]>
Commit | Line | Data |
---|---|---|
1da177e4 LT |
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 | * ROUTE - implementation of the IP router. | |
7 | * | |
02c30a84 | 8 | * Authors: Ross Biro |
1da177e4 LT |
9 | * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG> |
10 | * Alan Cox, <gw4pts@gw4pts.ampr.org> | |
11 | * Linus Torvalds, <Linus.Torvalds@helsinki.fi> | |
12 | * Alexey Kuznetsov, <kuznet@ms2.inr.ac.ru> | |
13 | * | |
14 | * Fixes: | |
15 | * Alan Cox : Verify area fixes. | |
16 | * Alan Cox : cli() protects routing changes | |
17 | * Rui Oliveira : ICMP routing table updates | |
18 | * (rco@di.uminho.pt) Routing table insertion and update | |
19 | * Linus Torvalds : Rewrote bits to be sensible | |
20 | * Alan Cox : Added BSD route gw semantics | |
e905a9ed | 21 | * Alan Cox : Super /proc >4K |
1da177e4 LT |
22 | * Alan Cox : MTU in route table |
23 | * Alan Cox : MSS actually. Also added the window | |
24 | * clamper. | |
25 | * Sam Lantinga : Fixed route matching in rt_del() | |
26 | * Alan Cox : Routing cache support. | |
27 | * Alan Cox : Removed compatibility cruft. | |
28 | * Alan Cox : RTF_REJECT support. | |
29 | * Alan Cox : TCP irtt support. | |
30 | * Jonathan Naylor : Added Metric support. | |
31 | * Miquel van Smoorenburg : BSD API fixes. | |
32 | * Miquel van Smoorenburg : Metrics. | |
33 | * Alan Cox : Use __u32 properly | |
34 | * Alan Cox : Aligned routing errors more closely with BSD | |
35 | * our system is still very different. | |
36 | * Alan Cox : Faster /proc handling | |
37 | * Alexey Kuznetsov : Massive rework to support tree based routing, | |
38 | * routing caches and better behaviour. | |
e905a9ed | 39 | * |
1da177e4 LT |
40 | * Olaf Erb : irtt wasn't being copied right. |
41 | * Bjorn Ekwall : Kerneld route support. | |
42 | * Alan Cox : Multicast fixed (I hope) | |
43 | * Pavel Krauz : Limited broadcast fixed | |
44 | * Mike McLagan : Routing by source | |
45 | * Alexey Kuznetsov : End of old history. Split to fib.c and | |
46 | * route.c and rewritten from scratch. | |
47 | * Andi Kleen : Load-limit warning messages. | |
48 | * Vitaly E. Lavrov : Transparent proxy revived after year coma. | |
49 | * Vitaly E. Lavrov : Race condition in ip_route_input_slow. | |
50 | * Tobias Ringstrom : Uninitialized res.type in ip_route_output_slow. | |
51 | * Vladimir V. Ivanov : IP rule info (flowid) is really useful. | |
52 | * Marc Boucher : routing by fwmark | |
53 | * Robert Olsson : Added rt_cache statistics | |
54 | * Arnaldo C. Melo : Convert proc stuff to seq_file | |
bb1d23b0 | 55 | * Eric Dumazet : hashed spinlocks and rt_check_expire() fixes. |
cef2685e IS |
56 | * Ilia Sotnikov : Ignore TOS on PMTUD and Redirect |
57 | * Ilia Sotnikov : Removed TOS from hash calculations | |
1da177e4 LT |
58 | * |
59 | * This program is free software; you can redistribute it and/or | |
60 | * modify it under the terms of the GNU General Public License | |
61 | * as published by the Free Software Foundation; either version | |
62 | * 2 of the License, or (at your option) any later version. | |
63 | */ | |
64 | ||
afd46503 JP |
65 | #define pr_fmt(fmt) "IPv4: " fmt |
66 | ||
1da177e4 LT |
67 | #include <linux/module.h> |
68 | #include <asm/uaccess.h> | |
1da177e4 LT |
69 | #include <linux/bitops.h> |
70 | #include <linux/types.h> | |
71 | #include <linux/kernel.h> | |
1da177e4 | 72 | #include <linux/mm.h> |
424c4b70 | 73 | #include <linux/bootmem.h> |
1da177e4 LT |
74 | #include <linux/string.h> |
75 | #include <linux/socket.h> | |
76 | #include <linux/sockios.h> | |
77 | #include <linux/errno.h> | |
78 | #include <linux/in.h> | |
79 | #include <linux/inet.h> | |
80 | #include <linux/netdevice.h> | |
81 | #include <linux/proc_fs.h> | |
82 | #include <linux/init.h> | |
39c90ece | 83 | #include <linux/workqueue.h> |
1da177e4 | 84 | #include <linux/skbuff.h> |
1da177e4 LT |
85 | #include <linux/inetdevice.h> |
86 | #include <linux/igmp.h> | |
87 | #include <linux/pkt_sched.h> | |
88 | #include <linux/mroute.h> | |
89 | #include <linux/netfilter_ipv4.h> | |
90 | #include <linux/random.h> | |
91 | #include <linux/jhash.h> | |
92 | #include <linux/rcupdate.h> | |
93 | #include <linux/times.h> | |
5a0e3ad6 | 94 | #include <linux/slab.h> |
b9eda06f | 95 | #include <linux/prefetch.h> |
352e512c | 96 | #include <net/dst.h> |
457c4cbc | 97 | #include <net/net_namespace.h> |
1da177e4 LT |
98 | #include <net/protocol.h> |
99 | #include <net/ip.h> | |
100 | #include <net/route.h> | |
101 | #include <net/inetpeer.h> | |
102 | #include <net/sock.h> | |
103 | #include <net/ip_fib.h> | |
104 | #include <net/arp.h> | |
105 | #include <net/tcp.h> | |
106 | #include <net/icmp.h> | |
107 | #include <net/xfrm.h> | |
8d71740c | 108 | #include <net/netevent.h> |
63f3444f | 109 | #include <net/rtnetlink.h> |
1da177e4 LT |
110 | #ifdef CONFIG_SYSCTL |
111 | #include <linux/sysctl.h> | |
7426a564 | 112 | #include <linux/kmemleak.h> |
1da177e4 | 113 | #endif |
6e5714ea | 114 | #include <net/secure_seq.h> |
1da177e4 | 115 | |
68a5e3dd | 116 | #define RT_FL_TOS(oldflp4) \ |
f61759e6 | 117 | ((oldflp4)->flowi4_tos & (IPTOS_RT_MASK | RTO_ONLINK)) |
1da177e4 LT |
118 | |
119 | #define IP_MAX_MTU 0xFFF0 | |
120 | ||
121 | #define RT_GC_TIMEOUT (300*HZ) | |
122 | ||
1da177e4 | 123 | static int ip_rt_max_size; |
817bc4db | 124 | static int ip_rt_gc_timeout __read_mostly = RT_GC_TIMEOUT; |
9f28a2fc | 125 | static int ip_rt_gc_interval __read_mostly = 60 * HZ; |
817bc4db SH |
126 | static int ip_rt_gc_min_interval __read_mostly = HZ / 2; |
127 | static int ip_rt_redirect_number __read_mostly = 9; | |
128 | static int ip_rt_redirect_load __read_mostly = HZ / 50; | |
129 | static int ip_rt_redirect_silence __read_mostly = ((HZ / 50) << (9 + 1)); | |
130 | static int ip_rt_error_cost __read_mostly = HZ; | |
131 | static int ip_rt_error_burst __read_mostly = 5 * HZ; | |
132 | static int ip_rt_gc_elasticity __read_mostly = 8; | |
133 | static int ip_rt_mtu_expires __read_mostly = 10 * 60 * HZ; | |
134 | static int ip_rt_min_pmtu __read_mostly = 512 + 20 + 20; | |
135 | static int ip_rt_min_advmss __read_mostly = 256; | |
9f28a2fc | 136 | |
1da177e4 LT |
137 | /* |
138 | * Interface to generic destination cache. | |
139 | */ | |
140 | ||
141 | static struct dst_entry *ipv4_dst_check(struct dst_entry *dst, u32 cookie); | |
0dbaee3b | 142 | static unsigned int ipv4_default_advmss(const struct dst_entry *dst); |
ebb762f2 | 143 | static unsigned int ipv4_mtu(const struct dst_entry *dst); |
1da177e4 | 144 | static void ipv4_dst_destroy(struct dst_entry *dst); |
1da177e4 LT |
145 | static struct dst_entry *ipv4_negative_advice(struct dst_entry *dst); |
146 | static void ipv4_link_failure(struct sk_buff *skb); | |
6700c270 DM |
147 | static void ip_rt_update_pmtu(struct dst_entry *dst, struct sock *sk, |
148 | struct sk_buff *skb, u32 mtu); | |
149 | static void ip_do_redirect(struct dst_entry *dst, struct sock *sk, | |
150 | struct sk_buff *skb); | |
1da177e4 | 151 | |
72cdd1d9 ED |
152 | static void ipv4_dst_ifdown(struct dst_entry *dst, struct net_device *dev, |
153 | int how) | |
154 | { | |
155 | } | |
1da177e4 | 156 | |
62fa8a84 DM |
157 | static u32 *ipv4_cow_metrics(struct dst_entry *dst, unsigned long old) |
158 | { | |
31248731 DM |
159 | WARN_ON(1); |
160 | return NULL; | |
62fa8a84 DM |
161 | } |
162 | ||
f894cbf8 DM |
163 | static struct neighbour *ipv4_neigh_lookup(const struct dst_entry *dst, |
164 | struct sk_buff *skb, | |
165 | const void *daddr); | |
d3aaeb38 | 166 | |
1da177e4 LT |
167 | static struct dst_ops ipv4_dst_ops = { |
168 | .family = AF_INET, | |
09640e63 | 169 | .protocol = cpu_to_be16(ETH_P_IP), |
1da177e4 | 170 | .check = ipv4_dst_check, |
0dbaee3b | 171 | .default_advmss = ipv4_default_advmss, |
ebb762f2 | 172 | .mtu = ipv4_mtu, |
62fa8a84 | 173 | .cow_metrics = ipv4_cow_metrics, |
1da177e4 LT |
174 | .destroy = ipv4_dst_destroy, |
175 | .ifdown = ipv4_dst_ifdown, | |
176 | .negative_advice = ipv4_negative_advice, | |
177 | .link_failure = ipv4_link_failure, | |
178 | .update_pmtu = ip_rt_update_pmtu, | |
e47a185b | 179 | .redirect = ip_do_redirect, |
1ac06e03 | 180 | .local_out = __ip_local_out, |
d3aaeb38 | 181 | .neigh_lookup = ipv4_neigh_lookup, |
1da177e4 LT |
182 | }; |
183 | ||
184 | #define ECN_OR_COST(class) TC_PRIO_##class | |
185 | ||
4839c52b | 186 | const __u8 ip_tos2prio[16] = { |
1da177e4 | 187 | TC_PRIO_BESTEFFORT, |
4a2b9c37 | 188 | ECN_OR_COST(BESTEFFORT), |
1da177e4 LT |
189 | TC_PRIO_BESTEFFORT, |
190 | ECN_OR_COST(BESTEFFORT), | |
191 | TC_PRIO_BULK, | |
192 | ECN_OR_COST(BULK), | |
193 | TC_PRIO_BULK, | |
194 | ECN_OR_COST(BULK), | |
195 | TC_PRIO_INTERACTIVE, | |
196 | ECN_OR_COST(INTERACTIVE), | |
197 | TC_PRIO_INTERACTIVE, | |
198 | ECN_OR_COST(INTERACTIVE), | |
199 | TC_PRIO_INTERACTIVE_BULK, | |
200 | ECN_OR_COST(INTERACTIVE_BULK), | |
201 | TC_PRIO_INTERACTIVE_BULK, | |
202 | ECN_OR_COST(INTERACTIVE_BULK) | |
203 | }; | |
d4a96865 | 204 | EXPORT_SYMBOL(ip_tos2prio); |
1da177e4 | 205 | |
2f970d83 | 206 | static DEFINE_PER_CPU(struct rt_cache_stat, rt_cache_stat); |
27f39c73 | 207 | #define RT_CACHE_STAT_INC(field) __this_cpu_inc(rt_cache_stat.field) |
1da177e4 | 208 | |
e84f84f2 DL |
209 | static inline int rt_genid(struct net *net) |
210 | { | |
211 | return atomic_read(&net->ipv4.rt_genid); | |
212 | } | |
213 | ||
1da177e4 | 214 | #ifdef CONFIG_PROC_FS |
1da177e4 LT |
215 | static void *rt_cache_seq_start(struct seq_file *seq, loff_t *pos) |
216 | { | |
29e75252 | 217 | if (*pos) |
89aef892 | 218 | return NULL; |
29e75252 | 219 | return SEQ_START_TOKEN; |
1da177e4 LT |
220 | } |
221 | ||
222 | static void *rt_cache_seq_next(struct seq_file *seq, void *v, loff_t *pos) | |
223 | { | |
1da177e4 | 224 | ++*pos; |
89aef892 | 225 | return NULL; |
1da177e4 LT |
226 | } |
227 | ||
228 | static void rt_cache_seq_stop(struct seq_file *seq, void *v) | |
229 | { | |
1da177e4 LT |
230 | } |
231 | ||
232 | static int rt_cache_seq_show(struct seq_file *seq, void *v) | |
233 | { | |
234 | if (v == SEQ_START_TOKEN) | |
235 | seq_printf(seq, "%-127s\n", | |
236 | "Iface\tDestination\tGateway \tFlags\t\tRefCnt\tUse\t" | |
237 | "Metric\tSource\t\tMTU\tWindow\tIRTT\tTOS\tHHRef\t" | |
238 | "HHUptod\tSpecDst"); | |
e905a9ed | 239 | return 0; |
1da177e4 LT |
240 | } |
241 | ||
f690808e | 242 | static const struct seq_operations rt_cache_seq_ops = { |
1da177e4 LT |
243 | .start = rt_cache_seq_start, |
244 | .next = rt_cache_seq_next, | |
245 | .stop = rt_cache_seq_stop, | |
246 | .show = rt_cache_seq_show, | |
247 | }; | |
248 | ||
249 | static int rt_cache_seq_open(struct inode *inode, struct file *file) | |
250 | { | |
89aef892 | 251 | return seq_open(file, &rt_cache_seq_ops); |
1da177e4 LT |
252 | } |
253 | ||
9a32144e | 254 | static const struct file_operations rt_cache_seq_fops = { |
1da177e4 LT |
255 | .owner = THIS_MODULE, |
256 | .open = rt_cache_seq_open, | |
257 | .read = seq_read, | |
258 | .llseek = seq_lseek, | |
89aef892 | 259 | .release = seq_release, |
1da177e4 LT |
260 | }; |
261 | ||
262 | ||
263 | static void *rt_cpu_seq_start(struct seq_file *seq, loff_t *pos) | |
264 | { | |
265 | int cpu; | |
266 | ||
267 | if (*pos == 0) | |
268 | return SEQ_START_TOKEN; | |
269 | ||
0f23174a | 270 | for (cpu = *pos-1; cpu < nr_cpu_ids; ++cpu) { |
1da177e4 LT |
271 | if (!cpu_possible(cpu)) |
272 | continue; | |
273 | *pos = cpu+1; | |
2f970d83 | 274 | return &per_cpu(rt_cache_stat, cpu); |
1da177e4 LT |
275 | } |
276 | return NULL; | |
277 | } | |
278 | ||
279 | static void *rt_cpu_seq_next(struct seq_file *seq, void *v, loff_t *pos) | |
280 | { | |
281 | int cpu; | |
282 | ||
0f23174a | 283 | for (cpu = *pos; cpu < nr_cpu_ids; ++cpu) { |
1da177e4 LT |
284 | if (!cpu_possible(cpu)) |
285 | continue; | |
286 | *pos = cpu+1; | |
2f970d83 | 287 | return &per_cpu(rt_cache_stat, cpu); |
1da177e4 LT |
288 | } |
289 | return NULL; | |
e905a9ed | 290 | |
