430
431 return skb->len;
432}
433
434void __init fib_rules_init(void)
435{
436 register_netdevice_notifier(&fib_rules_notifier);
437}