1da177e4 LT |
291 | } |
292 | ||
293 | static void rt_cpu_seq_stop(struct seq_file *seq, void *v) | |
294 | { | |
295 | ||
296 | } | |
297 | ||
298 | static int rt_cpu_seq_show(struct seq_file *seq, void *v) | |
299 | { | |
300 | struct rt_cache_stat *st = v; | |
301 | ||
302 | if (v == SEQ_START_TOKEN) { | |
5bec0039 | 303 | seq_printf(seq, "entries in_hit in_slow_tot in_slow_mc in_no_route in_brd in_martian_dst in_martian_src out_hit out_slow_tot out_slow_mc gc_total gc_ignored gc_goal_miss gc_dst_overflow in_hlist_search out_hlist_search\n"); |
1da177e4 LT |
304 | return 0; |
305 | } | |
e905a9ed | 306 | |
1da177e4 LT |
307 | seq_printf(seq,"%08x %08x %08x %08x %08x %08x %08x %08x " |
308 | " %08x %08x %08x %08x %08x %08x %08x %08x %08x \n", | |
fc66f95c | 309 | dst_entries_get_slow(&ipv4_dst_ops), |
1da177e4 LT |
310 | st->in_hit, |
311 | st->in_slow_tot, | |
312 | st->in_slow_mc, | |
313 | st->in_no_route, | |
314 | st->in_brd, | |
315 | st->in_martian_dst, | |
316 | st->in_martian_src, | |
317 | ||
318 | st->out_hit, | |
319 | st->out_slow_tot, | |
e905a9ed | 320 | st->out_slow_mc, |
1da177e4 LT |
321 | |
322 | st->gc_total, | |
323 | st->gc_ignored, | |
324 | st->gc_goal_miss, | |
325 | st->gc_dst_overflow, | |
326 | st->in_hlist_search, | |
327 | st->out_hlist_search | |
328 | ); | |
329 | return 0; | |
330 | } | |
331 | ||
f690808e | 332 | static const struct seq_operations rt_cpu_seq_ops = { |
1da177e4 LT |
333 | .start = rt_cpu_seq_start, |
334 | .next = rt_cpu_seq_next, | |
335 | .stop = rt_cpu_seq_stop, | |
336 | .show = rt_cpu_seq_show, | |
337 | }; | |
338 | ||
339 | ||
340 | static int rt_cpu_seq_open(struct inode *inode, struct file *file) | |
341 | { | |
342 | return seq_open(file, &rt_cpu_seq_ops); | |
343 | } | |
344 | ||
9a32144e | 345 | static const struct file_operations rt_cpu_seq_fops = { |
1da177e4 LT |
346 | .owner = THIS_MODULE, |
347 | .open = rt_cpu_seq_open, | |
348 | .read = seq_read, | |
349 | .llseek = seq_lseek, | |
350 | .release = seq_release, | |
351 | }; | |
352 | ||
c7066f70 | 353 | #ifdef CONFIG_IP_ROUTE_CLASSID |
a661c419 | 354 | static int rt_acct_proc_show(struct seq_file *m, void *v) |
78c686e9 | 355 | { |
a661c419 AD |
356 | struct ip_rt_acct *dst, *src; |
357 | unsigned int i, j; | |
358 | ||
359 | dst = kcalloc(256, sizeof(struct ip_rt_acct), GFP_KERNEL); | |
360 | if (!dst) | |
361 | return -ENOMEM; | |
362 | ||
363 | for_each_possible_cpu(i) { | |
364 | src = (struct ip_rt_acct *)per_cpu_ptr(ip_rt_acct, i); | |
365 | for (j = 0; j < 256; j++) { | |
366 | dst[j].o_bytes += src[j].o_bytes; | |
367 | dst[j].o_packets += src[j].o_packets; | |
368 | dst[j].i_bytes += src[j].i_bytes; | |
369 | dst[j].i_packets += src[j].i_packets; | |
370 | } | |
78c686e9 PE |
371 | } |
372 | ||
a661c419 AD |
373 | seq_write(m, dst, 256 * sizeof(struct ip_rt_acct)); |
374 | kfree(dst); | |
375 | return 0; | |
376 | } | |
78c686e9 | 377 | |
a661c419 AD |
378 | static int rt_acct_proc_open(struct inode *inode, struct file *file) |
379 | { | |
380 | return single_open(file, rt_acct_proc_show, NULL); | |
78c686e9 | 381 | } |
a661c419 AD |
382 | |
383 | static const struct file_operations rt_acct_proc_fops = { | |
384 | .owner = THIS_MODULE, | |
385 | .open = rt_acct_proc_open, | |
386 | .read = seq_read, | |
387 | .llseek = seq_lseek, | |
388 | .release = single_release, | |
389 | }; | |
78c686e9 | 390 | #endif |
107f1634 | 391 | |
73b38711 | 392 | static int __net_init ip_rt_do_proc_init(struct net *net) |
107f1634 PE |
393 | { |
394 | struct proc_dir_entry *pde; | |
395 | ||
396 | pde = proc_net_fops_create(net, "rt_cache", S_IRUGO, | |
397 | &rt_cache_seq_fops); | |
398 | if (!pde) | |
399 | goto err1; | |
400 | ||
77020720 WC |
401 | pde = proc_create("rt_cache", S_IRUGO, |
402 | net->proc_net_stat, &rt_cpu_seq_fops); | |
107f1634 PE |
403 | if (!pde) |
404 | goto err2; | |
405 | ||
c7066f70 | 406 | #ifdef CONFIG_IP_ROUTE_CLASSID |
a661c419 | 407 | pde = proc_create("rt_acct", 0, net->proc_net, &rt_acct_proc_fops); |
107f1634 PE |
408 | if (!pde) |
409 | goto err3; | |
410 | #endif | |
411 | return 0; | |
412 | ||
c7066f70 | 413 | #ifdef CONFIG_IP_ROUTE_CLASSID |
107f1634 PE |
414 | err3: |
415 | remove_proc_entry("rt_cache", net->proc_net_stat); | |
416 | #endif | |
417 | err2: | |
418 | remove_proc_entry("rt_cache", net->proc_net); | |
419 | err1: | |
420 | return -ENOMEM; | |
421 | } | |
73b38711 DL |
422 | |
423 | static void __net_exit ip_rt_do_proc_exit(struct net *net) | |
424 | { | |
425 | remove_proc_entry("rt_cache", net->proc_net_stat); | |
426 | remove_proc_entry("rt_cache", net->proc_net); | |
c7066f70 | 427 | #ifdef CONFIG_IP_ROUTE_CLASSID |
73b38711 | 428 | remove_proc_entry("rt_acct", net->proc_net); |
0a931acf | 429 | #endif |
73b38711 DL |
430 | } |
431 | ||
432 | static struct pernet_operations ip_rt_proc_ops __net_initdata = { | |
433 | .init = ip_rt_do_proc_init, | |
434 | .exit = ip_rt_do_proc_exit, | |
435 | }; | |
436 | ||
437 | static int __init ip_rt_proc_init(void) | |
438 | { | |
439 | return register_pernet_subsys(&ip_rt_proc_ops); | |
440 | } | |
441 | ||
107f1634 | 442 | #else |
73b38711 | 443 | static inline int ip_rt_proc_init(void) |
107f1634 PE |
444 | { |
445 | return 0; | |
446 | } | |
1da177e4 | 447 | #endif /* CONFIG_PROC_FS */ |
e905a9ed | 448 | |
e84f84f2 DL |
449 | static inline int rt_is_expired(struct rtable *rth) |
450 | { | |
d8d1f30b | 451 | return rth->rt_genid != rt_genid(dev_net(rth->dst.dev)); |
e84f84f2 DL |
452 | } |
453 | ||
29e75252 | 454 | /* |
25985edc | 455 | * Perturbation of rt_genid by a small quantity [1..256] |
29e75252 ED |
456 | * Using 8 bits of shuffling ensure we can call rt_cache_invalidate() |
457 | * many times (2^24) without giving recent rt_genid. | |
458 | * Jenkins hash is strong enough that litle changes of rt_genid are OK. | |
1da177e4 | 459 | */ |
86c657f6 | 460 | static void rt_cache_invalidate(struct net *net) |
1da177e4 | 461 | { |
29e75252 | 462 | unsigned char shuffle; |
1da177e4 | 463 | |
29e75252 | 464 | get_random_bytes(&shuffle, sizeof(shuffle)); |
e84f84f2 | 465 | atomic_add(shuffle + 1U, &net->ipv4.rt_genid); |
1da177e4 LT |
466 | } |
467 | ||
29e75252 ED |
468 | /* |
469 | * delay < 0 : invalidate cache (fast : entries will be deleted later) | |
470 | * delay >= 0 : invalidate & flush cache (can be long) | |
471 | */ | |
76e6ebfb | 472 | void rt_cache_flush(struct net *net, int delay) |
1da177e4 | 473 | { |
86c657f6 | 474 | rt_cache_invalidate(net); |
98376387 ED |
475 | } |
476 | ||
f894cbf8 DM |
477 | static struct neighbour *ipv4_neigh_lookup(const struct dst_entry *dst, |
478 | struct sk_buff *skb, | |
479 | const void *daddr) | |
3769cffb | 480 | { |
d3aaeb38 DM |
481 | struct net_device *dev = dst->dev; |
482 | const __be32 *pkey = daddr; | |
39232973 | 483 | const struct rtable *rt; |
3769cffb DM |
484 | struct neighbour *n; |
485 | ||
39232973 | 486 | rt = (const struct rtable *) dst; |
a263b309 | 487 | if (rt->rt_gateway) |
39232973 | 488 | pkey = (const __be32 *) &rt->rt_gateway; |
f894cbf8 DM |
489 | else if (skb) |
490 | pkey = &ip_hdr(skb)->daddr; | |
d3aaeb38 | 491 | |
80703d26 | 492 | n = __ipv4_neigh_lookup(dev, *(__force u32 *)pkey); |
d3aaeb38 DM |
493 | if (n) |
494 | return n; | |
32092ecf | 495 | return neigh_create(&arp_tbl, pkey, dev); |
d3aaeb38 DM |
496 | } |
497 | ||
1da177e4 LT |
498 | /* |
499 | * Peer allocation may fail only in serious out-of-memory conditions. However | |
500 | * we still can generate some output. | |
501 | * Random ID selection looks a bit dangerous because we have no chances to | |
502 | * select ID being unique in a reasonable period of time. | |
503 | * But broken packet identifier may be better than no packet at all. | |
504 | */ | |
505 | static void ip_select_fb_ident(struct iphdr *iph) | |
506 | { | |
507 | static DEFINE_SPINLOCK(ip_fb_id_lock); | |
508 | static u32 ip_fallback_id; | |
509 | u32 salt; | |
510 | ||
511 | spin_lock_bh(&ip_fb_id_lock); | |
e448515c | 512 | salt = secure_ip_id((__force __be32)ip_fallback_id ^ iph->daddr); |
1da177e4 LT |
513 | iph->id = htons(salt & 0xFFFF); |
514 | ip_fallback_id = salt; | |
515 | spin_unlock_bh(&ip_fb_id_lock); | |
516 | } | |
517 | ||
518 | void __ip_select_ident(struct iphdr *iph, struct dst_entry *dst, int more) | |
519 | { | |
1d861aa4 DM |
520 | struct net *net = dev_net(dst->dev); |
521 | struct inet_peer *peer; | |
1da177e4 | 522 | |
1d861aa4 DM |
523 | peer = inet_getpeer_v4(net->ipv4.peers, iph->daddr, 1); |
524 | if (peer) { | |
525 | iph->id = htons(inet_getid(peer, more)); | |
526 | inet_putpeer(peer); | |
527 | return; | |
528 | } | |
1da177e4 LT |
529 | |
530 | ip_select_fb_ident(iph); | |
531 | } | |
4bc2f18b | 532 | EXPORT_SYMBOL(__ip_select_ident); |
1da177e4 | 533 | |
5abf7f7e | 534 | static void __build_flow_key(struct flowi4 *fl4, const struct sock *sk, |
4895c771 DM |
535 | const struct iphdr *iph, |
536 | int oif, u8 tos, | |
537 | u8 prot, u32 mark, int flow_flags) | |
538 | { | |
539 | if (sk) { | |
540 | const struct inet_sock *inet = inet_sk(sk); | |
541 | ||
542 | oif = sk->sk_bound_dev_if; | |
543 | mark = sk->sk_mark; | |
544 | tos = RT_CONN_FLAGS(sk); | |
545 | prot = inet->hdrincl ? IPPROTO_RAW : sk->sk_protocol; | |
546 | } | |
547 | flowi4_init_output(fl4, oif, mark, tos, | |
548 | RT_SCOPE_UNIVERSE, prot, | |
549 | flow_flags, | |
550 | iph->daddr, iph->saddr, 0, 0); | |
551 | } | |
552 | ||
5abf7f7e ED |
553 | static void build_skb_flow_key(struct flowi4 *fl4, const struct sk_buff *skb, |
554 | const struct sock *sk) | |
4895c771 DM |
555 | { |
556 | const struct iphdr *iph = ip_hdr(skb); | |
557 | int oif = skb->dev->ifindex; | |
558 | u8 tos = RT_TOS(iph->tos); | |
559 | u8 prot = iph->protocol; | |
560 | u32 mark = skb->mark; | |
561 | ||
562 | __build_flow_key(fl4, sk, iph, oif, tos, prot, mark, 0); | |
563 | } | |
564 | ||
5abf7f7e | 565 | static void build_sk_flow_key(struct flowi4 *fl4, const struct sock *sk) |
4895c771 DM |
566 | { |
567 | const struct inet_sock *inet = inet_sk(sk); | |
5abf7f7e | 568 | const struct ip_options_rcu *inet_opt; |
4895c771 DM |
569 | __be32 daddr = inet->inet_daddr; |
570 | ||
571 | rcu_read_lock(); | |
572 | inet_opt = rcu_dereference(inet->inet_opt); | |
573 | if (inet_opt && inet_opt->opt.srr) | |
574 | daddr = inet_opt->opt.faddr; | |
575 | flowi4_init_output(fl4, sk->sk_bound_dev_if, sk->sk_mark, | |
576 | RT_CONN_FLAGS(sk), RT_SCOPE_UNIVERSE, | |
577 | inet->hdrincl ? IPPROTO_RAW : sk->sk_protocol, | |
578 | inet_sk_flowi_flags(sk), | |
579 | daddr, inet->inet_saddr, 0, 0); | |
580 | rcu_read_unlock(); | |
581 | } | |
582 | ||
5abf7f7e ED |
583 | static void ip_rt_build_flow_key(struct flowi4 *fl4, const struct sock *sk, |
584 | const struct sk_buff *skb) | |
4895c771 DM |
585 | { |
586 | if (skb) | |
587 | build_skb_flow_key(fl4, skb, sk); | |
588 | else | |
589 | build_sk_flow_key(fl4, sk); | |
590 | } | |
591 | ||
aee06da6 | 592 | static DEFINE_SEQLOCK(fnhe_seqlock); |
4895c771 | 593 | |
aee06da6 | 594 | static struct fib_nh_exception *fnhe_oldest(struct fnhe_hash_bucket *hash) |
4895c771 DM |
595 | { |
596 | struct fib_nh_exception *fnhe, *oldest; | |
597 | ||
598 | oldest = rcu_dereference(hash->chain); | |
599 | for (fnhe = rcu_dereference(oldest->fnhe_next); fnhe; | |
600 | fnhe = rcu_dereference(fnhe->fnhe_next)) { | |
601 | if (time_before(fnhe->fnhe_stamp, oldest->fnhe_stamp)) | |
602 | oldest = fnhe; | |
603 | } | |
604 | return oldest; | |
605 | } | |
606 | ||
d3a25c98 DM |
607 | static inline u32 fnhe_hashfun(__be32 daddr) |
608 | { | |
609 | u32 hval; | |
610 | ||
611 | hval = (__force u32) daddr; | |
612 | hval ^= (hval >> 11) ^ (hval >> 22); | |
613 | ||
614 | return hval & (FNHE_HASH_SIZE - 1); | |
615 | } | |
616 | ||
aee06da6 JA |
617 | static void update_or_create_fnhe(struct fib_nh *nh, __be32 daddr, __be32 gw, |
618 | u32 pmtu, unsigned long expires) | |
4895c771 | 619 | { |
aee06da6 | 620 | struct fnhe_hash_bucket *hash; |
4895c771 DM |
621 | struct fib_nh_exception *fnhe; |
622 | int depth; | |
aee06da6 JA |
623 | u32 hval = fnhe_hashfun(daddr); |
624 | ||
625 | write_seqlock_bh(&fnhe_seqlock); | |
4895c771 | 626 | |
aee06da6 | 627 | hash = nh->nh_exceptions; |
4895c771 | 628 | if (!hash) { |
aee06da6 | 629 | hash = kzalloc(FNHE_HASH_SIZE * sizeof(*hash), GFP_ATOMIC); |
4895c771 | 630 | if (!hash) |
aee06da6 JA |
631 | goto out_unlock; |
632 | nh->nh_exceptions = hash; | |
4895c771 DM |
633 | } |
634 | ||
4895c771 DM |
635 | hash += hval; |
636 | ||
637 | depth = 0; | |
638 | for (fnhe = rcu_dereference(hash->chain); fnhe; | |
639 | fnhe = rcu_dereference(fnhe->fnhe_next)) { | |
640 | if (fnhe->fnhe_daddr == daddr) | |
aee06da6 | 641 | break; |
4895c771 DM |
642 | depth++; |
643 | } | |
644 | ||
aee06da6 JA |
645 | if (fnhe) { |
646 | if (gw) | |
647 | fnhe->fnhe_gw = gw; | |
648 | if (pmtu) { | |
649 | fnhe->fnhe_pmtu = pmtu; | |
650 | fnhe->fnhe_expires = expires; | |
651 | } | |
652 | } else { | |
653 | if (depth > FNHE_RECLAIM_DEPTH) | |
654 | fnhe = fnhe_oldest(hash); | |
655 | else { | |
656 | fnhe = kzalloc(sizeof(*fnhe), GFP_ATOMIC); | |
657 | if (!fnhe) | |
658 | goto out_unlock; | |
659 | ||
660 | fnhe->fnhe_next = hash->chain; | |
661 | rcu_assign_pointer(hash->chain, fnhe); | |
662 | } | |
663 | fnhe->fnhe_daddr = daddr; | |
664 | fnhe->fnhe_gw = gw; | |
665 | fnhe->fnhe_pmtu = pmtu; | |
666 | fnhe->fnhe_expires = expires; | |
4895c771 | 667 | } |
4895c771 | 668 | |
4895c771 | 669 | fnhe->fnhe_stamp = jiffies; |
aee06da6 JA |
670 | |
671 | out_unlock: | |
672 | write_sequnlock_bh(&fnhe_seqlock); | |
673 | return; | |
4895c771 DM |
674 | } |
675 | ||
ceb33206 DM |
676 | static void __ip_do_redirect(struct rtable *rt, struct sk_buff *skb, struct flowi4 *fl4, |
677 | bool kill_route) | |
1da177e4 | 678 | { |
e47a185b | 679 | __be32 new_gw = icmp_hdr(skb)->un.gateway; |
94206125 | 680 | __be32 old_gw = ip_hdr(skb)->saddr; |
e47a185b | 681 | struct net_device *dev = skb->dev; |
e47a185b | 682 | struct in_device *in_dev; |
4895c771 | 683 | struct fib_result res; |
e47a185b | 684 | struct neighbour *n; |
317805b8 | 685 | struct net *net; |
1da177e4 | 686 | |
94206125 DM |
687 | switch (icmp_hdr(skb)->code & 7) { |
688 | case ICMP_REDIR_NET: | |
689 | case ICMP_REDIR_NETTOS: | |
690 | case ICMP_REDIR_HOST: | |
691 | case ICMP_REDIR_HOSTTOS: | |
692 | break; | |
693 | ||
694 | default: | |
695 | return; | |
696 | } | |
697 | ||
e47a185b DM |
698 | if (rt->rt_gateway != old_gw) |
699 | return; | |
700 | ||
701 | in_dev = __in_dev_get_rcu(dev); | |
702 | if (!in_dev) | |
703 | return; | |
704 | ||
c346dca1 | 705 | net = dev_net(dev); |
9d4fb27d JP |
706 | if (new_gw == old_gw || !IN_DEV_RX_REDIRECTS(in_dev) || |
707 | ipv4_is_multicast(new_gw) || ipv4_is_lbcast(new_gw) || | |
708 | ipv4_is_zeronet(new_gw)) | |
1da177e4 LT |
709 | goto reject_redirect; |
710 | ||
711 | if (!IN_DEV_SHARED_MEDIA(in_dev)) { | |
712 | if (!inet_addr_onlink(in_dev, new_gw, old_gw)) | |
713 | goto reject_redirect; | |
714 | if (IN_DEV_SEC_REDIRECTS(in_dev) && ip_fib_check_default(new_gw, dev)) | |
715 | goto reject_redirect; | |
716 | } else { | |
317805b8 | 717 | if (inet_addr_type(net, new_gw) != RTN_UNICAST) |
1da177e4 LT |
718 | goto reject_redirect; |
719 | } | |
720 | ||
4895c771 | 721 | n = ipv4_neigh_lookup(&rt->dst, NULL, &new_gw); |
e47a185b DM |
722 | if (n) { |
723 | if (!(n->nud_state & NUD_VALID)) { | |
724 | neigh_event_send(n, NULL); | |
725 | } else { | |
4895c771 DM |
726 | if (fib_lookup(net, fl4, &res) == 0) { |
727 | struct fib_nh *nh = &FIB_RES_NH(res); | |
4895c771 | 728 | |
aee06da6 JA |
729 | update_or_create_fnhe(nh, fl4->daddr, new_gw, |
730 | 0, 0); | |
4895c771 | 731 | } |
ceb33206 DM |
732 | if (kill_route) |
733 | rt->dst.obsolete = DST_OBSOLETE_KILL; | |
e47a185b DM |
734 | call_netevent_notifiers(NETEVENT_NEIGH_UPDATE, n); |
735 | } | |
736 | neigh_release(n); | |
737 | } | |
738 | return; | |
739 | ||
740 | reject_redirect: | |
741 | #ifdef CONFIG_IP_ROUTE_VERBOSE | |
99ee038d DM |
742 | if (IN_DEV_LOG_MARTIANS(in_dev)) { |
743 | const struct iphdr *iph = (const struct iphdr *) skb->data; | |
744 | __be32 daddr = iph->daddr; | |
745 | __be32 saddr = iph->saddr; | |
746 | ||
e47a185b DM |
747 | net_info_ratelimited("Redirect from %pI4 on %s about %pI4 ignored\n" |
748 | " Advised path = %pI4 -> %pI4\n", | |
749 | &old_gw, dev->name, &new_gw, | |
750 | &saddr, &daddr); | |
99ee038d | 751 | } |
e47a185b DM |
752 | #endif |
753 | ; | |
754 | } | |
755 | ||
4895c771 DM |
756 | static void ip_do_redirect(struct dst_entry *dst, struct sock *sk, struct sk_buff *skb) |
757 | { | |
758 | struct rtable *rt; | |
759 | struct flowi4 fl4; | |
760 | ||
761 | rt = (struct rtable *) dst; | |
762 | ||
763 | ip_rt_build_flow_key(&fl4, sk, skb); | |
ceb33206 | 764 | __ip_do_redirect(rt, skb, &fl4, true); |
4895c771 DM |
765 | } |
766 | ||
1da177e4 LT |
767 | static struct dst_entry *ipv4_negative_advice(struct dst_entry *dst) |
768 | { | |
ee6b9673 | 769 | struct rtable *rt = (struct rtable *)dst; |
1da177e4 LT |
770 | struct dst_entry *ret = dst; |
771 | ||
772 | if (rt) { | |
d11a4dc1 | 773 | if (dst->obsolete > 0) { |
1da177e4 LT |
774 | ip_rt_put(rt); |
775 | ret = NULL; | |
5943634f DM |
776 | } else if ((rt->rt_flags & RTCF_REDIRECTED) || |
777 | rt->dst.expires) { | |
89aef892 | 778 | ip_rt_put(rt); |
1da177e4 LT |
779 | ret = NULL; |
780 | } | |
781 | } | |
782 | return ret; | |
783 | } | |
784 | ||
785 | /* | |
786 | * Algorithm: | |
787 | * 1. The first ip_rt_redirect_number redirects are sent | |
788 | * with exponential backoff, then we stop sending them at all, | |
789 | * assuming that the host ignores our redirects. | |
790 | * 2. If we did not see packets requiring redirects | |
791 | * during ip_rt_redirect_silence, we assume that the host | |
792 | * forgot redirected route and start to send redirects again. | |
793 | * | |
794 | * This algorithm is much cheaper and more intelligent than dumb load limiting | |
795 | * in icmp.c. | |
796 | * | |
797 | * NOTE. Do not forget to inhibit load limiting for redirects (redundant) | |
798 | * and "frag. need" (breaks PMTU discovery) in icmp.c. | |
799 | */ | |
800 | ||
801 | void ip_rt_send_redirect(struct sk_buff *skb) | |
802 | { | |
511c3f92 | 803 | struct rtable *rt = skb_rtable(skb); |
30038fc6 | 804 | struct in_device *in_dev; |
92d86829 | 805 | struct inet_peer *peer; |
1d861aa4 | 806 | struct net *net; |
30038fc6 | 807 | int log_martians; |
1da177e4 | 808 | |
30038fc6 | 809 | rcu_read_lock(); |
d8d1f30b | 810 | in_dev = __in_dev_get_rcu(rt->dst.dev); |
30038fc6 ED |
811 | if (!in_dev || !IN_DEV_TX_REDIRECTS(in_dev)) { |
812 | rcu_read_unlock(); | |
1da177e4 | 813 | return; |
30038fc6 ED |
814 | } |
815 | log_martians = IN_DEV_LOG_MARTIANS(in_dev); | |
816 | rcu_read_unlock(); | |
1da177e4 | 817 | |
1d861aa4 DM |
818 | net = dev_net(rt->dst.dev); |
819 | peer = inet_getpeer_v4(net->ipv4.peers, ip_hdr(skb)->saddr, 1); | |
92d86829 DM |
820 | if (!peer) { |
821 | icmp_send(skb, ICMP_REDIRECT, ICMP_REDIR_HOST, rt->rt_gateway); | |
822 | return; | |
823 | } | |
824 | ||
1da177e4 LT |
825 | /* No redirected packets during ip_rt_redirect_silence; |
826 | * reset the algorithm. | |
827 | */ | |
92d86829 DM |
828 | if (time_after(jiffies, peer->rate_last + ip_rt_redirect_silence)) |
829 | peer->rate_tokens = 0; | |
1da177e4 LT |
830 | |
831 | /* Too many ignored redirects; do not send anything | |
d8d1f30b | 832 | * set dst.rate_last to the last seen redirected packet. |
1da177e4 | 833 | */ |
92d86829 DM |
834 | if (peer->rate_tokens >= ip_rt_redirect_number) { |
835 | peer->rate_last = jiffies; | |
1d861aa4 | 836 | goto out_put_peer; |
1da177e4 LT |
837 | } |
838 | ||
839 | /* Check for load limit; set rate_last to the latest sent | |
840 | * redirect. | |
841 | */ | |
92d86829 | 842 | if (peer->rate_tokens == 0 || |
14fb8a76 | 843 | time_after(jiffies, |
92d86829 DM |
844 | (peer->rate_last + |
845 | (ip_rt_redirect_load << peer->rate_tokens)))) { | |
1da177e4 | 846 | icmp_send(skb, ICMP_REDIRECT, ICMP_REDIR_HOST, rt->rt_gateway); |
92d86829 DM |
847 | peer->rate_last = jiffies; |
848 | ++peer->rate_tokens; | |
1da177e4 | 849 | #ifdef CONFIG_IP_ROUTE_VERBOSE |
30038fc6 | 850 | if (log_martians && |
e87cc472 JP |
851 | peer->rate_tokens == ip_rt_redirect_number) |
852 | net_warn_ratelimited("host %pI4/if%d ignores redirects for %pI4 to %pI4\n", | |
853 | &ip_hdr(skb)->saddr, rt->rt_iif, | |
f1ce3062 | 854 | &ip_hdr(skb)->daddr, &rt->rt_gateway); |
1da177e4 LT |
855 | #endif |
856 | } | |
1d861aa4 DM |
857 | out_put_peer: |
858 | inet_putpeer(peer); | |
1da177e4 LT |
859 | } |
860 | ||
861 | static int ip_error(struct sk_buff *skb) | |
862 | { | |
251da413 | 863 | struct in_device *in_dev = __in_dev_get_rcu(skb->dev); |
511c3f92 | 864 | struct rtable *rt = skb_rtable(skb); |
92d86829 | 865 | struct inet_peer *peer; |
1da177e4 | 866 | unsigned long now; |
251da413 | 867 | struct net *net; |
92d86829 | 868 | bool send; |
1da177e4 LT |
869 | int code; |
870 | ||
251da413 DM |
871 | net = dev_net(rt->dst.dev); |
872 | if (!IN_DEV_FORWARD(in_dev)) { | |
873 | switch (rt->dst.error) { | |
874 | case EHOSTUNREACH: | |
875 | IP_INC_STATS_BH(net, IPSTATS_MIB_INADDRERRORS); | |
876 | break; | |
877 | ||
878 | case ENETUNREACH: | |
879 | IP_INC_STATS_BH(net, IPSTATS_MIB_INNOROUTES); | |
880 | break; | |
881 | } | |
882 | goto out; | |
883 | } | |
884 | ||
d8d1f30b | 885 | switch (rt->dst.error) { |
4500ebf8 JP |
886 | case EINVAL: |
887 | default: | |
888 | goto out; | |
889 | case EHOSTUNREACH: | |
890 | code = ICMP_HOST_UNREACH; | |
891 | break; | |
892 | case ENETUNREACH: | |
893 | code = ICMP_NET_UNREACH; | |
251da413 | 894 | IP_INC_STATS_BH(net, IPSTATS_MIB_INNOROUTES); |
4500ebf8 JP |
895 | break; |
896 | case EACCES: | |
897 | code = ICMP_PKT_FILTERED; | |
898 | break; | |
1da177e4 LT |
899 | } |
900 | ||
1d861aa4 | 901 | peer = inet_getpeer_v4(net->ipv4.peers, ip_hdr(skb)->saddr, 1); |
92d86829 DM |
902 | |
903 | send = true; | |
904 | if (peer) { | |
905 | now = jiffies; | |
906 | peer->rate_tokens += now - peer->rate_last; | |
907 | if (peer->rate_tokens > ip_rt_error_burst) | |
908 | peer->rate_tokens = ip_rt_error_burst; | |
909 | peer->rate_last = now; | |
910 | if (peer->rate_tokens >= ip_rt_error_cost) | |
911 | peer->rate_tokens -= ip_rt_error_cost; | |
912 | else | |
913 | send = false; | |
1d861aa4 | 914 | inet_putpeer(peer); |
1da177e4 | 915 | } |
92d86829 DM |
916 | if (send) |
917 | icmp_send(skb, ICMP_DEST_UNREACH, code, 0); | |
1da177e4 LT |
918 | |
919 | out: kfree_skb(skb); | |
920 | return 0; | |
e905a9ed | 921 | } |
1da177e4 | 922 | |
ceb33206 | 923 | static u32 __ip_rt_update_pmtu(struct rtable *rt, struct flowi4 *fl4, u32 mtu) |
1da177e4 | 924 | { |
4895c771 | 925 | struct fib_result res; |
2c8cec5c | 926 | |
5943634f DM |
927 | if (mtu < ip_rt_min_pmtu) |
928 | mtu = ip_rt_min_pmtu; | |
2c8cec5c | 929 | |
4895c771 DM |
930 | if (fib_lookup(dev_net(rt->dst.dev), fl4, &res) == 0) { |
931 | struct fib_nh *nh = &FIB_RES_NH(res); | |
4895c771 | 932 | |
aee06da6 JA |
933 | update_or_create_fnhe(nh, fl4->daddr, 0, mtu, |
934 | jiffies + ip_rt_mtu_expires); | |
4895c771 | 935 | } |
ceb33206 | 936 | return mtu; |
1da177e4 LT |
937 | } |
938 | ||
4895c771 DM |
939 | static void ip_rt_update_pmtu(struct dst_entry *dst, struct sock *sk, |
940 | struct sk_buff *skb, u32 mtu) | |
941 | { | |
942 | struct rtable *rt = (struct rtable *) dst; | |
943 | struct flowi4 fl4; | |
944 | ||
945 | ip_rt_build_flow_key(&fl4, sk, skb); | |
ceb33206 DM |
946 | mtu = __ip_rt_update_pmtu(rt, &fl4, mtu); |
947 | ||
948 | if (!rt->rt_pmtu) { | |
949 | dst->obsolete = DST_OBSOLETE_KILL; | |
950 | } else { | |
951 | rt->rt_pmtu = mtu; | |
952 | dst_set_expires(&rt->dst, ip_rt_mtu_expires); | |
953 | } | |
4895c771 DM |
954 | } |
955 | ||
36393395 DM |
956 | void ipv4_update_pmtu(struct sk_buff *skb, struct net *net, u32 mtu, |
957 | int oif, u32 mark, u8 protocol, int flow_flags) | |
958 | { | |
4895c771 | 959 | const struct iphdr *iph = (const struct iphdr *) skb->data; |
36393395 DM |
960 | struct flowi4 fl4; |
961 | struct rtable *rt; | |
962 | ||
4895c771 DM |
963 | __build_flow_key(&fl4, NULL, iph, oif, |
964 | RT_TOS(iph->tos), protocol, mark, flow_flags); | |
36393395 DM |
965 | rt = __ip_route_output_key(net, &fl4); |
966 | if (!IS_ERR(rt)) { | |
4895c771 | 967 | __ip_rt_update_pmtu(rt, &fl4, mtu); |
36393395 DM |
968 | ip_rt_put(rt); |
969 | } | |
970 | } | |
971 | EXPORT_SYMBOL_GPL(ipv4_update_pmtu); | |
972 | ||
973 | void ipv4_sk_update_pmtu(struct sk_buff *skb, struct sock *sk, u32 mtu) | |
974 | { | |
4895c771 DM |
975 | const struct iphdr *iph = (const struct iphdr *) skb->data; |
976 | struct flowi4 fl4; | |
977 | struct rtable *rt; | |
36393395 | 978 | |
4895c771 DM |
979 | __build_flow_key(&fl4, sk, iph, 0, 0, 0, 0, 0); |
980 | rt = __ip_route_output_key(sock_net(sk), &fl4); | |
981 | if (!IS_ERR(rt)) { | |
982 | __ip_rt_update_pmtu(rt, &fl4, mtu); | |
983 | ip_rt_put(rt); | |
984 | } | |
36393395 DM |
985 | } |
986 | EXPORT_SYMBOL_GPL(ipv4_sk_update_pmtu); | |
f39925db | 987 | |
b42597e2 DM |
988 | void ipv4_redirect(struct sk_buff *skb, struct net *net, |
989 | int oif, u32 mark, u8 protocol, int flow_flags) | |
990 | { | |
4895c771 | 991 | const struct iphdr *iph = (const struct iphdr *) skb->data; |
b42597e2 DM |
992 | struct flowi4 fl4; |
993 | struct rtable *rt; | |
994 | ||
4895c771 DM |
995 | __build_flow_key(&fl4, NULL, iph, oif, |
996 | RT_TOS(iph->tos), protocol, mark, flow_flags); | |
b42597e2 DM |
997 | rt = __ip_route_output_key(net, &fl4); |
998 | if (!IS_ERR(rt)) { | |
ceb33206 | 999 | __ip_do_redirect(rt, skb, &fl4, false); |
b42597e2 DM |
1000 | ip_rt_put(rt); |
1001 | } | |
1002 | } | |
1003 | EXPORT_SYMBOL_GPL(ipv4_redirect); | |
1004 | ||
1005 | void ipv4_sk_redirect(struct sk_buff *skb, struct sock *sk) | |
1006 | { | |
4895c771 DM |
1007 | const struct iphdr *iph = (const struct iphdr *) skb->data; |
1008 | struct flowi4 fl4; | |
1009 | struct rtable *rt; | |
b42597e2 | 1010 | |
4895c771 DM |
1011 | __build_flow_key(&fl4, sk, iph, 0, 0, 0, 0, 0); |
1012 | rt = __ip_route_output_key(sock_net(sk), &fl4); | |
1013 | if (!IS_ERR(rt)) { | |
ceb33206 | 1014 | __ip_do_redirect(rt, skb, &fl4, false); |
4895c771 DM |
1015 | ip_rt_put(rt); |
1016 | } | |
b42597e2 DM |
1017 | } |
1018 | EXPORT_SYMBOL_GPL(ipv4_sk_redirect); | |
1019 | ||
efbc368d DM |
1020 | static struct dst_entry *ipv4_dst_check(struct dst_entry *dst, u32 cookie) |
1021 | { | |
1022 | struct rtable *rt = (struct rtable *) dst; | |
1023 | ||
ceb33206 DM |
1024 | /* All IPV4 dsts are created with ->obsolete set to the value |
1025 | * DST_OBSOLETE_FORCE_CHK which forces validation calls down | |
1026 | * into this function always. | |
1027 | * | |
1028 | * When a PMTU/redirect information update invalidates a | |
1029 | * route, this is indicated by setting obsolete to | |
1030 | * DST_OBSOLETE_KILL. | |
1031 | */ | |
1032 | if (dst->obsolete == DST_OBSOLETE_KILL || rt_is_expired(rt)) | |
efbc368d | 1033 | return NULL; |
d11a4dc1 | 1034 | return dst; |
1da177e4 LT |
1035 | } |
1036 | ||
1037 | static void ipv4_dst_destroy(struct dst_entry *dst) | |
1038 | { | |
1039 | struct rtable *rt = (struct rtable *) dst; | |
1da177e4 | 1040 | |
62fa8a84 DM |
1041 | if (rt->fi) { |
1042 | fib_info_put(rt->fi); | |
1043 | rt->fi = NULL; | |
1044 | } | |
1da177e4 LT |
1045 | } |
1046 | ||
1da177e4 LT |
1047 | |
1048 | static void ipv4_link_failure(struct sk_buff *skb) | |
1049 | { | |
1050 | struct rtable *rt; | |
1051 | ||
1052 | icmp_send(skb, ICMP_DEST_UNREACH, ICMP_HOST_UNREACH, 0); | |
1053 | ||
511c3f92 | 1054 | rt = skb_rtable(skb); |
5943634f DM |
1055 | if (rt) |
1056 | dst_set_expires(&rt->dst, 0); | |
1da177e4 LT |
1057 | } |
1058 | ||
1059 | static int ip_rt_bug(struct sk_buff *skb) | |
1060 | { | |
91df42be JP |
1061 | pr_debug("%s: %pI4 -> %pI4, %s\n", |
1062 | __func__, &ip_hdr(skb)->saddr, &ip_hdr(skb)->daddr, | |
1063 | skb->dev ? skb->dev->name : "?"); | |
1da177e4 | 1064 | kfree_skb(skb); |
c378a9c0 | 1065 | WARN_ON(1); |
1da177e4 LT |
1066 | return 0; |
1067 | } | |
1068 | ||
1069 | /* | |
1070 | We do not cache source address of outgoing interface, | |
1071 | because it is used only by IP RR, TS and SRR options, | |
1072 | so that it out of fast path. | |
1073 | ||
1074 | BTW remember: "addr" is allowed to be not aligned | |
1075 | in IP options! | |
1076 | */ | |
1077 | ||
8e36360a | 1078 | void ip_rt_get_source(u8 *addr, struct sk_buff *skb, struct rtable *rt) |
1da177e4 | 1079 | { |
a61ced5d | 1080 | __be32 src; |
1da177e4 | 1081 | |
c7537967 | 1082 | if (rt_is_output_route(rt)) |
c5be24ff | 1083 | src = ip_hdr(skb)->saddr; |
ebc0ffae | 1084 | else { |
8e36360a DM |
1085 | struct fib_result res; |
1086 | struct flowi4 fl4; | |
1087 | struct iphdr *iph; | |
1088 | ||
1089 | iph = ip_hdr(skb); | |
1090 | ||
1091 | memset(&fl4, 0, sizeof(fl4)); | |
1092 | fl4.daddr = iph->daddr; | |
1093 | fl4.saddr = iph->saddr; | |
b0fe4a31 | 1094 | fl4.flowi4_tos = RT_TOS(iph->tos); |
8e36360a DM |
1095 | fl4.flowi4_oif = rt->dst.dev->ifindex; |
1096 | fl4.flowi4_iif = skb->dev->ifindex; | |
1097 | fl4.flowi4_mark = skb->mark; | |
5e2b61f7 | 1098 | |
ebc0ffae | 1099 | rcu_read_lock(); |
68a5e3dd | 1100 | if (fib_lookup(dev_net(rt->dst.dev), &fl4, &res) == 0) |
436c3b66 | 1101 | src = FIB_RES_PREFSRC(dev_net(rt->dst.dev), res); |
ebc0ffae | 1102 | else |
f8126f1d DM |
1103 | src = inet_select_addr(rt->dst.dev, |
1104 | rt_nexthop(rt, iph->daddr), | |
1105 | RT_SCOPE_UNIVERSE); | |
ebc0ffae ED |
1106 | rcu_read_unlock(); |
1107 | } | |
1da177e4 LT |
1108 | memcpy(addr, &src, 4); |
1109 | } | |
1110 | ||
c7066f70 | 1111 | #ifdef CONFIG_IP_ROUTE_CLASSID |
1da177e4 LT |
1112 | static void set_class_tag(struct rtable *rt, u32 tag) |
1113 | { | |
d8d1f30b CG |
1114 | if (!(rt->dst.tclassid & 0xFFFF)) |
1115 | rt->dst.tclassid |= tag & 0xFFFF; | |
1116 | if (!(rt->dst.tclassid & 0xFFFF0000)) | |
1117 | rt->dst.tclassid |= tag & 0xFFFF0000; | |
1da177e4 LT |
1118 | } |
1119 | #endif | |
1120 | ||
0dbaee3b DM |
1121 | static unsigned int ipv4_default_advmss(const struct dst_entry *dst) |
1122 | { | |
1123 | unsigned int advmss = dst_metric_raw(dst, RTAX_ADVMSS); | |
1124 | ||
1125 | if (advmss == 0) { | |
1126 | advmss = max_t(unsigned int, dst->dev->mtu - 40, | |
1127 | ip_rt_min_advmss); | |
1128 | if (advmss > 65535 - 40) | |
1129 | advmss = 65535 - 40; | |
1130 | } | |
1131 | return advmss; | |
1132 | } | |
1133 | ||
ebb762f2 | 1134 | static unsigned int ipv4_mtu(const struct dst_entry *dst) |
d33e4553 | 1135 | { |
261663b0 | 1136 | const struct rtable *rt = (const struct rtable *) dst; |
5943634f DM |
1137 | unsigned int mtu = rt->rt_pmtu; |
1138 | ||
1139 | if (mtu && time_after_eq(jiffies, rt->dst.expires)) | |
1140 | mtu = 0; | |
1141 | ||
1142 | if (!mtu) | |
1143 | mtu = dst_metric_raw(dst, RTAX_MTU); | |
618f9bc7 | 1144 | |
261663b0 | 1145 | if (mtu && rt_is_output_route(rt)) |
618f9bc7 SK |
1146 | return mtu; |
1147 | ||
1148 | mtu = dst->dev->mtu; | |
d33e4553 DM |
1149 | |
1150 | if (unlikely(dst_metric_locked(dst, RTAX_MTU))) { | |
f8126f1d | 1151 | if (rt->rt_gateway && mtu > 576) |
d33e4553 DM |
1152 | mtu = 576; |
1153 | } | |
1154 | ||
1155 | if (mtu > IP_MAX_MTU) | |
1156 | mtu = IP_MAX_MTU; | |
1157 | ||
1158 | return mtu; | |
1159 | } | |
1160 | ||
813b3b5d | 1161 | static void rt_init_metrics(struct rtable *rt, const struct flowi4 *fl4, |
5e2b61f7 | 1162 | struct fib_info *fi) |
a4daad6b | 1163 | { |
f185071d DM |
1164 | if (fi->fib_metrics != (u32 *) dst_default_metrics) { |
1165 | rt->fi = fi; | |
1166 | atomic_inc(&fi->fib_clntref); | |
a4daad6b | 1167 | } |
f185071d | 1168 | dst_init_metrics(&rt->dst, fi->fib_metrics, true); |
a4daad6b DM |
1169 | } |
1170 | ||
4895c771 DM |
1171 | static void rt_bind_exception(struct rtable *rt, struct fib_nh *nh, __be32 daddr) |
1172 | { | |
1173 | struct fnhe_hash_bucket *hash = nh->nh_exceptions; | |
1174 | struct fib_nh_exception *fnhe; | |
1175 | u32 hval; | |
1176 | ||
d3a25c98 | 1177 | hval = fnhe_hashfun(daddr); |
4895c771 | 1178 | |
aee06da6 | 1179 | restart: |
4895c771 DM |
1180 | for (fnhe = rcu_dereference(hash[hval].chain); fnhe; |
1181 | fnhe = rcu_dereference(fnhe->fnhe_next)) { | |
aee06da6 JA |
1182 | __be32 fnhe_daddr, gw; |
1183 | unsigned long expires; | |
1184 | unsigned int seq; | |
1185 | u32 pmtu; | |
1186 | ||
1187 | seq = read_seqbegin(&fnhe_seqlock); | |
1188 | fnhe_daddr = fnhe->fnhe_daddr; | |
1189 | gw = fnhe->fnhe_gw; | |
1190 | pmtu = fnhe->fnhe_pmtu; | |
1191 | expires = fnhe->fnhe_expires; | |
1192 | if (read_seqretry(&fnhe_seqlock, seq)) | |
1193 | goto restart; | |
1194 | if (daddr != fnhe_daddr) | |
1195 | continue; | |
1196 | if (pmtu) { | |
f31fd383 | 1197 | unsigned long diff = expires - jiffies; |
aee06da6 JA |
1198 | |
1199 | if (time_before(jiffies, expires)) { | |
1200 | rt->rt_pmtu = pmtu; | |
1201 | dst_set_expires(&rt->dst, diff); | |
4895c771 | 1202 | } |
4895c771 | 1203 | } |
ceb33206 DM |
1204 | if (gw) { |
1205 | rt->rt_flags |= RTCF_REDIRECTED; | |
aee06da6 | 1206 | rt->rt_gateway = gw; |
ceb33206 | 1207 | } |
aee06da6 JA |
1208 | fnhe->fnhe_stamp = jiffies; |
1209 | break; | |
4895c771 DM |
1210 | } |
1211 | } | |
1212 | ||
813b3b5d | 1213 | static void rt_set_nexthop(struct rtable *rt, const struct flowi4 *fl4, |
5e2b61f7 | 1214 | const struct fib_result *res, |
982721f3 | 1215 | struct fib_info *fi, u16 type, u32 itag) |
1da177e4 | 1216 | { |
1da177e4 | 1217 | if (fi) { |
4895c771 DM |
1218 | struct fib_nh *nh = &FIB_RES_NH(*res); |
1219 | ||
1220 | if (nh->nh_gw && nh->nh_scope == RT_SCOPE_LINK) | |
1221 | rt->rt_gateway = nh->nh_gw; | |
1222 | if (unlikely(nh->nh_exceptions)) | |
1223 | rt_bind_exception(rt, nh, fl4->daddr); | |
813b3b5d | 1224 | rt_init_metrics(rt, fl4, fi); |
c7066f70 | 1225 | #ifdef CONFIG_IP_ROUTE_CLASSID |
710ab6c0 | 1226 | rt->dst.tclassid = FIB_RES_NH(*res).nh_tclassid; |
1da177e4 | 1227 | #endif |
d33e4553 | 1228 | } |
defb3519 | 1229 | |
c7066f70 | 1230 | #ifdef CONFIG_IP_ROUTE_CLASSID |
1da177e4 | 1231 | #ifdef CONFIG_IP_MULTIPLE_TABLES |
85b91b03 | 1232 | set_class_tag(rt, res->tclassid); |
1da177e4 LT |
1233 | #endif |
1234 | set_class_tag(rt, itag); | |
1235 | #endif | |
1da177e4 LT |
1236 | } |
1237 | ||
5c1e6aa3 DM |
1238 | static struct rtable *rt_dst_alloc(struct net_device *dev, |
1239 | bool nopolicy, bool noxfrm) | |
0c4dcd58 | 1240 | { |
f5b0a874 | 1241 | return dst_alloc(&ipv4_dst_ops, dev, 1, DST_OBSOLETE_FORCE_CHK, |
89aef892 | 1242 | DST_HOST | DST_NOCACHE | |
5c1e6aa3 DM |
1243 | (nopolicy ? DST_NOPOLICY : 0) | |
1244 | (noxfrm ? DST_NOXFRM : 0)); | |
0c4dcd58 DM |
1245 | } |
1246 | ||
96d36220 | 1247 | /* called in rcu_read_lock() section */ |
9e12bb22 | 1248 | static int ip_route_input_mc(struct sk_buff *skb, __be32 daddr, __be32 saddr, |
1da177e4 LT |
1249 | u8 tos, struct net_device *dev, int our) |
1250 | { | |
1da177e4 | 1251 | struct rtable *rth; |
96d36220 | 1252 | struct in_device *in_dev = __in_dev_get_rcu(dev); |
1da177e4 | 1253 | u32 itag = 0; |
b5f7e755 | 1254 | int err; |
1da177e4 LT |
1255 | |
1256 | /* Primary sanity checks. */ | |
1257 | ||
1258 | if (in_dev == NULL) | |
1259 | return -EINVAL; | |
1260 | ||
1e637c74 | 1261 | if (ipv4_is_multicast(saddr) || ipv4_is_lbcast(saddr) || |
d0daebc3 | 1262 | skb->protocol != htons(ETH_P_IP)) |
1da177e4 LT |
1263 | goto e_inval; |
1264 | ||
d0daebc3 TG |
1265 | if (likely(!IN_DEV_ROUTE_LOCALNET(in_dev))) |
1266 | if (ipv4_is_loopback(saddr)) | |
1267 | goto e_inval; | |
1268 | ||
f97c1e0c JP |
1269 | if (ipv4_is_zeronet(saddr)) { |
1270 | if (!ipv4_is_local_multicast(daddr)) | |
1da177e4 | 1271 | goto e_inval; |
b5f7e755 | 1272 | } else { |
9e56e380 DM |
1273 | err = fib_validate_source(skb, saddr, 0, tos, 0, dev, |
1274 | in_dev, &itag); | |
b5f7e755 ED |
1275 | if (err < 0) |
1276 | goto e_err; | |
1277 | } | |
4e7b2f14 | 1278 | rth = rt_dst_alloc(dev_net(dev)->loopback_dev, |
5c1e6aa3 | 1279 | IN_DEV_CONF_GET(in_dev, NOPOLICY), false); |
1da177e4 LT |
1280 | if (!rth) |
1281 | goto e_nobufs; | |
1282 | ||
cf911662 DM |
1283 | #ifdef CONFIG_IP_ROUTE_CLASSID |
1284 | rth->dst.tclassid = itag; | |
1285 | #endif | |
d8d1f30b | 1286 | rth->dst.output = ip_rt_bug; |
1da177e4 | 1287 | |
cf911662 DM |
1288 | rth->rt_genid = rt_genid(dev_net(dev)); |
1289 | rth->rt_flags = RTCF_MULTICAST; | |
1290 | rth->rt_type = RTN_MULTICAST; | |
1b86a58f | 1291 | rth->rt_route_iif = dev->ifindex; |
5e2b61f7 | 1292 | rth->rt_iif = dev->ifindex; |
5e2b61f7 | 1293 | rth->rt_oif = 0; |
5943634f | 1294 | rth->rt_pmtu = 0; |
f8126f1d | 1295 | rth->rt_gateway = 0; |
cf911662 | 1296 | rth->fi = NULL; |
1da177e4 | 1297 | if (our) { |
d8d1f30b | 1298 | rth->dst.input= ip_local_deliver; |
1da177e4 LT |
1299 | rth->rt_flags |= RTCF_LOCAL; |
1300 | } | |
1301 | ||
1302 | #ifdef CONFIG_IP_MROUTE | |
f97c1e0c | 1303 | if (!ipv4_is_local_multicast(daddr) && IN_DEV_MFORWARD(in_dev)) |
d8d1f30b | 1304 | rth->dst.input = ip_mr_input; |
1da177e4 LT |
1305 | #endif |
1306 | RT_CACHE_STAT_INC(in_slow_mc); | |
1307 | ||
89aef892 DM |
1308 | skb_dst_set(skb, &rth->dst); |
1309 | return 0; | |
1da177e4 LT |
1310 | |
1311 | e_nobufs: | |
1da177e4 | 1312 | return -ENOBUFS; |
1da177e4 | 1313 | e_inval: |
96d36220 | 1314 | return -EINVAL; |
b5f7e755 | 1315 | e_err: |
b5f7e755 | 1316 | return err; |
1da177e4 LT |
1317 | } |
1318 | ||
1319 | ||
1320 | static void ip_handle_martian_source(struct net_device *dev, | |
1321 | struct in_device *in_dev, | |
1322 | struct sk_buff *skb, | |
9e12bb22 AV |
1323 | __be32 daddr, |
1324 | __be32 saddr) | |
1da177e4 LT |
1325 | { |
1326 | RT_CACHE_STAT_INC(in_martian_src); | |
1327 | #ifdef CONFIG_IP_ROUTE_VERBOSE | |
1328 | if (IN_DEV_LOG_MARTIANS(in_dev) && net_ratelimit()) { | |
1329 | /* | |
1330 | * RFC1812 recommendation, if source is martian, | |
1331 | * the only hint is MAC header. | |
1332 | */ | |
058bd4d2 | 1333 | pr_warn("martian source %pI4 from %pI4, on dev %s\n", |
673d57e7 | 1334 | &daddr, &saddr, dev->name); |
98e399f8 | 1335 | if (dev->hard_header_len && skb_mac_header_was_set(skb)) { |
058bd4d2 JP |
1336 | print_hex_dump(KERN_WARNING, "ll header: ", |
1337 | DUMP_PREFIX_OFFSET, 16, 1, | |
1338 | skb_mac_header(skb), | |
1339 | dev->hard_header_len, true); | |
1da177e4 LT |
1340 | } |
1341 | } | |
1342 | #endif | |
1343 | } | |
1344 | ||
47360228 | 1345 | /* called in rcu_read_lock() section */ |
5969f71d | 1346 | static int __mkroute_input(struct sk_buff *skb, |
982721f3 | 1347 | const struct fib_result *res, |
5969f71d SH |
1348 | struct in_device *in_dev, |
1349 | __be32 daddr, __be32 saddr, u32 tos, | |
1350 | struct rtable **result) | |
1da177e4 | 1351 | { |
1da177e4 LT |
1352 | struct rtable *rth; |
1353 | int err; | |
1354 | struct in_device *out_dev; | |
47360228 | 1355 | unsigned int flags = 0; |
d9c9df8c | 1356 | u32 itag; |
1da177e4 LT |
1357 | |
1358 | /* get a working reference to the output device */ | |
47360228 | 1359 | out_dev = __in_dev_get_rcu(FIB_RES_DEV(*res)); |
1da177e4 | 1360 | if (out_dev == NULL) { |
e87cc472 | 1361 | net_crit_ratelimited("Bug in ip_route_input_slow(). Please report.\n"); |
1da177e4 LT |
1362 | return -EINVAL; |
1363 | } | |
1364 | ||
1365 | ||
5c04c819 | 1366 | err = fib_validate_source(skb, saddr, daddr, tos, FIB_RES_OIF(*res), |
9e56e380 | 1367 | in_dev->dev, in_dev, &itag); |
1da177e4 | 1368 | if (err < 0) { |
e905a9ed | 1369 | ip_handle_martian_source(in_dev->dev, in_dev, skb, daddr, |
1da177e4 | 1370 | saddr); |
e905a9ed | 1371 | |
1da177e4 LT |
1372 | goto cleanup; |
1373 | } | |
1374 | ||
1375 | if (err) | |
1376 | flags |= RTCF_DIRECTSRC; | |
1377 | ||
51b77cae | 1378 | if (out_dev == in_dev && err && |
1da177e4 LT |
1379 | (IN_DEV_SHARED_MEDIA(out_dev) || |
1380 | inet_addr_onlink(out_dev, saddr, FIB_RES_GW(*res)))) | |
1381 | flags |= RTCF_DOREDIRECT; | |
1382 | ||
1383 | if (skb->protocol != htons(ETH_P_IP)) { | |
1384 | /* Not IP (i.e. ARP). Do not create route, if it is | |
1385 | * invalid for proxy arp. DNAT routes are always valid. | |
65324144 JDB |
1386 | * |
1387 | * Proxy arp feature have been extended to allow, ARP | |
1388 | * replies back to the same interface, to support | |
1389 | * Private VLAN switch technologies. See arp.c. | |
1da177e4 | 1390 | */ |
65324144 JDB |
1391 | if (out_dev == in_dev && |
1392 | IN_DEV_PROXY_ARP_PVLAN(in_dev) == 0) { | |
1da177e4 LT |
1393 | err = -EINVAL; |
1394 | goto cleanup; | |
1395 | } | |
1396 | } | |
1397 | ||
5c1e6aa3 DM |
1398 | rth = rt_dst_alloc(out_dev->dev, |
1399 | IN_DEV_CONF_GET(in_dev, NOPOLICY), | |
0c4dcd58 | 1400 | IN_DEV_CONF_GET(out_dev, NOXFRM)); |
1da177e4 LT |
1401 | if (!rth) { |
1402 | err = -ENOBUFS; | |
1403 | goto cleanup; | |
1404 | } | |
1405 | ||
cf911662 DM |
1406 | rth->rt_genid = rt_genid(dev_net(rth->dst.dev)); |
1407 | rth->rt_flags = flags; | |
1408 | rth->rt_type = res->type; | |
1b86a58f | 1409 | rth->rt_route_iif = in_dev->dev->ifindex; |
5e2b61f7 | 1410 | rth->rt_iif = in_dev->dev->ifindex; |
5e2b61f7 | 1411 | rth->rt_oif = 0; |
5943634f | 1412 | rth->rt_pmtu = 0; |
f8126f1d | 1413 | rth->rt_gateway = 0; |
cf911662 | 1414 | rth->fi = NULL; |
1da177e4 | 1415 | |
d8d1f30b CG |
1416 | rth->dst.input = ip_forward; |
1417 | rth->dst.output = ip_output; | |
1da177e4 | 1418 | |
5e2b61f7 | 1419 | rt_set_nexthop(rth, NULL, res, res->fi, res->type, itag); |
1da177e4 | 1420 | |
1da177e4 LT |
1421 | *result = rth; |
1422 | err = 0; | |
1423 | cleanup: | |
1da177e4 | 1424 | return err; |
e905a9ed | 1425 | } |
1da177e4 | 1426 | |
5969f71d SH |
1427 | static int ip_mkroute_input(struct sk_buff *skb, |
1428 | struct fib_result *res, | |
68a5e3dd | 1429 | const struct flowi4 *fl4, |
5969f71d SH |
1430 | struct in_device *in_dev, |
1431 | __be32 daddr, __be32 saddr, u32 tos) | |
1da177e4 | 1432 | { |
5e73ea1a | 1433 | struct rtable *rth = NULL; |
1da177e4 | 1434 | int err; |
1da177e4 LT |
1435 | |
1436 | #ifdef CONFIG_IP_ROUTE_MULTIPATH | |
ff3fccb3 | 1437 | if (res->fi && res->fi->fib_nhs > 1) |
1b7fe593 | 1438 | fib_select_multipath(res); |
1da177e4 LT |
1439 | #endif |
1440 | ||
1441 | /* create a routing cache entry */ | |
1442 | err = __mkroute_input(skb, res, in_dev, daddr, saddr, tos, &rth); | |
1443 | if (err) | |
1444 | return err; | |
1da177e4 | 1445 | |
89aef892 | 1446 | skb_dst_set(skb, &rth->dst); |
b23dd4fe | 1447 | return 0; |
1da177e4 LT |
1448 | } |
1449 | ||
1da177e4 LT |
1450 | /* |
1451 | * NOTE. We drop all the packets that has local source | |
1452 | * addresses, because every properly looped back packet | |
1453 | * must have correct destination already attached by output routine. | |
1454 | * | |
1455 | * Such approach solves two big problems: | |
1456 | * 1. Not simplex devices are handled properly. | |
1457 | * 2. IP spoofing attempts are filtered with 100% of guarantee. | |
ebc0ffae | 1458 | * called with rcu_read_lock() |
1da177e4 LT |
1459 | */ |
1460 | ||
9e12bb22 | 1461 | static int ip_route_input_slow(struct sk_buff *skb, __be32 daddr, __be32 saddr, |
c10237e0 | 1462 | u8 tos, struct net_device *dev) |
1da177e4 LT |
1463 | { |
1464 | struct fib_result res; | |
96d36220 | 1465 | struct in_device *in_dev = __in_dev_get_rcu(dev); |
68a5e3dd | 1466 | struct flowi4 fl4; |
95c96174 | 1467 | unsigned int flags = 0; |
1da177e4 | 1468 | u32 itag = 0; |
95c96174 | 1469 | struct rtable *rth; |
1da177e4 | 1470 | int err = -EINVAL; |
5e73ea1a | 1471 | struct net *net = dev_net(dev); |
1da177e4 LT |
1472 | |
1473 | /* IP on this device is disabled. */ | |
1474 | ||
1475 | if (!in_dev) | |
1476 | goto out; | |
1477 | ||
1478 | /* Check for the most weird martians, which can be not detected | |
1479 | by fib_lookup. | |
1480 | */ | |
1481 | ||
d0daebc3 | 1482 | if (ipv4_is_multicast(saddr) || ipv4_is_lbcast(saddr)) |
1da177e4 LT |
1483 | goto martian_source; |
1484 | ||
27a954bd | 1485 | if (ipv4_is_lbcast(daddr) || (saddr == 0 && daddr == 0)) |
1da177e4 LT |
1486 | goto brd_input; |
1487 | ||
1488 | /* Accept zero addresses only to limited broadcast; | |
1489 | * I even do not know to fix it or not. Waiting for complains :-) | |
1490 | */ | |
f97c1e0c | 1491 | if (ipv4_is_zeronet(saddr)) |
1da177e4 LT |
1492 | goto martian_source; |
1493 | ||
d0daebc3 | 1494 | if (ipv4_is_zeronet(daddr)) |
1da177e4 LT |
1495 | goto martian_destination; |
1496 | ||
d0daebc3 TG |
1497 | if (likely(!IN_DEV_ROUTE_LOCALNET(in_dev))) { |
1498 | if (ipv4_is_loopback(daddr)) | |
1499 | goto martian_destination; | |
1500 | ||
1501 | if (ipv4_is_loopback(saddr)) | |
1502 | goto martian_source; | |
1503 | } | |
1504 | ||
1da177e4 LT |
1505 | /* |
1506 | * Now we are ready to route packet. | |
1507 | */ | |
68a5e3dd DM |
1508 | fl4.flowi4_oif = 0; |
1509 | fl4.flowi4_iif = dev->ifindex; | |
1510 | fl4.flowi4_mark = skb->mark; | |
1511 | fl4.flowi4_tos = tos; | |
1512 | fl4.flowi4_scope = RT_SCOPE_UNIVERSE; | |
1513 | fl4.daddr = daddr; | |
1514 | fl4.saddr = saddr; | |
1515 | err = fib_lookup(net, &fl4, &res); | |
251da413 | 1516 | if (err != 0) |
1da177e4 | 1517 | goto no_route; |
1da177e4 LT |
1518 | |
1519 | RT_CACHE_STAT_INC(in_slow_tot); | |
1520 | ||
1521 | if (res.type == RTN_BROADCAST) | |
1522 | goto brd_input; | |
1523 | ||
1524 | if (res.type == RTN_LOCAL) { | |
5c04c819 | 1525 | err = fib_validate_source(skb, saddr, daddr, tos, |
ebc0ffae | 1526 | net->loopback_dev->ifindex, |
9e56e380 | 1527 | dev, in_dev, &itag); |
b5f7e755 ED |
1528 | if (err < 0) |
1529 | goto martian_source_keep_err; | |
1530 | if (err) | |
1da177e4 | 1531 | flags |= RTCF_DIRECTSRC; |
1da177e4 LT |
1532 | goto local_input; |
1533 | } | |
1534 | ||
1535 | if (!IN_DEV_FORWARD(in_dev)) | |
251da413 | 1536 | goto no_route; |
1da177e4 LT |
1537 | if (res.type != RTN_UNICAST) |
1538 | goto martian_destination; | |
1539 | ||
68a5e3dd | 1540 | err = ip_mkroute_input(skb, &res, &fl4, in_dev, daddr, saddr, tos); |
1da177e4 LT |
1541 | out: return err; |
1542 | ||
1543 | brd_input: | |
1544 | if (skb->protocol != htons(ETH_P_IP)) | |
1545 | goto e_inval; | |
1546 | ||
41347dcd | 1547 | if (!ipv4_is_zeronet(saddr)) { |
9e56e380 DM |
1548 | err = fib_validate_source(skb, saddr, 0, tos, 0, dev, |
1549 | in_dev, &itag); | |
1da177e4 | 1550 | if (err < 0) |
b5f7e755 | 1551 | goto martian_source_keep_err; |
1da177e4 LT |
1552 | if (err) |
1553 | flags |= RTCF_DIRECTSRC; | |
1554 | } | |
1555 | flags |= RTCF_BROADCAST; | |
1556 | res.type = RTN_BROADCAST; | |
1557 | RT_CACHE_STAT_INC(in_brd); | |
1558 | ||
1559 | local_input: | |
5c1e6aa3 DM |
1560 | rth = rt_dst_alloc(net->loopback_dev, |
1561 | IN_DEV_CONF_GET(in_dev, NOPOLICY), false); | |
1da177e4 LT |
1562 | if (!rth) |
1563 | goto e_nobufs; | |
1564 | ||
cf911662 | 1565 | rth->dst.input= ip_local_deliver; |
d8d1f30b | 1566 | rth->dst.output= ip_rt_bug; |
cf911662 DM |
1567 | #ifdef CONFIG_IP_ROUTE_CLASSID |
1568 | rth->dst.tclassid = itag; | |
1569 | #endif | |
1da177e4 | 1570 | |
cf911662 DM |
1571 | rth->rt_genid = rt_genid(net); |
1572 | rth->rt_flags = flags|RTCF_LOCAL; | |
1573 | rth->rt_type = res.type; | |
1b86a58f | 1574 | rth->rt_route_iif = dev->ifindex; |
5e2b61f7 | 1575 | rth->rt_iif = dev->ifindex; |
cf911662 | 1576 | rth->rt_oif = 0; |
5943634f | 1577 | rth->rt_pmtu = 0; |
f8126f1d | 1578 | rth->rt_gateway = 0; |
cf911662 | 1579 | rth->fi = NULL; |
1da177e4 | 1580 | if (res.type == RTN_UNREACHABLE) { |
d8d1f30b CG |
1581 | rth->dst.input= ip_error; |
1582 | rth->dst.error= -err; | |
1da177e4 LT |
1583 | rth->rt_flags &= ~RTCF_LOCAL; |
1584 | } | |
89aef892 | 1585 | skb_dst_set(skb, &rth->dst); |
b23dd4fe | 1586 | err = 0; |
ebc0ffae | 1587 | goto out; |
1da177e4 LT |
1588 | |
1589 | no_route: | |
1590 | RT_CACHE_STAT_INC(in_no_route); | |
1da177e4 | 1591 | res.type = RTN_UNREACHABLE; |
7f53878d MC |
1592 | if (err == -ESRCH) |
1593 | err = -ENETUNREACH; | |
1da177e4 LT |
1594 | goto local_input; |
1595 | ||
1596 | /* | |
1597 | * Do not cache martian addresses: they should be logged (RFC1812) | |
1598 | */ | |
1599 | martian_destination: | |
1600 | RT_CACHE_STAT_INC(in_martian_dst); | |
1601 | #ifdef CONFIG_IP_ROUTE_VERBOSE | |
e87cc472 JP |
1602 | if (IN_DEV_LOG_MARTIANS(in_dev)) |
1603 | net_warn_ratelimited("martian destination %pI4 from %pI4, dev %s\n", | |
1604 | &daddr, &saddr, dev->name); | |
1da177e4 | 1605 | #endif |
2c2910a4 | 1606 | |
1da177e4 LT |
1607 | e_inval: |
1608 | err = -EINVAL; | |
ebc0ffae | 1609 | goto out; |
1da177e4 LT |
1610 | |
1611 | e_nobufs: | |
1612 | err = -ENOBUFS; | |
ebc0ffae | 1613 | goto out; |
1da177e4 LT |
1614 | |
1615 | martian_source: | |
b5f7e755 ED |
1616 | err = -EINVAL; |
1617 | martian_source_keep_err: | |
1da177e4 | 1618 | ip_handle_martian_source(dev, in_dev, skb, daddr, saddr); |
ebc0ffae | 1619 | goto out; |
1da177e4 LT |
1620 | } |
1621 | ||
38a424e4 DM |
1622 | int ip_route_input(struct sk_buff *skb, __be32 daddr, __be32 saddr, |
1623 | u8 tos, struct net_device *dev) | |
1da177e4 | 1624 | { |
96d36220 | 1625 | int res; |
1da177e4 | 1626 | |
96d36220 ED |
1627 | rcu_read_lock(); |
1628 | ||
1da177e4 LT |
1629 | /* Multicast recognition logic is moved from route cache to here. |
1630 | The problem was that too many Ethernet cards have broken/missing | |
1631 | hardware multicast filters :-( As result the host on multicasting | |
1632 | network acquires a lot of useless route cache entries, sort of | |
1633 | SDR messages from all the world. Now we try to get rid of them. | |
1634 | Really, provided software IP multicast filter is organized | |
1635 | reasonably (at least, hashed), it does not result in a slowdown | |
1636 | comparing with route cache reject entries. | |
1637 | Note, that multicast routers are not affected, because | |
1638 | route cache entry is created eventually. | |
1639 | */ | |
f97c1e0c | 1640 | if (ipv4_is_multicast(daddr)) { |
96d36220 | 1641 | struct in_device *in_dev = __in_dev_get_rcu(dev); |
1da177e4 | 1642 | |
96d36220 | 1643 | if (in_dev) { |
dbdd9a52 DM |
1644 | int our = ip_check_mc_rcu(in_dev, daddr, saddr, |
1645 | ip_hdr(skb)->protocol); | |
1da177e4 LT |
1646 | if (our |
1647 | #ifdef CONFIG_IP_MROUTE | |
9d4fb27d JP |
1648 | || |
1649 | (!ipv4_is_local_multicast(daddr) && | |
1650 | IN_DEV_MFORWARD(in_dev)) | |
1da177e4 | 1651 | #endif |
9d4fb27d | 1652 | ) { |
96d36220 ED |
1653 | int res = ip_route_input_mc(skb, daddr, saddr, |
1654 | tos, dev, our); | |
1da177e4 | 1655 | rcu_read_unlock(); |
96d36220 | 1656 | return res; |
1da177e4 LT |
1657 | } |
1658 | } | |
1659 | rcu_read_unlock(); | |
1660 | return -EINVAL; | |
1661 | } | |
c10237e0 | 1662 | res = ip_route_input_slow(skb, daddr, saddr, tos, dev); |
96d36220 ED |
1663 | rcu_read_unlock(); |
1664 | return res; | |
1da177e4 | 1665 | } |
38a424e4 | 1666 | EXPORT_SYMBOL(ip_route_input); |
1da177e4 | 1667 | |
ebc0ffae | 1668 | /* called with rcu_read_lock() */ |
982721f3 | 1669 | static struct rtable *__mkroute_output(const struct fib_result *res, |
1a00fee4 | 1670 | const struct flowi4 *fl4, int orig_oif, |
f61759e6 | 1671 | struct net_device *dev_out, |
5ada5527 | 1672 | unsigned int flags) |
1da177e4 | 1673 | { |
982721f3 | 1674 | struct fib_info *fi = res->fi; |
5ada5527 | 1675 | struct in_device *in_dev; |
982721f3 | 1676 | u16 type = res->type; |
5ada5527 | 1677 | struct rtable *rth; |
1da177e4 | 1678 | |
d0daebc3 TG |
1679 | in_dev = __in_dev_get_rcu(dev_out); |
1680 | if (!in_dev) | |
5ada5527 | 1681 | return ERR_PTR(-EINVAL); |
1da177e4 | 1682 | |
d0daebc3 TG |
1683 | if (likely(!IN_DEV_ROUTE_LOCALNET(in_dev))) |
1684 | if (ipv4_is_loopback(fl4->saddr) && !(dev_out->flags & IFF_LOOPBACK)) | |
1685 | return ERR_PTR(-EINVAL); | |
1686 | ||
68a5e3dd | 1687 | if (ipv4_is_lbcast(fl4->daddr)) |
982721f3 | 1688 | type = RTN_BROADCAST; |
68a5e3dd | 1689 | else if (ipv4_is_multicast(fl4->daddr)) |
982721f3 | 1690 | type = RTN_MULTICAST; |
68a5e3dd | 1691 | else if (ipv4_is_zeronet(fl4->daddr)) |
5ada5527 | 1692 | return ERR_PTR(-EINVAL); |
1da177e4 LT |
1693 | |
1694 | if (dev_out->flags & IFF_LOOPBACK) | |
1695 | flags |= RTCF_LOCAL; | |
1696 | ||
982721f3 | 1697 | if (type == RTN_BROADCAST) { |
1da177e4 | 1698 | flags |= RTCF_BROADCAST | RTCF_LOCAL; |
982721f3 DM |
1699 | fi = NULL; |
1700 | } else if (type == RTN_MULTICAST) { | |
dd28d1a0 | 1701 | flags |= RTCF_MULTICAST | RTCF_LOCAL; |
813b3b5d DM |
1702 | if (!ip_check_mc_rcu(in_dev, fl4->daddr, fl4->saddr, |
1703 | fl4->flowi4_proto)) | |
1da177e4 LT |
1704 | flags &= ~RTCF_LOCAL; |
1705 | /* If multicast route do not exist use | |
dd28d1a0 ED |
1706 | * default one, but do not gateway in this case. |
1707 | * Yes, it is hack. | |
1da177e4 | 1708 | */ |
982721f3 DM |
1709 | if (fi && res->prefixlen < 4) |
1710 | fi = NULL; | |
1da177e4 LT |
1711 | } |
1712 | ||
5c1e6aa3 DM |
1713 | rth = rt_dst_alloc(dev_out, |
1714 | IN_DEV_CONF_GET(in_dev, NOPOLICY), | |
0c4dcd58 | 1715 | IN_DEV_CONF_GET(in_dev, NOXFRM)); |
8391d07b | 1716 | if (!rth) |
5ada5527 | 1717 | return ERR_PTR(-ENOBUFS); |
8391d07b | 1718 | |
cf911662 DM |
1719 | rth->dst.output = ip_output; |
1720 | ||
cf911662 DM |
1721 | rth->rt_genid = rt_genid(dev_net(dev_out)); |
1722 | rth->rt_flags = flags; | |
1723 | rth->rt_type = type; | |
1b86a58f | 1724 | rth->rt_route_iif = 0; |
813b3b5d DM |
1725 | rth->rt_iif = orig_oif ? : dev_out->ifindex; |
1726 | rth->rt_oif = orig_oif; | |
5943634f | 1727 | rth->rt_pmtu = 0; |
f8126f1d | 1728 | rth->rt_gateway = 0; |
cf911662 | 1729 | rth->fi = NULL; |
1da177e4 LT |
1730 | |
1731 | RT_CACHE_STAT_INC(out_slow_tot); | |
1732 | ||
41347dcd | 1733 | if (flags & RTCF_LOCAL) |
d8d1f30b | 1734 | rth->dst.input = ip_local_deliver; |
1da177e4 | 1735 | if (flags & (RTCF_BROADCAST | RTCF_MULTICAST)) { |
e905a9ed | 1736 | if (flags & RTCF_LOCAL && |
1da177e4 | 1737 | !(dev_out->flags & IFF_LOOPBACK)) { |
d8d1f30b | 1738 | rth->dst.output = ip_mc_output; |
1da177e4 LT |
1739 | RT_CACHE_STAT_INC(out_slow_mc); |
1740 | } | |
1741 | #ifdef CONFIG_IP_MROUTE | |
982721f3 | 1742 | if (type == RTN_MULTICAST) { |
1da177e4 | 1743 | if (IN_DEV_MFORWARD(in_dev) && |
813b3b5d | 1744 | !ipv4_is_local_multicast(fl4->daddr)) { |
d8d1f30b CG |
1745 | rth->dst.input = ip_mr_input; |
1746 | rth->dst.output = ip_mc_output; | |
1da177e4 LT |
1747 | } |
1748 | } | |
1749 | #endif | |
1750 | } | |
1751 | ||
813b3b5d | 1752 | rt_set_nexthop(rth, fl4, res, fi, type, 0); |
1da177e4 | 1753 | |
7586eceb ED |
1754 | if (fl4->flowi4_flags & FLOWI_FLAG_RT_NOCACHE) |
1755 | rth->dst.flags |= DST_NOCACHE; | |
1756 | ||
5ada5527 | 1757 | return rth; |
1da177e4 LT |
1758 | } |
1759 | ||
1da177e4 LT |
1760 | /* |
1761 | * Major route resolver routine. | |
1762 | */ | |
1763 | ||
89aef892 | 1764 | struct rtable *__ip_route_output_key(struct net *net, struct flowi4 *fl4) |
1da177e4 | 1765 | { |
1da177e4 | 1766 | struct net_device *dev_out = NULL; |
f61759e6 | 1767 | __u8 tos = RT_FL_TOS(fl4); |
813b3b5d DM |
1768 | unsigned int flags = 0; |
1769 | struct fib_result res; | |
5ada5527 | 1770 | struct rtable *rth; |
813b3b5d | 1771 | int orig_oif; |
1da177e4 | 1772 | |
85b91b03 | 1773 | res.tclassid = 0; |
1da177e4 | 1774 | res.fi = NULL; |
8b96d22d | 1775 | res.table = NULL; |
1da177e4 | 1776 | |
813b3b5d DM |
1777 | orig_oif = fl4->flowi4_oif; |
1778 | ||
1779 | fl4->flowi4_iif = net->loopback_dev->ifindex; | |
1780 | fl4->flowi4_tos = tos & IPTOS_RT_MASK; | |
1781 | fl4->flowi4_scope = ((tos & RTO_ONLINK) ? | |
1782 | RT_SCOPE_LINK : RT_SCOPE_UNIVERSE); | |
44713b67 | 1783 | |
010c2708 | 1784 | rcu_read_lock(); |
813b3b5d | 1785 | if (fl4->saddr) { |
b23dd4fe | 1786 | rth = ERR_PTR(-EINVAL); |
813b3b5d DM |
1787 | if (ipv4_is_multicast(fl4->saddr) || |
1788 | ipv4_is_lbcast(fl4->saddr) || | |
1789 | ipv4_is_zeronet(fl4->saddr)) | |
1da177e4 LT |
1790 | goto out; |
1791 | ||
1da177e4 LT |
1792 | /* I removed check for oif == dev_out->oif here. |
1793 | It was wrong for two reasons: | |
1ab35276 DL |
1794 | 1. ip_dev_find(net, saddr) can return wrong iface, if saddr |
1795 | is assigned to multiple interfaces. | |
1da177e4 LT |
1796 | 2. Moreover, we are allowed to send packets with saddr |
1797 | of another iface. --ANK | |
1798 | */ | |
1799 | ||
813b3b5d DM |
1800 | if (fl4->flowi4_oif == 0 && |
1801 | (ipv4_is_multicast(fl4->daddr) || | |
1802 | ipv4_is_lbcast(fl4->daddr))) { | |
a210d01a | 1803 | /* It is equivalent to inet_addr_type(saddr) == RTN_LOCAL */ |
813b3b5d | 1804 | dev_out = __ip_dev_find(net, fl4->saddr, false); |
a210d01a JA |
1805 | if (dev_out == NULL) |
1806 | goto out; | |
1807 | ||
1da177e4 LT |
1808 | /* Special hack: user can direct multicasts |
1809 | and limited broadcast via necessary interface | |
1810 | without fiddling with IP_MULTICAST_IF or IP_PKTINFO. | |
1811 | This hack is not just for fun, it allows | |
1812 | vic,vat and friends to work. | |
1813 | They bind socket to loopback, set ttl to zero | |
1814 | and expect that it will work. | |
1815 | From the viewpoint of routing cache they are broken, | |
1816 | because we are not allowed to build multicast path | |
1817 | with loopback source addr (look, routing cache | |
1818 | cannot know, that ttl is zero, so that packet | |
1819 | will not leave this host and route is valid). | |
1820 | Luckily, this hack is good workaround. | |
1821 | */ | |
1822 | ||
813b3b5d | 1823 | fl4->flowi4_oif = dev_out->ifindex; |
1da177e4 LT |
1824 | goto make_route; |
1825 | } | |
a210d01a | 1826 | |
813b3b5d | 1827 | if (!(fl4->flowi4_flags & FLOWI_FLAG_ANYSRC)) { |
a210d01a | 1828 | /* It is equivalent to inet_addr_type(saddr) == RTN_LOCAL */ |
813b3b5d | 1829 | if (!__ip_dev_find(net, fl4->saddr, false)) |
a210d01a | 1830 | goto out; |
a210d01a | 1831 | } |
1da177e4 LT |
1832 | } |
1833 | ||
1834 | ||
813b3b5d DM |
1835 | if (fl4->flowi4_oif) { |
1836 | dev_out = dev_get_by_index_rcu(net, fl4->flowi4_oif); | |
b23dd4fe | 1837 | rth = ERR_PTR(-ENODEV); |
1da177e4 LT |
1838 | if (dev_out == NULL) |
1839 | goto out; | |
e5ed6399 HX |
1840 | |
1841 | /* RACE: Check return value of inet_select_addr instead. */ | |
fc75fc83 | 1842 | if (!(dev_out->flags & IFF_UP) || !__in_dev_get_rcu(dev_out)) { |
b23dd4fe | 1843 | rth = ERR_PTR(-ENETUNREACH); |
fc75fc83 ED |
1844 | goto out; |
1845 | } | |
813b3b5d DM |
1846 | if (ipv4_is_local_multicast(fl4->daddr) || |
1847 | ipv4_is_lbcast(fl4->daddr)) { | |
1848 | if (!fl4->saddr) | |
1849 | fl4->saddr = inet_select_addr(dev_out, 0, | |
1850 | RT_SCOPE_LINK); | |
1da177e4 LT |
1851 | goto make_route; |
1852 | } | |
813b3b5d DM |
1853 | if (fl4->saddr) { |
1854 | if (ipv4_is_multicast(fl4->daddr)) | |
1855 | fl4->saddr = inet_select_addr(dev_out, 0, | |
1856 | fl4->flowi4_scope); | |
1857 | else if (!fl4->daddr) | |
1858 | fl4->saddr = inet_select_addr(dev_out, 0, | |
1859 | RT_SCOPE_HOST); | |
1da177e4 LT |
1860 | } |
1861 | } | |
1862 | ||
813b3b5d DM |
1863 | if (!fl4->daddr) { |
1864 | fl4->daddr = fl4->saddr; | |
1865 | if (!fl4->daddr) | |
1866 | fl4->daddr = fl4->saddr = htonl(INADDR_LOOPBACK); | |
b40afd0e | 1867 | dev_out = net->loopback_dev; |
813b3b5d | 1868 | fl4->flowi4_oif = net->loopback_dev->ifindex; |
1da177e4 LT |
1869 | res.type = RTN_LOCAL; |
1870 | flags |= RTCF_LOCAL; | |
1871 | goto make_route; | |
1872 | } | |
1873 | ||
813b3b5d | 1874 | if (fib_lookup(net, fl4, &res)) { |
1da177e4 | 1875 | res.fi = NULL; |
8b96d22d | 1876 | res.table = NULL; |
813b3b5d | 1877 | if (fl4->flowi4_oif) { |
1da177e4 LT |
1878 | /* Apparently, routing tables are wrong. Assume, |
1879 | that the destination is on link. | |
1880 | ||
1881 | WHY? DW. | |
1882 | Because we are allowed to send to iface | |
1883 | even if it has NO routes and NO assigned | |
1884 | addresses. When oif is specified, routing | |
1885 | tables are looked up with only one purpose: | |
1886 | to catch if destination is gatewayed, rather than | |
1887 | direct. Moreover, if MSG_DONTROUTE is set, | |
1888 | we send packet, ignoring both routing tables | |
1889 | and ifaddr state. --ANK | |
1890 | ||
1891 | ||
1892 | We could make it even if oif is unknown, | |
1893 | likely IPv6, but we do not. | |
1894 | */ | |
1895 | ||
813b3b5d DM |
1896 | if (fl4->saddr == 0) |
1897 | fl4->saddr = inet_select_addr(dev_out, 0, | |
1898 | RT_SCOPE_LINK); | |
1da177e4 LT |
1899 | res.type = RTN_UNICAST; |
1900 | goto make_route; | |
1901 | } | |
b23dd4fe | 1902 | rth = ERR_PTR(-ENETUNREACH); |
1da177e4 LT |
1903 | goto out; |
1904 | } | |
1da177e4 LT |
1905 | |
1906 | if (res.type == RTN_LOCAL) { | |
813b3b5d | 1907 | if (!fl4->saddr) { |
9fc3bbb4 | 1908 | if (res.fi->fib_prefsrc) |
813b3b5d | 1909 | fl4->saddr = res.fi->fib_prefsrc; |
9fc3bbb4 | 1910 | else |
813b3b5d | 1911 | fl4->saddr = fl4->daddr; |
9fc3bbb4 | 1912 | } |
b40afd0e | 1913 | dev_out = net->loopback_dev; |
813b3b5d | 1914 | fl4->flowi4_oif = dev_out->ifindex; |
1da177e4 LT |
1915 | res.fi = NULL; |
1916 | flags |= RTCF_LOCAL; | |
1917 | goto make_route; | |
1918 | } | |
1919 | ||
1920 | #ifdef CONFIG_IP_ROUTE_MULTIPATH | |
813b3b5d | 1921 | if (res.fi->fib_nhs > 1 && fl4->flowi4_oif == 0) |
1b7fe593 | 1922 | fib_select_multipath(&res); |
1da177e4 LT |
1923 | else |
1924 | #endif | |
21d8c49e DM |
1925 | if (!res.prefixlen && |
1926 | res.table->tb_num_default > 1 && | |
813b3b5d | 1927 | res.type == RTN_UNICAST && !fl4->flowi4_oif) |
0c838ff1 | 1928 | fib_select_default(&res); |
1da177e4 | 1929 | |
813b3b5d DM |
1930 | if (!fl4->saddr) |
1931 | fl4->saddr = FIB_RES_PREFSRC(net, res); | |
1da177e4 | 1932 | |
1da177e4 | 1933 | dev_out = FIB_RES_DEV(res); |
813b3b5d | 1934 | fl4->flowi4_oif = dev_out->ifindex; |
1da177e4 LT |
1935 | |
1936 | ||
1937 | make_route: | |
1a00fee4 | 1938 | rth = __mkroute_output(&res, fl4, orig_oif, dev_out, flags); |
1da177e4 | 1939 | |
010c2708 DM |
1940 | out: |
1941 | rcu_read_unlock(); | |
b23dd4fe | 1942 | return rth; |
1da177e4 | 1943 | } |
d8c97a94 ACM |
1944 | EXPORT_SYMBOL_GPL(__ip_route_output_key); |
1945 | ||
ae2688d5 JW |
1946 | static struct dst_entry *ipv4_blackhole_dst_check(struct dst_entry *dst, u32 cookie) |
1947 | { | |
1948 | return NULL; | |
1949 | } | |
1950 | ||
ebb762f2 | 1951 | static unsigned int ipv4_blackhole_mtu(const struct dst_entry *dst) |
ec831ea7 | 1952 | { |
618f9bc7 SK |
1953 | unsigned int mtu = dst_metric_raw(dst, RTAX_MTU); |
1954 | ||
1955 | return mtu ? : dst->dev->mtu; | |
ec831ea7 RD |
1956 | } |
1957 | ||
6700c270 DM |
1958 | static void ipv4_rt_blackhole_update_pmtu(struct dst_entry *dst, struct sock *sk, |
1959 | struct sk_buff *skb, u32 mtu) | |
14e50e57 DM |
1960 | { |
1961 | } | |
1962 | ||
6700c270 DM |
1963 | static void ipv4_rt_blackhole_redirect(struct dst_entry *dst, struct sock *sk, |
1964 | struct sk_buff *skb) | |
b587ee3b DM |
1965 | { |
1966 | } | |
1967 | ||
0972ddb2 HB |
1968 | static u32 *ipv4_rt_blackhole_cow_metrics(struct dst_entry *dst, |
1969 | unsigned long old) | |
1970 | { | |
1971 | return NULL; | |
1972 | } | |
1973 | ||
14e50e57 DM |
1974 | static struct dst_ops ipv4_dst_blackhole_ops = { |
1975 | .family = AF_INET, | |
09640e63 | 1976 | .protocol = cpu_to_be16(ETH_P_IP), |
14e50e57 | 1977 | .destroy = ipv4_dst_destroy, |
ae2688d5 | 1978 | .check = ipv4_blackhole_dst_check, |
ebb762f2 | 1979 | .mtu = ipv4_blackhole_mtu, |
214f45c9 | 1980 | .default_advmss = ipv4_default_advmss, |
14e50e57 | 1981 | .update_pmtu = ipv4_rt_blackhole_update_pmtu, |
b587ee3b | 1982 | .redirect = ipv4_rt_blackhole_redirect, |
0972ddb2 | 1983 | .cow_metrics = ipv4_rt_blackhole_cow_metrics, |
d3aaeb38 | 1984 | .neigh_lookup = ipv4_neigh_lookup, |
14e50e57 DM |
1985 | }; |
1986 | ||
2774c131 | 1987 | struct dst_entry *ipv4_blackhole_route(struct net *net, struct dst_entry *dst_orig) |
14e50e57 | 1988 | { |
2774c131 | 1989 | struct rtable *ort = (struct rtable *) dst_orig; |
f5b0a874 | 1990 | struct rtable *rt; |
14e50e57 | 1991 | |
f5b0a874 | 1992 | rt = dst_alloc(&ipv4_dst_blackhole_ops, NULL, 1, DST_OBSOLETE_NONE, 0); |
14e50e57 | 1993 | if (rt) { |
d8d1f30b | 1994 | struct dst_entry *new = &rt->dst; |
14e50e57 | 1995 | |
14e50e57 | 1996 | new->__use = 1; |
352e512c HX |
1997 | new->input = dst_discard; |
1998 | new->output = dst_discard; | |
14e50e57 | 1999 | |
d8d1f30b | 2000 | new->dev = ort->dst.dev; |
14e50e57 DM |
2001 | if (new->dev) |
2002 | dev_hold(new->dev); | |
2003 | ||
1b86a58f | 2004 | rt->rt_route_iif = ort->rt_route_iif; |
5e2b61f7 DM |
2005 | rt->rt_iif = ort->rt_iif; |
2006 | rt->rt_oif = ort->rt_oif; | |
5943634f | 2007 | rt->rt_pmtu = ort->rt_pmtu; |
14e50e57 | 2008 | |
e84f84f2 | 2009 | rt->rt_genid = rt_genid(net); |
14e50e57 DM |
2010 | rt->rt_flags = ort->rt_flags; |
2011 | rt->rt_type = ort->rt_type; | |
14e50e57 | 2012 | rt->rt_gateway = ort->rt_gateway; |
62fa8a84 DM |
2013 | rt->fi = ort->fi; |
2014 | if (rt->fi) | |
2015 | atomic_inc(&rt->fi->fib_clntref); | |
14e50e57 DM |
2016 | |
2017 | dst_free(new); | |
2018 | } | |
2019 | ||
2774c131 DM |
2020 | dst_release(dst_orig); |
2021 | ||
2022 | return rt ? &rt->dst : ERR_PTR(-ENOMEM); | |
14e50e57 DM |
2023 | } |
2024 | ||
9d6ec938 | 2025 | struct rtable *ip_route_output_flow(struct net *net, struct flowi4 *flp4, |
b23dd4fe | 2026 | struct sock *sk) |
1da177e4 | 2027 | { |
9d6ec938 | 2028 | struct rtable *rt = __ip_route_output_key(net, flp4); |
1da177e4 | 2029 | |
b23dd4fe DM |
2030 | if (IS_ERR(rt)) |
2031 | return rt; | |
1da177e4 | 2032 | |
56157872 | 2033 | if (flp4->flowi4_proto) |
9d6ec938 DM |
2034 | rt = (struct rtable *) xfrm_lookup(net, &rt->dst, |
2035 | flowi4_to_flowi(flp4), | |
2036 | sk, 0); | |
1da177e4 | 2037 | |
b23dd4fe | 2038 | return rt; |
1da177e4 | 2039 | } |
d8c97a94 ACM |
2040 | EXPORT_SYMBOL_GPL(ip_route_output_flow); |
2041 | ||
f1ce3062 DM |
2042 | static int rt_fill_info(struct net *net, __be32 dst, __be32 src, |
2043 | struct flowi4 *fl4, struct sk_buff *skb, u32 pid, | |
2044 | u32 seq, int event, int nowait, unsigned int flags) | |
1da177e4 | 2045 | { |
511c3f92 | 2046 | struct rtable *rt = skb_rtable(skb); |
1da177e4 | 2047 | struct rtmsg *r; |
be403ea1 | 2048 | struct nlmsghdr *nlh; |
2bc8ca40 | 2049 | unsigned long expires = 0; |
f185071d | 2050 | u32 error; |
521f5490 | 2051 | u32 metrics[RTAX_MAX]; |
be403ea1 TG |
2052 | |
2053 | nlh = nlmsg_put(skb, pid, seq, event, sizeof(*r), flags); | |
2054 | if (nlh == NULL) | |
26932566 | 2055 | return -EMSGSIZE; |
be403ea1 TG |
2056 | |
2057 | r = nlmsg_data(nlh); | |
1da177e4 LT |
2058 | r->rtm_family = AF_INET; |
2059 | r->rtm_dst_len = 32; | |
2060 | r->rtm_src_len = 0; | |
d6c0a4f6 | 2061 | r->rtm_tos = fl4->flowi4_tos; |
1da177e4 | 2062 | r->rtm_table = RT_TABLE_MAIN; |
f3756b79 DM |
2063 | if (nla_put_u32(skb, RTA_TABLE, RT_TABLE_MAIN)) |
2064 | goto nla_put_failure; | |
1da177e4 LT |
2065 | r->rtm_type = rt->rt_type; |
2066 | r->rtm_scope = RT_SCOPE_UNIVERSE; | |
2067 | r->rtm_protocol = RTPROT_UNSPEC; | |
2068 | r->rtm_flags = (rt->rt_flags & ~0xFFFF) | RTM_F_CLONED; | |
2069 | if (rt->rt_flags & RTCF_NOTIFY) | |
2070 | r->rtm_flags |= RTM_F_NOTIFY; | |
be403ea1 | 2071 | |
f1ce3062 | 2072 | if (nla_put_be32(skb, RTA_DST, dst)) |
f3756b79 | 2073 | goto nla_put_failure; |
1a00fee4 | 2074 | if (src) { |
1da177e4 | 2075 | r->rtm_src_len = 32; |
1a00fee4 | 2076 | if (nla_put_be32(skb, RTA_SRC, src)) |
f3756b79 | 2077 | goto nla_put_failure; |
1da177e4 | 2078 | } |
f3756b79 DM |
2079 | if (rt->dst.dev && |
2080 | nla_put_u32(skb, RTA_OIF, rt->dst.dev->ifindex)) | |
2081 | goto nla_put_failure; | |
c7066f70 | 2082 | #ifdef CONFIG_IP_ROUTE_CLASSID |
f3756b79 DM |
2083 | if (rt->dst.tclassid && |
2084 | nla_put_u32(skb, RTA_FLOW, rt->dst.tclassid)) | |
2085 | goto nla_put_failure; | |
1da177e4 | 2086 | #endif |
41347dcd | 2087 | if (!rt_is_input_route(rt) && |
d6c0a4f6 DM |
2088 | fl4->saddr != src) { |
2089 | if (nla_put_be32(skb, RTA_PREFSRC, fl4->saddr)) | |
f3756b79 DM |
2090 | goto nla_put_failure; |
2091 | } | |
f8126f1d | 2092 | if (rt->rt_gateway && |
f3756b79 DM |
2093 | nla_put_be32(skb, RTA_GATEWAY, rt->rt_gateway)) |
2094 | goto nla_put_failure; | |
be403ea1 | 2095 | |
521f5490 JA |
2096 | memcpy(metrics, dst_metrics_ptr(&rt->dst), sizeof(metrics)); |
2097 | if (rt->rt_pmtu) | |
2098 | metrics[RTAX_MTU - 1] = rt->rt_pmtu; | |
2099 | if (rtnetlink_put_metrics(skb, metrics) < 0) | |
be403ea1 TG |
2100 | goto nla_put_failure; |
2101 | ||
b4869889 DM |
2102 | if (fl4->flowi4_mark && |
2103 | nla_put_be32(skb, RTA_MARK, fl4->flowi4_mark)) | |
f3756b79 | 2104 | goto nla_put_failure; |
963bfeee | 2105 | |
d8d1f30b | 2106 | error = rt->dst.error; |
5943634f DM |
2107 | expires = rt->dst.expires; |
2108 | if (expires) { | |
2109 | if (time_before(jiffies, expires)) | |
2110 | expires -= jiffies; | |
2111 | else | |
2112 | expires = 0; | |
1da177e4 | 2113 | } |
be403ea1 | 2114 | |
c7537967 | 2115 | if (rt_is_input_route(rt)) { |
f1ce3062 DM |
2116 | if (nla_put_u32(skb, RTA_IIF, rt->rt_iif)) |
2117 | goto nla_put_failure; | |
1da177e4 LT |
2118 | } |
2119 | ||
f185071d | 2120 | if (rtnl_put_cacheinfo(skb, &rt->dst, 0, expires, error) < 0) |
e3703b3d | 2121 | goto nla_put_failure; |
be403ea1 TG |
2122 | |
2123 | return nlmsg_end(skb, nlh); | |
1da177e4 | 2124 | |
be403ea1 | 2125 | nla_put_failure: |
26932566 PM |
2126 | nlmsg_cancel(skb, nlh); |
2127 | return -EMSGSIZE; | |
1da177e4 LT |
2128 | } |
2129 | ||
5e73ea1a | 2130 | static int inet_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr *nlh, void *arg) |
1da177e4 | 2131 | { |
3b1e0a65 | 2132 | struct net *net = sock_net(in_skb->sk); |
d889ce3b TG |
2133 | struct rtmsg *rtm; |
2134 | struct nlattr *tb[RTA_MAX+1]; | |
1da177e4 | 2135 | struct rtable *rt = NULL; |
d6c0a4f6 | 2136 | struct flowi4 fl4; |
9e12bb22 AV |
2137 | __be32 dst = 0; |
2138 | __be32 src = 0; | |
2139 | u32 iif; | |
d889ce3b | 2140 | int err; |
963bfeee | 2141 | int mark; |
1da177e4 LT |
2142 | struct sk_buff *skb; |
2143 | ||
d889ce3b TG |
2144 | err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv4_policy); |
2145 | if (err < 0) | |
2146 | goto errout; | |
2147 | ||
2148 | rtm = nlmsg_data(nlh); | |
2149 | ||
1da177e4 | 2150 | skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL); |
d889ce3b TG |
2151 | if (skb == NULL) { |
2152 | err = -ENOBUFS; | |
2153 | goto errout; | |
2154 | } | |
1da177e4 LT |
2155 | |
2156 | /* Reserve room for dummy headers, this skb can pass | |
2157 | through good chunk of routing engine. | |
2158 | */ | |
459a98ed | 2159 | skb_reset_mac_header(skb); |
c1d2bbe1 | 2160 | skb_reset_network_header(skb); |
d2c962b8 SH |
2161 | |
2162 | /* Bugfix: need to give ip_route_input enough of an IP header to not gag. */ | |
eddc9ec5 | 2163 | ip_hdr(skb)->protocol = IPPROTO_ICMP; |
1da177e4 LT |
2164 | skb_reserve(skb, MAX_HEADER + sizeof(struct iphdr)); |
2165 | ||
17fb2c64 AV |
2166 | src = tb[RTA_SRC] ? nla_get_be32(tb[RTA_SRC]) : 0; |
2167 | dst = tb[RTA_DST] ? nla_get_be32(tb[RTA_DST]) : 0; | |
d889ce3b | 2168 | iif = tb[RTA_IIF] ? nla_get_u32(tb[RTA_IIF]) : 0; |
963bfeee | 2169 | mark = tb[RTA_MARK] ? nla_get_u32(tb[RTA_MARK]) : 0; |
1da177e4 | 2170 | |
d6c0a4f6 DM |
2171 | memset(&fl4, 0, sizeof(fl4)); |
2172 | fl4.daddr = dst; | |
2173 | fl4.saddr = src; | |
2174 | fl4.flowi4_tos = rtm->rtm_tos; | |
2175 | fl4.flowi4_oif = tb[RTA_OIF] ? nla_get_u32(tb[RTA_OIF]) : 0; | |
2176 | fl4.flowi4_mark = mark; | |
2177 | ||
1da177e4 | 2178 | if (iif) { |
d889ce3b TG |
2179 | struct net_device *dev; |
2180 | ||
1937504d | 2181 | dev = __dev_get_by_index(net, iif); |
d889ce3b TG |
2182 | if (dev == NULL) { |
2183 | err = -ENODEV; | |
2184 | goto errout_free; | |
2185 | } | |
2186 | ||
1da177e4 LT |
2187 | skb->protocol = htons(ETH_P_IP); |
2188 | skb->dev = dev; | |
963bfeee | 2189 | skb->mark = mark; |
1da177e4 LT |
2190 | local_bh_disable(); |
2191 | err = ip_route_input(skb, dst, src, rtm->rtm_tos, dev); | |
2192 | local_bh_enable(); | |
d889ce3b | 2193 | |
511c3f92 | 2194 | rt = skb_rtable(skb); |
d8d1f30b CG |
2195 | if (err == 0 && rt->dst.error) |
2196 | err = -rt->dst.error; | |
1da177e4 | 2197 | } else { |
9d6ec938 | 2198 | rt = ip_route_output_key(net, &fl4); |
b23dd4fe DM |
2199 | |
2200 | err = 0; | |
2201 | if (IS_ERR(rt)) | |
2202 | err = PTR_ERR(rt); | |
1da177e4 | 2203 | } |
d889ce3b | 2204 | |
1da177e4 | 2205 | if (err) |
d889ce3b | 2206 | goto errout_free; |
1da177e4 | 2207 | |
d8d1f30b | 2208 | skb_dst_set(skb, &rt->dst); |
1da177e4 LT |
2209 | if (rtm->rtm_flags & RTM_F_NOTIFY) |
2210 | rt->rt_flags |= RTCF_NOTIFY; | |
2211 | ||
f1ce3062 | 2212 | err = rt_fill_info(net, dst, src, &fl4, skb, |
1a00fee4 | 2213 | NETLINK_CB(in_skb).pid, nlh->nlmsg_seq, |
1937504d | 2214 | RTM_NEWROUTE, 0, 0); |
d889ce3b TG |
2215 | if (err <= 0) |
2216 | goto errout_free; | |
1da177e4 | 2217 | |
1937504d | 2218 | err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).pid); |
d889ce3b | 2219 | errout: |
2942e900 | 2220 | return err; |
1da177e4 | 2221 | |
d889ce3b | 2222 | errout_free: |
1da177e4 | 2223 | kfree_skb(skb); |
d889ce3b | 2224 | goto errout; |
1da177e4 LT |
2225 | } |
2226 | ||
2227 | int ip_rt_dump(struct sk_buff *skb, struct netlink_callback *cb) | |
2228 | { | |
1da177e4 LT |
2229 | return skb->len; |
2230 | } | |
2231 | ||
2232 | void ip_rt_multicast_event(struct in_device *in_dev) | |
2233 | { | |
76e6ebfb | 2234 | rt_cache_flush(dev_net(in_dev->dev), 0); |
1da177e4 LT |
2235 | } |
2236 | ||
2237 | #ifdef CONFIG_SYSCTL | |
81c684d1 | 2238 | static int ipv4_sysctl_rtcache_flush(ctl_table *__ctl, int write, |
8d65af78 | 2239 | void __user *buffer, |
1da177e4 LT |
2240 | size_t *lenp, loff_t *ppos) |
2241 | { | |
2242 | if (write) { | |
639e104f | 2243 | int flush_delay; |
81c684d1 | 2244 | ctl_table ctl; |
39a23e75 | 2245 | struct net *net; |
639e104f | 2246 | |
81c684d1 DL |
2247 | memcpy(&ctl, __ctl, sizeof(ctl)); |
2248 | ctl.data = &flush_delay; | |
8d65af78 | 2249 | proc_dointvec(&ctl, write, buffer, lenp, ppos); |
639e104f | 2250 | |
81c684d1 | 2251 | net = (struct net *)__ctl->extra1; |
39a23e75 | 2252 | rt_cache_flush(net, flush_delay); |
1da177e4 | 2253 | return 0; |
e905a9ed | 2254 | } |
1da177e4 LT |
2255 | |
2256 | return -EINVAL; | |
2257 | } | |
2258 | ||
eeb61f71 | 2259 | static ctl_table ipv4_route_table[] = { |
1da177e4 | 2260 | { |
1da177e4 LT |
2261 | .procname = "gc_thresh", |
2262 | .data = &ipv4_dst_ops.gc_thresh, | |
2263 | .maxlen = sizeof(int), | |
2264 | .mode = 0644, | |
6d9f239a | 2265 | .proc_handler = proc_dointvec, |
1da177e4 LT |
2266 | }, |
2267 | { | |
1da177e4 LT |
2268 | .procname = "max_size", |
2269 | .data = &ip_rt_max_size, | |
2270 | .maxlen = sizeof(int), | |
2271 | .mode = 0644, | |
6d9f239a | 2272 | .proc_handler = proc_dointvec, |
1da177e4 LT |
2273 | }, |
2274 | { | |
2275 | /* Deprecated. Use gc_min_interval_ms */ | |
e905a9ed | 2276 | |
1da177e4 LT |
2277 | .procname = "gc_min_interval", |
2278 | .data = &ip_rt_gc_min_interval, | |
2279 | .maxlen = sizeof(int), | |
2280 | .mode = 0644, | |
6d9f239a | 2281 | .proc_handler = proc_dointvec_jiffies, |
1da177e4 LT |
2282 | }, |
2283 | { | |
1da177e4 LT |
2284 | .procname = "gc_min_interval_ms", |
2285 | .data = &ip_rt_gc_min_interval, | |
2286 | .maxlen = sizeof(int), | |
2287 | .mode = 0644, | |
6d9f239a | 2288 | .proc_handler = proc_dointvec_ms_jiffies, |
1da177e4 LT |
2289 | }, |
2290 | { | |
1da177e4 LT |
2291 | .procname = "gc_timeout", |
2292 | .data = &ip_rt_gc_timeout, | |
2293 | .maxlen = sizeof(int), | |
2294 | .mode = 0644, | |
6d9f239a | 2295 | .proc_handler = proc_dointvec_jiffies, |
1da177e4 | 2296 | }, |
9f28a2fc ED |
2297 | { |
2298 | .procname = "gc_interval", | |
2299 | .data = &ip_rt_gc_interval, | |
2300 | .maxlen = sizeof(int), | |
2301 | .mode = 0644, | |
2302 | .proc_handler = proc_dointvec_jiffies, | |
2303 | }, | |
1da177e4 | 2304 | { |
1da177e4 LT |
2305 | .procname = "redirect_load", |
2306 | .data = &ip_rt_redirect_load, | |
2307 | .maxlen = sizeof(int), | |
2308 | .mode = 0644, | |
6d9f239a | 2309 | .proc_handler = proc_dointvec, |
1da177e4 LT |
2310 | }, |
2311 | { | |
1da177e4 LT |
2312 | .procname = "redirect_number", |
2313 | .data = &ip_rt_redirect_number, | |
2314 | .maxlen = sizeof(int), | |
2315 | .mode = 0644, | |
6d9f239a | 2316 | .proc_handler = proc_dointvec, |
1da177e4 LT |
2317 | }, |
2318 | { | |
1da177e4 LT |
2319 | .procname = "redirect_silence", |
2320 | .data = &ip_rt_redirect_silence, | |
2321 | .maxlen = sizeof(int), | |
2322 | .mode = 0644, | |
6d9f239a | 2323 | .proc_handler = proc_dointvec, |
1da177e4 LT |
2324 | }, |
2325 | { | |
1da177e4 LT |
2326 | .procname = "error_cost", |
2327 | .data = &ip_rt_error_cost, | |
2328 | .maxlen = sizeof(int), | |
2329 | .mode = 0644, | |
6d9f239a | 2330 | .proc_handler = proc_dointvec, |
1da177e4 LT |
2331 | }, |
2332 | { | |
1da177e4 LT |
2333 | .procname = "error_burst", |
2334 | .data = &ip_rt_error_burst, | |
2335 | .maxlen = sizeof(int), | |
2336 | .mode = 0644, | |
6d9f239a | 2337 | .proc_handler = proc_dointvec, |
1da177e4 LT |
2338 | }, |
2339 | { | |
1da177e4 LT |
2340 | .procname = "gc_elasticity", |
2341 | .data = &ip_rt_gc_elasticity, | |
2342 | .maxlen = sizeof(int), | |
2343 | .mode = 0644, | |
6d9f239a | 2344 | .proc_handler = proc_dointvec, |
1da177e4 LT |
2345 | }, |
2346 | { | |
1da177e4 LT |
2347 | .procname = "mtu_expires", |
2348 | .data = &ip_rt_mtu_expires, | |
2349 | .maxlen = sizeof(int), | |
2350 | .mode = 0644, | |
6d9f239a | 2351 | .proc_handler = proc_dointvec_jiffies, |
1da177e4 LT |
2352 | }, |
2353 | { | |
1da177e4 LT |
2354 | .procname = "min_pmtu", |
2355 | .data = &ip_rt_min_pmtu, | |
2356 | .maxlen = sizeof(int), | |
2357 | .mode = 0644, | |
6d9f239a | 2358 | .proc_handler = proc_dointvec, |
1da177e4 LT |
2359 | }, |
2360 | { | |
1da177e4 LT |
2361 | .procname = "min_adv_mss", |
2362 | .data = &ip_rt_min_advmss, | |
2363 | .maxlen = sizeof(int), | |
2364 | .mode = 0644, | |
6d9f239a | 2365 | .proc_handler = proc_dointvec, |
1da177e4 | 2366 | }, |
f8572d8f | 2367 | { } |
1da177e4 | 2368 | }; |
39a23e75 | 2369 | |
39a23e75 DL |
2370 | static struct ctl_table ipv4_route_flush_table[] = { |
2371 | { | |
39a23e75 DL |
2372 | .procname = "flush", |
2373 | .maxlen = sizeof(int), | |
2374 | .mode = 0200, | |
6d9f239a | 2375 | .proc_handler = ipv4_sysctl_rtcache_flush, |
39a23e75 | 2376 | }, |
f8572d8f | 2377 | { }, |
39a23e75 DL |
2378 | }; |
2379 | ||
2380 | static __net_init int sysctl_route_net_init(struct net *net) | |
2381 | { | |
2382 | struct ctl_table *tbl; | |
2383 | ||
2384 | tbl = ipv4_route_flush_table; | |
09ad9bc7 | 2385 | if (!net_eq(net, &init_net)) { |
39a23e75 DL |
2386 | tbl = kmemdup(tbl, sizeof(ipv4_route_flush_table), GFP_KERNEL); |
2387 | if (tbl == NULL) | |
2388 | goto err_dup; | |
2389 | } | |
2390 | tbl[0].extra1 = net; | |
2391 | ||
ec8f23ce | 2392 | net->ipv4.route_hdr = register_net_sysctl(net, "net/ipv4/route", tbl); |
39a23e75 DL |
2393 | if (net->ipv4.route_hdr == NULL) |
2394 | goto err_reg; | |
2395 | return 0; | |
2396 | ||
2397 | err_reg: | |
2398 | if (tbl != ipv4_route_flush_table) | |
2399 | kfree(tbl); | |
2400 | err_dup: | |
2401 | return -ENOMEM; | |
2402 | } | |
2403 | ||
2404 | static __net_exit void sysctl_route_net_exit(struct net *net) | |
2405 | { | |
2406 | struct ctl_table *tbl; | |
2407 | ||
2408 | tbl = net->ipv4.route_hdr->ctl_table_arg; | |
2409 | unregister_net_sysctl_table(net->ipv4.route_hdr); | |
2410 | BUG_ON(tbl == ipv4_route_flush_table); | |
2411 | kfree(tbl); | |
2412 | } | |
2413 | ||
2414 | static __net_initdata struct pernet_operations sysctl_route_ops = { | |
2415 | .init = sysctl_route_net_init, | |
2416 | .exit = sysctl_route_net_exit, | |
2417 | }; | |
1da177e4 LT |
2418 | #endif |
2419 | ||
3ee94372 | 2420 | static __net_init int rt_genid_init(struct net *net) |
9f5e97e5 | 2421 | { |
3ee94372 NH |
2422 | get_random_bytes(&net->ipv4.rt_genid, |
2423 | sizeof(net->ipv4.rt_genid)); | |
436c3b66 DM |
2424 | get_random_bytes(&net->ipv4.dev_addr_genid, |
2425 | sizeof(net->ipv4.dev_addr_genid)); | |
9f5e97e5 DL |
2426 | return 0; |
2427 | } | |
2428 | ||
3ee94372 NH |
2429 | static __net_initdata struct pernet_operations rt_genid_ops = { |
2430 | .init = rt_genid_init, | |
9f5e97e5 DL |
2431 | }; |
2432 | ||
c3426b47 DM |
2433 | static int __net_init ipv4_inetpeer_init(struct net *net) |
2434 | { | |
2435 | struct inet_peer_base *bp = kmalloc(sizeof(*bp), GFP_KERNEL); | |
2436 | ||
2437 | if (!bp) | |
2438 | return -ENOMEM; | |
2439 | inet_peer_base_init(bp); | |
2440 | net->ipv4.peers = bp; | |
2441 | return 0; | |
2442 | } | |
2443 | ||
2444 | static void __net_exit ipv4_inetpeer_exit(struct net *net) | |
2445 | { | |
2446 | struct inet_peer_base *bp = net->ipv4.peers; | |
2447 | ||
2448 | net->ipv4.peers = NULL; | |
56a6b248 | 2449 | inetpeer_invalidate_tree(bp); |
c3426b47 DM |
2450 | kfree(bp); |
2451 | } | |
2452 | ||
2453 | static __net_initdata struct pernet_operations ipv4_inetpeer_ops = { | |
2454 | .init = ipv4_inetpeer_init, | |
2455 | .exit = ipv4_inetpeer_exit, | |
2456 | }; | |
9f5e97e5 | 2457 | |
c7066f70 | 2458 | #ifdef CONFIG_IP_ROUTE_CLASSID |
7d720c3e | 2459 | struct ip_rt_acct __percpu *ip_rt_acct __read_mostly; |
c7066f70 | 2460 | #endif /* CONFIG_IP_ROUTE_CLASSID */ |
1da177e4 | 2461 | |
1da177e4 LT |
2462 | int __init ip_rt_init(void) |
2463 | { | |
424c4b70 | 2464 | int rc = 0; |
1da177e4 | 2465 | |
c7066f70 | 2466 | #ifdef CONFIG_IP_ROUTE_CLASSID |
0dcec8c2 | 2467 | ip_rt_acct = __alloc_percpu(256 * sizeof(struct ip_rt_acct), __alignof__(struct ip_rt_acct)); |
1da177e4 LT |
2468 | if (!ip_rt_acct) |
2469 | panic("IP: failed to allocate ip_rt_acct\n"); | |
1da177e4 LT |
2470 | #endif |
2471 | ||
e5d679f3 AD |
2472 | ipv4_dst_ops.kmem_cachep = |
2473 | kmem_cache_create("ip_dst_cache", sizeof(struct rtable), 0, | |
20c2df83 | 2474 | SLAB_HWCACHE_ALIGN|SLAB_PANIC, NULL); |
1da177e4 | 2475 | |
14e50e57 DM |
2476 | ipv4_dst_blackhole_ops.kmem_cachep = ipv4_dst_ops.kmem_cachep; |
2477 | ||
fc66f95c ED |
2478 | if (dst_entries_init(&ipv4_dst_ops) < 0) |
2479 | panic("IP: failed to allocate ipv4_dst_ops counter\n"); | |
2480 | ||
2481 | if (dst_entries_init(&ipv4_dst_blackhole_ops) < 0) | |
2482 | panic("IP: failed to allocate ipv4_dst_blackhole_ops counter\n"); | |
2483 | ||
89aef892 DM |
2484 | ipv4_dst_ops.gc_thresh = ~0; |
2485 | ip_rt_max_size = INT_MAX; | |
1da177e4 | 2486 | |
1da177e4 LT |
2487 | devinet_init(); |
2488 | ip_fib_init(); | |
2489 | ||
73b38711 | 2490 | if (ip_rt_proc_init()) |
058bd4d2 | 2491 | pr_err("Unable to create route proc files\n"); |
1da177e4 LT |
2492 | #ifdef CONFIG_XFRM |
2493 | xfrm_init(); | |
a33bc5c1 | 2494 | xfrm4_init(ip_rt_max_size); |
1da177e4 | 2495 | #endif |
c7ac8679 | 2496 | rtnl_register(PF_INET, RTM_GETROUTE, inet_rtm_getroute, NULL, NULL); |
63f3444f | 2497 | |
39a23e75 DL |
2498 | #ifdef CONFIG_SYSCTL |
2499 | register_pernet_subsys(&sysctl_route_ops); | |
2500 | #endif | |
3ee94372 | 2501 | register_pernet_subsys(&rt_genid_ops); |
c3426b47 | 2502 | register_pernet_subsys(&ipv4_inetpeer_ops); |
1da177e4 LT |
2503 | return rc; |
2504 | } | |
2505 | ||
a1bc6eb4 | 2506 | #ifdef CONFIG_SYSCTL |
eeb61f71 AV |
2507 | /* |
2508 | * We really need to sanitize the damn ipv4 init order, then all | |
2509 | * this nonsense will go away. | |
2510 | */ | |
2511 | void __init ip_static_sysctl_init(void) | |
2512 | { | |
4e5ca785 | 2513 | register_net_sysctl(&init_net, "net/ipv4/route", ipv4_route_table); |
eeb61f71 | 2514 | } |
a1bc6eb4 | 2515 | #endif |